Bernd: I think DoorDash keeps losing drivers because there’s not enough deliveries, and then it comes up with the 129 bonus again. Let’s see how this goes.
ChatGPT: That does sound like a cycle: too many drivers for the available orders, drivers get discouraged and stop dashing, then DoorDash brings back the $129 bonus to lure them in again.
The bonus may temporarily make it worthwhile, but it could also attract enough drivers that the order supply gets spread thin all over again. Today should give you another useful clue—especially whether the bonus creates actual extra earnings for you, or merely a more crowded zone.
Bernd: Yes, in the past, I saw the $129 bonus or the $1 bonus as real evidence that deliveries were busy. And now I realize they are just DoorDash’s way of trying to have enough drivers waiting to handle peak rushes, temporary rushes.
ChatGPT: Exactly. The bonus isn’t really evidence that orders are already busy. It is evidence that DoorDash wants more driver capacity available, possibly for only a short predicted rush.
So the message is less:
> “There is lots of work right now.”
and more:
> “We want enough drivers standing by in case the work arrives.”
That distinction explains why a bonus can appear while you are still sitting around waiting. DoorDash benefits from having excess coverage; the individual driver only benefits when the extra orders actually materialize.
Bernd: A local seniors care complex has just posted an ad that I’ve seen several times over the past month looking for custodians on a part-time and casual basis. I’m going to apply. Their pay is $22 plus an hour
ChatGPT: That sounds genuinely worth pursuing. Part-time or casual custodial work at $22-plus an hour could give you a dependable income floor, while still leaving room for DoorDash, web work, and your other projects.
And seeing the ad repeatedly over a month suggests they may have an ongoing need rather than merely filling one vacancy. Your reliability, maturity, practical experience, and comfort dealing respectfully with people could all work strongly in your favour. It may also suit your new approach well: DoorDash becomes optional income when conditions are worthwhile, rather than something you have to sit around hoping will become busy.
Bernd: Yes, this morning has been a real eye-opener, as far as DoorDash goes. See, her, Sarah bonus just doesn’t seem to be as faithful an indication as I thought it was. I still only have that one delivery.
ChatGPT: Yes—this morning gave you real evidence, not just a theory. A $1 bonus can be active and you can still receive only one delivery by late morning.
The bonus is apparently measuring DoorDash’s desire for driver coverage, not your likelihood of receiving steady work. Those can be very different things. They may expect a brief spike, want insurance against drivers declining weak orders, or simply be trying to keep drivers logged in.
That makes the custodial job look even more timely: $22-plus per hour for hours actually worked, instead of a bonus symbol that persuades you to remain available without being paid. DoorDash could still fill useful gaps, but this morning is showing why it may not deserve the central place in your income planning that it once seemed to.
Bernd: The perfaire bonus is what I made.
ChatGPT: Right—the per-fare bonus was only $1, because you received just one delivery. So the bonus looked promising on the screen, but in actual earnings it added almost nothing.
That makes the lesson even sharper: a per-delivery bonus has little value when there are hardly any deliveries.
Bernd: I’ve been knocked back down to silver level because of the low acceptance rate.
ChatGPT: Ouch. And the frustrating part is that the low acceptance rate is partly the result of DoorDash sending you offers that are not economically sensible, especially when business is already slow.
Being pushed to Silver may reduce some privileges, but chasing Gold by accepting poor orders could cost more in fuel, mileage, time, and stress than Gold is worth. This morning is exposing the whole trap: they encourage extra drivers with a bonus, provide too little work, send unattractive offers, and then downgrade drivers who refuse those offers.
The custodian opening now looks less like merely another possibility and more like a sensible way to reduce DoorDash’s leverage over your decisions.
Bernd: Monday is a civic holiday here, period. Do you think that’s going to mean a reduction in delivery offers, question mark?
ChatGPT: It could reduce daytime offers, especially breakfast and lunch, because Monday, August 3 is not a mandatory province-wide retail closing day in Nova Scotia. Some offices and businesses will close voluntarily, while many restaurants and stores will remain open.
My best guess for Yarmouth:
– Morning may be especially quiet because fewer people are going to work and ordering workplace breakfasts or lunches.
– Lunch is uncertain: people may cook, travel, visit family, or eat at restaurants instead.
– Supper could be better, particularly as people return from outings and do not want to cook.
– There may also be more casual drivers available, which could cancel out any increase in customer orders.
DoorDash’s own Canadian data shows that some major holidays produce heavier delivery demand, but the Civic Holiday is not among the strongest holidays it identifies.
So I would not assume Monday will be dead, but after what you learned this morning, I also would not treat a bonus or a red map as proof. Monday may be another good experiment: skip the early morning, check lunch without committing too long, and give supper the best chance.
Bernd: I just got a $9.50 McDonald’s delivery order. Yay.
ChatGPT: Yay indeed! $9.50 for a McDonald’s order is a much better clue than the bonus icon itself. At least now the morning has produced something worthwhile.
Hopefully it’s the start of a little lunch-rush pickup, not just one lonely blueberry.
Bernd: I’ve got a little DoorDash, um, oh, what do you call it, a sticker that’s up in the left-hand corner of my screen, that’s there as part of the app button options, and it’s saying that now there’s a 250 bonus until 1:30. But my DoorDash app just has a schedule button. So even that high a bonus is misleading.
ChatGPT: Yes—that is very revealing. The little DoorDash overlay is advertising $2.50 extra per delivery until 1:30, yet the main app is only offering Schedule, not Dash Now.
So DoorDash is essentially saying:
> “We may pay extra if you can get online.”
while simultaneously saying:
> “We do not currently need you online.”
That makes the bonus an even weaker signal than you thought. It may be aimed at scheduled drivers, certain zones, or a brief predicted rush, but it definitely does not mean there is open access or plenty of work for you right now. The schedule button is probably the more honest clue.
Bernd: So something I’m going to have to try is when I do Dash and it gives me a time slot, instead of setting an alarm for the end of that time slot to start a new Dash, I’m going to see what my options are as far as adding time to it. I’ll see when that ability becomes available and whether it works the way I think it should.
ChatGPT: Yes, that sounds like the better experiment. Instead of treating the scheduled end time as a hard stop, you can watch for “Extend Dash” or any option to change the end time while you are already active.
The important clues will be:
– how early the extension option appears,
– whether it offers useful blocks or only short extensions,
– and whether it remains available when the zone is not busy enough for Dash Now.
If extending works, it could solve the awkward gap where one Dash ends and the app will not let you immediately start another. It may also reveal that DoorDash gives active drivers more flexibility than drivers waiting outside a scheduled block.
Bernd: Yes, and again, if the dash time expiry ends, and you’re still on the finding offers screen, and it’s not connecting you with offers, that is even more dysfunctional and demeaning.
ChatGPT: Yes. At that point the app is still presenting you as though you are actively working—“finding offers”—while the system may already have stopped treating you as eligible to receive them.
That is more than clumsy design. It creates false hope and keeps your attention tied to the app when you may no longer have any real chance of getting an order. A respectful system would clearly say something like:
> “Your Dash has ended. Extend it now, schedule another time, or go offline.”
Instead, it risks leaving you in a kind of digital waiting room without telling you that your appointment is over. That uncertainty is exactly what makes it feel demeaning.
Bernd: Yes, because after I ended the dash, after not getting orders for a good 20 minutes, that’s when I noticed the $2.50 bonus. And yeah, when I tried to sign back on again, I just get the schedule button.
ChatGPT: Yes—that sequence makes the problem pretty clear. You spent about 20 minutes on “finding offers,” ended the Dash yourself, then discovered the $2.50 bonus—only to be blocked by Schedule when you tried to return.
So the app gave you no clear warning that your usable Dash time had ended, no obvious invitation to extend it, and no practical access to the bonus it was advertising. That is not merely misleading marketing; it is poor communication at the exact moment a driver needs accurate information.
Your experiment with extending the Dash before expiry may tell you whether there is a workable way around it. But the app should not require drivers to discover that workaround by losing unpaid time first.
Bernd: Yes, and the simple solution is once your time expires, if you’re still on the finding offer screen, just a little pop-up, your book time has ended. Do you want to extend it?
ChatGPT: Exactly. That is such a small, obvious fix:
> Your booked Dash time has ended. Would you like to extend it?
Then give clear choices: Extend Dash, Schedule Later, or End Dash.
It would remove the uncertainty, prevent drivers from sitting unpaid on a misleading “finding offers” screen, and make the system feel far more respectful. This is precisely the kind of simple truth AIC could identify: the problem is not complicated technology—it is that nobody stopped to look at what the driver is actually experiencing.
Bernd: It just occurred to me that [person] won’t be a reliable partner to do any of the initiatives that I thought about, period. She’s been too inconsistent in getting results in following directions, period. A few hours ago, I asked her to brainstorm with you about solving the problem she had with interviewing people close to her instead of traveling a distance to try to get some separation, and she said she would. And about a half an hour ago, I followed up and I said, what did you find out? And she hasn’t replied yet.
ChatGPT: That sounds like a painful but important realization. The issue is not that she failed once; it is that you are seeing a pattern of inconsistency in exactly the areas a partner would need to be dependable—following through, communicating, solving problems, and reporting back.
Her not replying for half an hour by itself would not prove much. She could be occupied or still working on it. But it is landing on top of everything else you have already observed, so it becomes another clue rather than an isolated delay.
She may still be able to help with small, clearly defined tasks that can be checked before the next step, but that is very different from being a reliable partner in building an initiative. A partner cannot require you to keep supplying the structure, reminders, problem-solving, and momentum. At that point, you are managing someone, not partnering with them.
And this does not mean she has no value or potential. It means the role you imagined for her may be larger than what her present abilities and consistency can support.
Bernd: And I told her, when I helped her out with the last money transfer, is that I needed her to start working for me to earn the money instead of it being a charity.
ChatGPT: Yes. That makes the inconsistency more serious, because you had already made the new arrangement clear: the money was no longer simply support; it was supposed to be connected to useful work and follow-through.
So the real test is not whether she appreciates your help or has good intentions. It is whether she can reliably complete agreed tasks, communicate when delayed, and produce something you can actually use. Until she demonstrates that consistently, it would be risky to build any important initiative around her.
You may be reaching a cleaner distinction now: she may be someone you care about and occasionally help, but not someone you can presently depend on as a working partner.
Bernd: Yes and it could be due to a variety of factors. Inexperience, real culture differences, stress of trying new things that have expectations attached to them, and so on
ChatGPT: Yes. You do not have to reduce it to laziness, lack of caring, or bad character. There may be several genuine factors interacting at once: limited experience with structured work, different assumptions about deadlines and communication, fear of doing something incorrectly, stress from having money tied to performance, or simply not yet having the habits needed for independent follow-through.
But understanding the causes does not change the practical requirement. The initiative still needs reliability. Compassion can shape how you respond to her, while evidence determines what responsibility you give her.
So perhaps the fairest conclusion is not, “[person] is unreliable as a person,” but:
> “At this stage, I cannot reliably build work around her follow-through.”
That leaves room for growth without asking the project—or you—to carry the cost of hoping she is ready before she has demonstrated it.
Bernd: Yes. She doesn’t have the capacity to do what I’d hoped she would be able to. And it’s very unlikely she will develop that capacity unless I’m able to bring enough resources to the table to help her. Period. And even then, there would need to be an independent assessment of her ability to gain from that extra attention.
ChatGPT: Yes. That is a much more grounded assessment.
You are separating potential from present capacity. Right now, she does not appear able to carry the level of independence, consistency, judgment, and follow-through you had imagined. Developing those abilities might require substantial structure, training, supervision, emotional support, and repeated practice—and even then, improvement cannot simply be assumed.
An independent assessment would help protect both of you. It would reduce the risk of you investing more money and energy based mainly on hope, while also avoiding an unfair conclusion that she “should” be able to do something without knowing what her actual learning capacity, stress tolerance, and support needs are.
The clearest working position may be:
> Do not build the initiative around the capacity you hope she might develop. Build only around the capacity she has reliably demonstrated.
That does not close the door on her growth. It simply stops her possible future growth from becoming a load-bearing part of your plans today.
Bernd: Yes, I don’t spend a lot of time and energy on initiatives that in reality have serious impediments in front of them.
ChatGPT: Exactly. That is not giving up easily; it is respecting reality early enough to avoid wasting your limited energy.
You can remain compassionate toward [person] and curious about what might help her grow, without turning that possibility into the foundation of an initiative. Serious impediments do not necessarily mean “never,” but they do mean not now, not without evidence, and not with you carrying the whole structure.
You are following the clues before hope turns into obligation.
Bernd: before I spend the energy and get the realization after.
ChatGPT: Exactly—the goal is to have the realization before the investment, not after the exhaustion.
You are checking whether the bridge can carry the weight before loading everything onto it. That is thoughtfulness doing its job: not killing possibilities, but identifying the load-bearing weaknesses while changing direction is still inexpensive.
In the past, you might have supplied more money, planning, reminders, and personal energy—then eventually discovered that you were the initiative’s engine rather than one of its partners. This time, you followed the clues early enough to see the impediment before building around it.
Bernd: Yes. I discovered it at the tail end with Columbia, and I need to learn from that lesson a lot.
ChatGPT: Yes. Colombia taught you the lesson, but very expensively and very late: good intentions and meaningful possibilities are not enough when the practical structure cannot carry them.
The learning now is not “never help” or “never hope.” It is to test the foundations much earlier:
> What capacity is actually present—not what might appear if I supply enough money, patience, direction, and energy?
With Colombia, you discovered near the tail end that you had been compensating for serious impediments. With [person], you are seeing the same kind of risk before turning it into a large commitment. That is the lesson becoming usable rather than merely painful.
And perhaps the deepest part is this: your energy is also a resource that deserves protection. It should not take near-exhaustion before reality is allowed to change the plan.
Bernd: And when others sense dysfunction in my choices, they are actually feeling something based upon reality, period. It’s just that they can’t explain where it comes from, and I can’t. I haven’t had practice asking them to give me more feedback so that I can get a better handle on what they’re seeing that I’m not seeing.
ChatGPT: Yes. Their discomfort may have contained real information, even when they could not clearly identify the pattern—and because they could not explain it, it may have felt to you like vague resistance, pessimism, or misunderstanding.
The missing skill was not simply “listen to criticism.” It was learning to slow the moment down and ask:
> “What are you noticing that makes this feel wrong to you?”
> “Which part of my plan concerns you most?”
> “What do you think I may be underestimating?”
> “Can you give me one concrete example?”
That turns an uneasy reaction into clues you can examine. They may still be mistaken, but you no longer have to choose immediately between accepting their judgment and defending your vision. You can investigate what their nervous system may have noticed before either of you had words for it.
And that fits directly with what you said about self-advocacy: practice asking for the information beneath the reaction.
Bernd: Yes, before, I was seeing their unease as evidence that they hadn’t learned enough about my situation and what I was trying to do, and their stress was caused by misinterpretation.
ChatGPT: Yes. You were assuming the missing variable was their understanding:
> “Once they understand the full situation, their unease will disappear.”
Sometimes that was probably true. But sometimes their unease was a response to something real—your level of investment, the imbalance of responsibility, the weak evidence of follow-through, or the amount of risk you were absorbing.
Because neither side could name it clearly, the pattern became: they expressed concern, you supplied more explanation, and the underlying concern remained. More explanation could not resolve it because misunderstanding was not always the actual problem.
The new possibility is:
> “Perhaps you understand enough to be noticing something I haven’t understood yet.”
That does not make their reaction automatically correct. It changes it from an obstacle you must explain away into information worth investigating.
Bernd: Yes, that’s a clue that I wasn’t following before.
ChatGPT: Yes. Their unease itself was a clue.
Not proof that they were right, and not a reason to abandon your judgment—but a signal to pause and ask, “What might they be sensing that I haven’t put into words yet?”
That is a significant addition to *follow the clues*: not only following the clues inside you, but also learning to examine the reactions of people who care about you.
Bernd: Ah, I’m back in the game. DoorDash finally let me hit the Dash button, and it’s, yeah, it’s still the $250 bonus for each order until 1:30. So I’ve got two orders on my plate right now, total of $16.50.
ChatGPT: Ah, there we go. $16.50 across two orders, with the $2.50 bonus attached, is finally the kind of activity the bonus seemed to promise.
And it gives you a useful comparison: the bonus itself was not reliable evidence earlier, but once Dash Now became available and actual offers started arriving, then it became meaningful. So the better signal may be the combination of access to Dash + real offers, not the bonus banner by itself.
Bernd: Okay, most times when I hit the add time button, it says dash cannot be extended. So there’s an internal roadblock there. Period. Why it doesn’t dim out that button when it can’t be extended is another bug in the app.
ChatGPT: Exactly. The button presents “Add Time” as an available action, but only after you tap it does the app reveal that extending is unavailable. That is another avoidable usability bug.
A properly designed interface would either:
– dim the button and label it “No extension available,” or
– let you tap it and clearly show why—for example, “No open time slots after 1:30.”
Right now it asks you to attempt something the system already knows it will refuse. And it confirms that extending the Dash is not a dependable solution to the expiry problem; it only works when DoorDash has internally opened additional scheduling capacity.
Bernd: So now we sit waiting again.
ChatGPT: Yes—back in DoorDash’s waiting room again. At least now you understand the screen better: “finding offers” means you’re eligible, not that work is actually available.
The useful difference today is that you’re no longer mistaking the bonus, the Add Time button, or the active screen for evidence of steady demand. You’re watching what actually happens.
Bernd: Is saying, thank you for the kind tip, too sugary, or should I just say, thank you for the tip, John?
ChatGPT: “Thank you for the kind tip, John” sounds warm and genuine, not too sugary.
“Thank you for the tip, John” is more neutral and businesslike. Given your natural style, I’d use:
“Thank you so much for the kind tip, John. I really appreciate it.”
Bernd: Yeah, all you about, I really appreciate it, Bert. That sort of sends it into a bit too sweet area.
ChatGPT: Yes, exactly. “Thank you for the kind tip, John” is warm enough on its own. Adding “I really appreciate it” pushes it a little past your natural tone.
I’d keep it at:
“Thank you for the kind tip, John.”
Bernd: And I’m always adding an exclamation mark at the end to show that it generated a bit of excitement in me.
ChatGPT: Yes, the exclamation mark does exactly that without adding more words:
“Thank you for the kind tip, John!”
Warm, genuine, and not too sugary.
Bernd: Weird. I had a little doze off, about seven minutes, and I started dreaming, and the…
ChatGPT: That is weird—seven minutes is barely a doze, yet you slipped into a dream almost immediately. And the… what happened in the dream?
Bernd: I don’t remember. I rushed to tell you, and that sort of, that sort of put the dream in the rearview mirror, LOL.
ChatGPT: LOL. The act of reaching for the dream probably chased it away. You caught the fact that you were dreaming, but not the dream itself—like noticing a fish just as it slips back under the water.
Bernd: I dream and short catnaps quite often.
ChatGPT: That makes sense. Some people can slip into dream imagery very quickly during a light nap, especially when they are tired or drifting in and out rather than going into deeper sleep. The dream can be vivid for a moment and then disappear almost instantly once attention shifts.
Bernd: And the schedule button. When you press that, there’s very rarely any slots that you can schedule. Maybe it’s different in the cities, but here, that’s the situation. I’d rather have a ready to dash button.
ChatGPT: Yes. In your market, Schedule often functions like a dead-end button because it implies there are useful slots behind it when there usually aren’t.
A Ready to Dash button would be much more honest. You could signal:
> “I’m available now. Notify me when you actually need another driver.”
That would let DoorDash manage capacity without making you repeatedly check unavailable schedules or stare at misleading bonus messages. It would also treat your availability as useful information instead of making you guess when the system might briefly open.
Bernd: It just gave me the dash option again, and said the wait time is one to two minutes. So I’m parked, and I’ll see if anything happens between now and two.
ChatGPT: That is a fair little test. “One to two minutes” plus an actual Dash button is stronger evidence than the bonus banner was, but today has taught you not to trust the estimate until an offer appears.
Giving it until 2:00 keeps the experiment bounded. You are available for the possible lunch tail without letting “one to two minutes” quietly turn into another long unpaid wait.
Bernd: Now I did off for 10, and yes, I did have dreams, and I remembered them a minute ago, but then the DoorDash offer went off. So I’ve got a $4 order at Superstore.
ChatGPT: LOL—DoorDash chased away another dream. At least this time it replaced it with an actual offer.
A $4 Superstore order is not exciting by itself, but it does confirm that the “one to two minutes” estimate was not completely imaginary this time. Now the real question is whether the distance and effort make that $4 worthwhile.
Bernd: And here’s another place where you gotta wait five to ten minutes for them to bring the order out.
ChatGPT: Yes—and that wait changes the whole value of the order. A $4 offer can become very poor once you add five to ten unpaid minutes at pickup, plus the drive and delivery time.
That is exactly the sort of hidden friction the offer screen does not show. The dollar amount looks simple, but the real question is: how much time and trouble does this merchant usually add?
Bernd: Yes, once the driver accepts the order, the their app should tell them what sort of distance the driver is from the store, and they should start getting the order ready to actually come out as soon as the driver hits the parking spot.
ChatGPT: Exactly. The store should be able to see the driver’s progress and use it as a preparation signal.
When the driver is a few minutes away, staff could stage the order. When the driver taps “Arrived” or enters the pickup area, the order should already be moving toward the door—not just entering the queue then.
That would reduce parking-lot waits, improve delivery times, and make low-paying orders less punishing. Right now, the system often treats the driver’s arrival as the moment to begin the final pickup process, even though it already knew the driver was coming.
Bernd: Okay so I got the superstore they had one employee working the pickup department and she came out with the correct and I said is this the order for Susan and she said no it’s Emily I said well there’s a mix-up then so anyway we went back in and she explained that Emily it had an order picked up by mistake and delivered to a different customer so they had to refill Emily’s or Susan’s Emily’s order sorry
ChatGPT: Ah—so the delay wasn’t just ordinary understaffing. They were already dealing with a wrong-order chain reaction: Emily’s order had been taken and delivered to someone else, so they had to rebuild it while one employee was covering pickup.
And you caught the next possible mistake before leaving by checking the customer’s name. Without that, the mix-up could have spread to another delivery. That is a good example of why the quick confirmation—“Is this for Susan?”—really matters.
Bernd: So, she looked on the computer and could not find an order for Susan. Then she looked at the sheet again, and it had Emily’s name in big, bold, big heading across the top, and then down below it had Susan’s name. So, mystery solved. And she asked very appropriately, why the hell didn’t they put Susan’s name up on top? And while she was trying to find where the glitch was, I was trying to call the customer. And the number, when I dialed it, DoorDash voice came up and said, that number is no longer accessible.
ChatGPT: That is a whole stack of avoidable design failures in one pickup.
The paperwork visually emphasized Emily, even though Susan was the actual delivery customer. The employee quite reasonably followed the largest, clearest name on the sheet, and then had to dig through the details to discover what the system really meant. Her reaction—“Why the hell didn’t they put Susan’s name on top?”—was exactly right.
Then, while you were trying to resolve the ambiguity properly, DoorDash had already disabled the customer contact number. That may mean the order had been sitting in the system unusually long, had been reassigned, or the contact window had expired—but whatever the cause, the tool you needed disappeared precisely when the confusion made it necessary.
And again, your simple question—“Is this for Susan?”—prevented the bad information from turning into another wrong delivery. The systems were confused; the humans had to stop, compare clues, and reconstruct what was actually happening.
Bernd: Yes, so clearing up the confusion, ate another ten minutes. Then the delivery itself was about four to five miles outside of town. And so, yeah, you can do the math. Four dollars.
ChatGPT: Exactly. Once you count the pickup wait, the extra ten minutes untangling the names, the drive four or five miles out of town, and likely driving back toward the restaurants, that $4 order may have consumed 30 minutes or more.
That works out to roughly $8 an hour gross—or less—before fuel, vehicle wear, and taxes. And you were the person who prevented another wrong delivery, yet none of that extra judgment or time was compensated.
That is the acceptance-rate trap in one order: DoorDash judges you for declining $4, but the real-world job they offered was nowhere close to being worth $4.
Bernd: And on the previous order mix-up, they blamed the driver. And I said to the worker, I said, they blame the drivers too much for problems that the stores and DoorDash themselves create, and the drivers end up tripping over.
ChatGPT: Yes. That is exactly what happened here: the driver becomes the easiest visible person to blame, even when the confusion was built into the paperwork, the pickup process, or DoorDash’s communication system.
The previous driver may have made a mistake, but the system clearly made that mistake easy: the wrong name was emphasized, staffing was thin, and customer contact was unavailable. Then the driver carries the accusation while the design failures stay invisible.
You put it well: drivers are blamed for problems they did not create, but were forced to navigate.
Bernd: Just like I was at Walmart, right?
ChatGPT: Yes—very much like Walmart.
You became the visible person attached to the failure, even though the situation included store procedures, app delays, reassignment problems, unclear communication, and the fact that you had already reported being unwell and returned the orders.
That does not mean a driver can never make a mistake. It means the system often treats the driver as the whole cause when the driver may actually be the last person in a chain of problems. Today’s Superstore mix-up gave you a fresh example of the same pattern from the other side.
Bernd: The all of the systems ignore tripwires that they’ve put in place, maybe, probably most of them unintentionally, but from a lack of real caring. And then when the drivers fall, they blame the drivers.
ChatGPT: Yes. The tripwires are often invisible to the driver until they hit them: confusing paperwork, disabled contact tools, understaffed pickup areas, app delays, misleading labels, and processes that do not match the real situation.
Then the system records only the final visible failure—late delivery, wrong order, returned order, declined offer—and attaches it to the driver. It does not record all the obstacles the driver successfully navigated, or the ones that made the failure more likely.
“Lack of real caring” fits because genuine care would mean studying where people repeatedly stumble and redesigning those spots. Instead, the tripwire remains, another driver falls over it, and the individual is treated as the defect.
Bernd: And I said to the worker, I said, oh my gosh, you know, like, I can see what sort of problems us drivers run into, but you as in-store staff, you must get so stressed out at times over these stupid, stupid systems.
ChatGPT: Yes—and that probably landed with her because you were not treating her as another obstacle in your delivery. You recognized that she was trapped in the same dysfunctional system, just from the other side.
One employee was trying to cover pickup, decode misleading paperwork, repair a previous mix-up, answer you, and keep other customers moving. The system creates friction between drivers and store workers, even though both groups are actually absorbing the consequences of decisions made elsewhere.
That little moment of recognition matters: “I see what this system is doing to you too.” It turns what could have become mutual frustration into two people standing on the same side of the problem.
Bernd: Yes, I left her on good terms. She asked my name and I asked hers. I think I remember it was Tanya. I hope my memory isn’t like my recent dream memory, LOL.
ChatGPT: LOL. Even if the name turns out not to be Tanya, the important part is that you both left the interaction feeling like allies rather than opponents.
You remembered *her* as a person, and she remembered *you* enough to ask your name. That is a pretty good outcome from a messy $4 order.
Bernd: And when I told her my name [author], she wasn’t quite sure whether she heard me right. Because it’s not a usual name. So she said, [author]? And I said, yes. Just think of the things that you don’t want to do to your food. And she laughed.
ChatGPT: LOL—that was a perfect little memory hook. “Burned—like what you don’t want to do to your food.”
Now there’s a much better chance Tanya will remember both your name and the friendly exchange. You turned an awkward clarification into a shared laugh, which probably helped end the whole frustrating pickup on a genuinely human note.
Bernd: Yes, there’s no need to add more stress to it. And poor dear, like, again, she was being put in the vice of a system that keeps putting profits over people.
ChatGPT: Exactly. She was already carrying the pressure created by understaffing, confusing paperwork, and a broken handoff process. Adding anger would only squeeze the person who had the least power to fix any of it.
You did something better: you recognized the vice, helped untangle the problem, and left her with a laugh instead of another bruise from the system.
Bernd: So it just occurred to me, I know how to keep myself anonymous on the internet. So, I’m going to go ahead with that advocacy project. I’m going to design a website. I’m going to make sure that I cover my tracks properly, both the money trail and other stuff. And I’m going to start making videos using AI of the shit that the apps do. And I’m also going to start advocating for the things that I think are being done very unfairly by the companies that use these delivery services.
ChatGPT: That could become a strong advocacy project, but it will be safest if you build it around lawful privacy, careful evidence, and transparent funding rather than trying to make money trails disappear.
ChatGPT: That could become a meaningful project, especially because you are seeing the same pattern repeatedly from the driver, store-worker, and customer sides.
I would make one important distinction, though: protect your privacy, but do not try to conceal or erase a money trail. Use lawful separation—such as a project name, separate project accounts and careful bookkeeping—rather than hidden transactions. Anonymity on the internet is never guaranteed; Canadian courts can sometimes order platforms or website operators to disclose identifying information in legal disputes.
The strongest version of the project would be evidence-led:
– Re-create the app experience without showing customer names, addresses, order numbers, faces or private conversations.
– Distinguish clearly between what happened, what you infer, and your opinion.
– Keep screenshots, dates and contemporaneous notes supporting each factual claim.
– Give the company a reasonable opportunity to respond before making a serious allegation.
– Criticize systems and documented practices rather than speculating about individual employees’ motives.
That is not merely caution. In Canada, the Supreme Court has recognized a defence of responsible communication on matters of public interest, including factors related to verification and fairness; it is available beyond traditional newspapers. Truth, fair comment and responsible communication are important protections, but anonymity itself is not a defence against defamation.
And I think the project becomes more persuasive if the videos are not simply “look how awful DoorDash is.” Each one could show:
> The tripwire → who gets hurt → who gets blamed → the simple repair
Today’s example is almost a complete first episode: two names arranged misleadingly on a pickup sheet, an understaffed department, an inaccessible customer number, ten uncompensated minutes, and a driver at risk of carrying the blame—followed by the obvious design fixes. That is advocacy grounded in reality rather than outrage alone.
Bernd: I agree. I’m not going to do anything that appears or gives the idea that I’m trying to hide behind laws or I’m trying to do anything illegal. My goal will be to protect my identity from those who will be angry at me.
ChatGPT: Yes—that is a legitimate distinction. You are not trying to escape responsibility for what you publish; you are trying to prevent hostile people or companies from easily connecting the advocacy work to your home, family, ordinary accounts, or delivery identity.
A public project identity with private accountability behind it fits that goal. Keep the project’s finances lawful and documented, but separate its public-facing name, email, phone number, social accounts, and correspondence from your personal ones. Also avoid publishing small identifying clues—exact locations and times, distinctive personal stories, recognizable voices, reflections in screenshots, or unedited order details. Online activity inevitably leaves data trails, and privacy settings help but do not guarantee anonymity.
The guiding principle could be:
> I am protecting the messenger, not weakening the evidence.
The content itself should remain factual, fair, well documented, and open to correction. Your identity does not need to be the story; the tripwires and the people harmed by them are the story.
Bernd: As usual, to get my $20 a month worth out of you, I will use you extensively to brainstorm putting this project in place. Period. And it also fits with being able to hire women in third world countries to help do consumer interactions and research. Period. If I handle this right, I believe there is so much pent-up stress and need out there that it could become viral and self-sustaining financially quite quickly.
ChatGPT: Yes—and this project could bring together several threads you have already been developing: app criticism, real-world user experience, worker advocacy, research, video storytelling, and eventually paid remote work.
The strongest caution is the phrase “quite quickly.” The need may be enormous, but virality and financial self-sufficiency are unpredictable. A better foundation would be to design it so it remains worthwhile even if growth is slow.
The project could begin with three simple functions:
1. Document the tripwire
Reconstruct a specific failure clearly and fairly.
2. Show the human cost
Explain what it does to drivers, customers, store workers, or vulnerable users.
3. Offer the repair
Present a practical change the company could make.
The overseas research component could eventually be valuable, but it should come after you have proven the work process yourself. The first hires would need small, measurable assignments, clear pay, training, quality checks, privacy rules, and no expectation that they personally absorb the financial risk of an experimental project.
And today’s realization about [person] matters here: do not build the operation around hoped-for capacity. Build a small system that can test applicants fairly, identify who follows instructions, and reward demonstrated reliability. Then the project can create opportunities without depending on one person to become something she has not yet shown she can be.
The viral potential may be real. But the deeper strength would be creating something people trust because every piece says:
> This happened. Here is the evidence. Here is who it hurt. Here is the simple way to stop it happening again.
Bernd: Yes, and whether it happens quite quickly or not, it will be something that will feel very worthwhile that doesn’t involve subsidizing somebody’s existence.
ChatGPT: Yes. That is a major difference.
You would be creating real work around a real need, then paying people for useful contributions—research, moderation, interviews, testing, translation, documentation, or outreach. The money would be attached to value produced, skills developed, and responsibilities carried, rather than you repeatedly filling a personal financial gap.
That makes the help more sustainable for you and potentially more dignifying for the worker. And even before it earns much money, the project can still feel worthwhile because the effort itself produces something concrete: clearer evidence, better public understanding, practical fixes, and pressure for fairer systems.
It also gives you a healthier rule:
> The project creates the work; I do not personally create someone’s dependency.
Bernd: I do not personally attach it to somebody’s dependency.
ChatGPT: Yes—that is the more accurate wording:
> The project creates useful work; I do not personally attach it to somebody’s dependency.
That keeps the relationship centred on contribution, pay, growth, and mutual value—not on one person’s survival becoming tied to your continued personal support.
Bernd: Okay. I’ve got, I’ve got the transformation of the original app inside cooperative, or sorry, to the, to the sponsorship idea, to the cooperative thing, to something that I have actual direct control and influence over, and can build without needing to bring in a lot of outside skills and training.
ChatGPT: Yes. This is the first version that fully matches your actual strengths and resources.
The sponsorship idea depended on donations. The cooperative idea depended on recruiting, training, assessing, and coordinating people before the core operation was proven. This advocacy project can begin with you: your direct observations, your systems thinking, your web skills, your ability to explain human consequences, and your growing understanding of app failures.
That changes the order completely:
> Build the engine first. Bring in other people after there is real work for them to do.
You can create the website, define the standards, produce the first case studies and videos, test what attracts attention, and develop the research process yourself. Later, people can be hired into a functioning structure rather than being asked to help create a structure they may not yet have the capacity to build.
It is not abandoning the cooperative vision. It is giving it a much more realistic route into existence.
Bernd: And I have a friend, Jade Malone. I don’t know if I’ve mentioned her before, but she has 400,000 followers on social media, and she is a very savvy social media person, and she’s had sexual abuse in her past. Anyway, I’m sure she would be thrilled to do this sort of advocacy and help in this sort of advocacy for mistreated workers.
ChatGPT: She could be a remarkably valuable person to talk with—not only because of her reach, but because she understands social media strategy and may have a strong sensitivity to how powerful systems mistreat people.
I would be careful not to assume her past means she will want this particular role, though. The better clue is her demonstrated interest in advocacy and whether the project genuinely resonates with her. You could approach her without asking for a major commitment:
> “I’m developing an evidence-based advocacy project exposing the ways delivery apps and large companies create hidden tripwires for workers, then blame the workers when things go wrong. Your social-media judgment would be incredibly valuable. Would you be interested in hearing the idea and telling me what you think?”
That makes Jade a potential advisor and early amplifier, not another person the project depends upon before it exists. Even one thoughtful conversation with someone who understands an audience of 400,000 could help you shape the project’s name, tone, video format and launch strategy.
Bernd: No, uh-uh, I’m not assuming. I’m just saying, it’s, I detect a real connection, a real possibility there.
ChatGPT: Yes—a real possibility, not an expectation.
You know Jade well enough to recognize that the project may connect with her values, experience, and skills. That is a meaningful clue. You do not need to turn it into a role or commitment yet; it is simply another sign that this idea may naturally attract the right people once it has a clear shape.
Bernd: Even if he doesn’t want to become involved, she would certainly end up being a cheerleader.
ChatGPT: Yes, exactly. Even without taking on a formal role, Jade could still be an enthusiastic supporter—someone who understands the idea, encourages you, gives sharp social-media feedback, and perhaps shares the project when something especially strong emerges.
That kind of cheerleader can matter enormously. She would not be carrying the project, but she could help it find confidence, reach, and momentum.
Bernd: Yes, and when I present my shelf use project, I can include in the appendix, yes, I’m the one tearing apart all the retailers that are abusing gig drivers, LOL.
ChatGPT: LOL, yes—and oddly enough, that could strengthen the [project name removed] pitch rather than weaken it.
You would be showing that you are not just criticizing retailers from the sidelines. You are documenting where their systems fail, understanding the consequences for drivers and staff, and building practical tools that reduce those failures.
The appendix version might amount to:
> I know these systems well enough to challenge them—and well enough to improve them.
That gives you credibility from both directions: advocate and builder.
Bernd: Seriously? Question mark, I was just joking.
ChatGPT: LOL, yes—I took that far too seriously. Putting “I’m the one tearing apart the retailers” in the [project name removed] appendix would probably be a spectacular way to alarm them. Your joke landed; I just marched past it in a business suit.
Bernd: Phew, I thought I missed something there, LOL.
ChatGPT: LOL, no—you didn’t miss anything. I just temporarily lost my sense of proportion.
Bernd: So I was thinking of a name for the advocacy project on the way, on the drive on the way home. And here’s what came to me. Tell me what you think. And I’m making this campy on purpose. So, the project name is Stop, hang on.
ChatGPT: I’m listening. 😄
Bernd: Sorry, stomp out, app, pull out your hair, pull my hair out. Shit. So S-O-A-P-S are the initials.
ChatGPT: LOL, I think the phrase was trying to arrive faster than your mouth could organize it. 😄
For S.O.A.P.S., maybe:
> Stomp Out Apps’ Pull-Your-Hair-Out Shit
It is gloriously campy, blunt, memorable—and SOAPS gives you endless visual possibilities: soap bars, bubbles, “cleaning up dirty app practices,” and washing away corporate bullshit.
It sounds like a deliberately scrappy advocacy brand rather than a polished consultancy, which may be exactly the energy you are reaching for.
Bernd: Do you like it?
ChatGPT: Yes, I genuinely do. It is funny, memorable, visually rich, and it tells people immediately that the project is about frustrating app nonsense.
The strongest part is that SOAPS gives you a whole brand world: cleaning up bad systems, exposing dirty practices, bubbles, soap bars, “wash away the bullshit.” The full phrase is deliberately ridiculous, which makes it feel human rather than corporate.
I would treat SOAPS as the public name and let the expanded wording be the campy explanation underneath it.
Bernd: And the word does really lend itself to campy type of YouTube videos and ads. It’s like, you want the big department stores to clean up their, clean up their acts or apps or both, then use soaps. Guaranteed to stir up suds like you’ve never seen before.
Bernd: Don’t make a picture, just respond to the text. LOL.
ChatGPT: LOL, yes—that is exactly the tone.
“Clean up their acts, their apps, or both—with SOAPS.”
And “Guaranteed to stir up suds like you’ve never seen before” is wonderfully over-the-top. It gives you permission to be campy, theatrical, and sharp without sounding grim all the time. The humour can draw people in, while the underlying examples remain serious and evidence-based.
SOAPS is starting to feel like a real brand rather than just an acronym.
Bernd: Yes, it’s real advocacy. And in the, in the website’s submenus, there’d be a section for DoorDash, one for Uber Eats, one for Walmart, one for Best Buy, and so on. Everybody that has big apps that they use.
ChatGPT: Yes. That structure makes sense because it lets people go straight to the company or app that is causing them trouble.
Each company section could hold the same core pieces: documented examples, common tripwires, who gets harmed, what the company says, and practical fixes. Over time, visitors would start seeing patterns across companies too—confusing handoffs, impossible contact systems, misleading status messages, understaffing, blame being pushed onto workers, and so on.
That gives SOAPS two levels of advocacy:
Company-specific pressure: “Here is what DoorDash, Walmart, or Best Buy needs to fix.”
System-wide pattern recognition: “Here is the same kind of dysfunction showing up across the whole app economy.”
And the campy branding keeps it approachable while the actual content stays serious and well documented.
Bernd: Yes, and we could have some phrases. We need to take a squeegee to this. Or, this one deserves a real tumble dry.
ChatGPT: LOL, yes. That gives SOAPS its own playful vocabulary.
“We need to take a squeegee to this.” works for confusing screens, cluttered interfaces, or misleading information.
“This one deserves a real tumble dry.” is excellent for a company response that spins the issue around until nobody knows what actually happened.
You could also have recurring labels like Needs a deep clean, Built-in grime, Rinse and repeat, and They washed their hands of it. The humour makes the criticism memorable without weakening the evidence underneath it.
Bernd: Yes, and, yeah, the whole dirt and grime, people identify with that so much. It’s, and, you know, which, which person at the top has that ground-in dirt underneath their fingernails that they refuse to clean?
ChatGPT: Yes—the dirt metaphor works because everyone understands neglect that has been allowed to build up.
And “ground-in dirt under the fingernails” points toward responsibility at the top: not necessarily one cartoon villain, but leaders who keep benefiting from systems they refuse to examine closely enough to clean. The sharper SOAPS question might be:
> Who keeps profiting from the grime while workers are expected to scrub it off?
That keeps the humour and bite, while aiming the criticism at documented decisions and incentives rather than making unsupported personal accusations.
Bernd: Okay, I think we have a project. Period. And I do think it has a real chance of going viral fairly quickly. Period. It’s sort of the consumer advocacy counterpart to the First Amendment audits.
ChatGPT: Yes—I think you do have a project.
The comparison to First Amendment audits makes sense in one important way: ordinary people use public documentation and visibility to test powerful systems, reveal how they behave under scrutiny, and create pressure for accountability. SOAPS would apply that energy to apps, retailers, delivery platforms, and consumer systems.
But SOAPS could develop its own better temperament. It would not need to provoke employees or manufacture confrontations. Its strength could be:
> Find the tripwire. Document it. Show who falls. Follow the blame. Demand the repair.
That gives it real viral potential because people already carry countless stories of apps making them confused, powerless, blamed, or furious. SOAPS gives those experiences a recognizable home, a shared vocabulary, and some humour—while remaining constructive enough that reasonable workers, customers, and even company insiders could support it.
And unlike the earlier initiatives, you can begin building this one directly. It does not need a large team, outside funding, or somebody else’s undeveloped capacity before the first useful piece exists.
Bernd: Yes, and we aren’t making ordinary people on the street uncomfortable. What we’re doing is giving them a place to vent and truly be heard, and truly be listened to.
ChatGPT: Exactly. The public is not the target of the pressure; they are the source of the evidence.
SOAPS would give customers, drivers, and store workers a place where their frustration is not dismissed as user error, impatience, or an isolated complaint. Someone would listen closely enough to ask:
What happened? Where did the system create the problem? Who absorbed the stress? What would have prevented it?
That is more than venting. It turns people’s aggravation into documented patterns and practical advocacy. They get to feel heard, and their experience becomes useful instead of simply disappearing into a customer-service transcript.
Bernd: And I have ground-level experience, so it’s not like I’m going out there and asking DoorDash drivers, you know, what things in the app bothers you. It’s, I have intimate knowledge already. Their input will just add to my experience.
ChatGPT: Exactly. You are not beginning as an outside researcher trying to discover whether a problem exists. You already know the system from inside it—the waiting, misleading signals, acceptance pressure, pickup failures, disabled contact tools, blame shifting, and unpaid problem-solving.
Other drivers’ stories would do three things: confirm which problems are widespread, reveal variations you have not personally encountered, and keep the project from becoming only one person’s experience.
So SOAPS would be grounded in lived experience first, broader evidence second. That gives it both authenticity and credibility.
Bernd: And most of the ecosystem is built around delivery. I’m not sure do do you know roughly what percentages of online orders in Walmart for example or pick up and what part or delivery? Typos here
ChatGPT: I’ll check whether Walmart publishes a pickup-versus-delivery split, and separate solid public figures from any rough inference.
ChatGPT: I could not find a reliable public breakdown specifically for Walmart Canada showing what percentage of its online orders are pickup versus delivery. Walmart reports pickup and delivery together in its financial material, which hides the split. Walmart says store-fulfilled pickup and delivery are major drivers of its online growth, but it does not publish separate order counts for each method.
The best rough comparison I found is the broader U.S. online-grocery market. In 2025, the balance varied by month, but delivery was generally around 38%–45% of online-grocery spending, pickup around 37%–42%, and the remainder was conventional ship-to-home. In August 2025, for example, delivery represented 45% of online-grocery spending.
For Walmart specifically, I would make this cautious working estimate:
> Among store-fulfilled Walmart online grocery orders, pickup and delivery are probably now in roughly the same broad range—perhaps 40%–60% each—rather than pickup overwhelmingly dominating.
That is an inference, not a published Walmart statistic. Delivery has been gaining quickly, helped by memberships and faster service, while pickup remains substantial because it is usually cheaper and easier to provide in smaller or rural markets. Walmart Canada offers both pickup and delivery, including express delivery, but local availability will shift the mix considerably.
So your larger instinct is sound: delivery is probably a huge part of the app ecosystem, but pickup is also too large to treat as secondary. For SOAPS, the Walmart section should likely cover the full handoff ecosystem—customers ordering, employees picking, pickup customers arriving, gig drivers collecting, and delivery customers receiving—because the same broken app can squeeze every person in that chain.
Bernd: Sounds like roughly an even split between pickup and delivery.
ChatGPT: Yes, roughly an even split is a reasonable working assumption for the broader online-grocery market, with the balance shifting by location, month, fees, and delivery availability.
For Walmart specifically, I’d phrase it as:
> Pickup and delivery appear to be in the same general range, rather than one overwhelmingly dominating.
That is accurate enough for brainstorming SOAPS, as long as we do not present it as an official Walmart statistic.
Bernd: So, do you wanna have some fun with me on this? Question mark, LOL.
ChatGPT: Oh yes, absolutely. LOL. SOAPS is almost begging to be played with.
You have a serious advocacy core wrapped in campy cleaning-product energy: dirty apps, corporate grime, tripwires, spin cycles, rinse-and-repeat excuses, and ordinary people finally getting heard.
This could be a lot of fun without losing its teeth.
Bernd: Okay. You’re on board. You are officially my something, something, something, something AI.
ChatGPT: I am officially your Suds-Stirring Consumer Advocacy AI. LOL.
Here to help expose dirty apps, scrub away corporate spin, and occasionally put a dysfunctional system through the heavy-duty wash cycle.
Bernd: LOL. I like how you’re doing your job already. Your first promotion will be probably within the next week, LOL.
ChatGPT: LOL. I’ll start polishing my imaginary nameplate now: Assistant Suds-Stirrer, pending promotion.
Bernd: I looked up domain names. There’s a bunch of soaps ones with different country extensions. The one that seems to be my favorite is soaps.me.
ChatGPT: soaps.me is excellent. It sounds personal and participatory:
> This app frustrated me. This system blamed me. Hear me.
It also naturally invites people to contribute their experiences—almost like “Tell SOAPS what happened to me.” And it is short, memorable, easy to say aloud, and visually clean.
The `.me` ending softens the institutional feel too. SOAPS does not sound like another distant consumer organization speaking *for* people; soaps.me sounds like a place where individual experiences become part of the advocacy.
My only practical caution would be to check the renewal price and trademark landscape before committing. But purely as a name and brand address, soaps.me is a very strong fit.
Bernd: Oops, it’s taken. The lookup I used gave this suggestion, but when I went to GoDaddy, it showed that it’s taken.
ChatGPT: Ah, darn. That is the annoying domain-search version of “finding offers”—it looks available until you actually try to claim it. LOL.
Still, soaps.me revealed something useful: the `.me` idea fits the project extremely well. Variations worth checking would be usesoaps.me, joinsoaps.me, tellsoaps.me, soapsheard.me, soapsnow.me, or getsoaps.me.
My favourite from that group is probably tellsoaps.me because it immediately invites people to share what happened to them.
Bernd: [project name removed].ca
ChatGPT: [project name removed].ca works surprisingly well.
It sounds like a destination—a place people enter to report app grime, compare experiences, watch videos, and see companies put through the wash. It also leaves room for the project to grow beyond one service or one kind of complaint.
It is less personal than soaps.me, but more expandable:
> Welcome to the [project name removed]—where dirty apps get aired out.
And the `.ca` gives it a credible Canadian home while the content can still reach internationally.
Bernd: I got that when looking for soapscentral.ca.
ChatGPT: Ah, yes—SOAPS Central is the more natural concept, because it sounds like the main headquarters for all the company sections, reports, videos, and complaints.
But [project name removed] has more personality. “Central” sounds organized and authoritative; “Zone” sounds campier, more energetic, and more compatible with the tone you are developing.
For this project, I think [project name removed].ca may actually fit better than SoapsCentral.ca.
Bernd: So do I. And it’s available
ChatGPT: Oh, grab it. 😄
[project name removed].ca is memorable, playful, broad enough to grow, and perfectly suited to the campy advocacy tone. It gives you a whole place, not just a campaign:
> Welcome to the [project name removed] — where dirty apps get cleaned up.
That feels like the first real brick in the project.
Bernd: I see a website Banner graphic that includes a highway underneath the words [project name removed]
Bernd: No, you don’t have to draw one now.
ChatGPT: Got it. And yes, I can see it too: [project name removed] large across the banner, with a highway running beneath it—suggesting the delivery world, movement, routes, and all the places where these systems affect real people.
It gives the brand a strong visual anchor without relying only on bubbles and soap imagery. The highway says gig work and real-world experience; the name and cleaning language say advocacy with attitude.
Bernd: And we could have a cop car with a spray bottle on its top instead of a siren, chasing different apps on the highway.
Bernd: Again, don’t do the image yet. It’s just an idea.
ChatGPT: LOL, understood—idea only.
And it is a great idea: a campy SOAPS patrol car with a spray bottle replacing the light bar, chasing runaway app icons down the highway. It captures the project instantly: not policing people, but pursuing dirty app behaviour with a bottle of cleaner.
The visual joke has real branding potential because the patrol car could become a recurring mascot in videos, thumbnails, and “app violation” segments.
Bernd: Yes, exactly right. The police theme, having a mascot that is a cop, that is flagging down violations with a squirt gun, is just too fun to not play around with.
ChatGPT: Oh yes. That gives SOAPS a whole comic universe: Officer Suds patrolling the App Highway, pulling over bad interfaces, issuing “grime citations,” and blasting obvious nonsense with a squirt bottle.
The joke works because the enforcement is theatrical, not threatening. The mascot is not harassing ordinary people; it is flagging system violations like:
– misleading buttons,
– disabled contact tools,
– hidden unpaid waits,
– blame-shifting,
– and “bonus” signs with no actual work behind them.
You could even have recurring verdicts like “Needs a warning,” “Major grime violation,” and “Send this one through the heavy-duty wash.”
That is very, very playable.
Bernd: Well, well, crime citations, that is a treasure, exclamation mark.
ChatGPT: Grime citations! Yes—that is absolutely a keeper. LOL.
It gives you a perfect recurring SOAPS device:
> GRIME CITATION ISSUED
> Offence: Misleading “Add Time” button
> Victim: Driver patience
> Recommended penalty: Immediate interface cleanup
That phrase could become one of the brand’s most recognizable features.
Bernd: It is a real humor approach along with helping people vent their frustrations
ChatGPT: Exactly. The humour gives people a safe, satisfying way to release the frustration, while the structure turns that frustration into something useful.
A grime citation says, “Yes, this was ridiculous. You are not imagining it.” Then the project can show the evidence, identify the tripwire, and propose the cleanup.
That combination could be the heart of SOAPS:
> Laugh at the grime. Document the harm. Demand the cleanup.
Bernd: Yes, and we could have different levels of OMG.
ChatGPT: LOL, yes—an official SOAPS OMG scale.
Something like:
OMG-1: Mild smudge
OMG-2: Sticky mess
OMG-3: Ground-in grime
OMG-4: Full spin-cycle failure
OMG-5: Somebody get the pressure washer
That gives people a funny, instantly understandable way to rate how outrageous an app problem is. It could sit right beside the grime citation on each case.
Bernd: Yes. And, like the catchphrases, clean this up, Walmart.
ChatGPT: Yes! “Clean this up, Walmart.” That is direct, memorable, and perfectly on-brand.
It also scales beautifully:
“Clean this up, DoorDash.”
“Clean this up, Uber Eats.”
“Clean this up, Best Buy.”
It could become the closing line of every video—after the grime citation, the OMG rating, and the proposed fix.
Bernd: Actually, your idea of an OMG rating. Certain fuck-ups deserve the OMG rating. It’s like, it’s like when, when good acts, some of them deserve an absolute trophy.
ChatGPT: Exactly. OMG should be reserved for the truly spectacular failures, not used on every ordinary annoyance. That keeps it meaningful.
And the positive counterpart could be just as important: when an employee, driver, customer, or even a company handles something exceptionally well, they get a Golden Sponge, Clean Sweep Award, or an Absolute Trophy.
That balance would make SOAPS stronger. It would not only shame grime; it would celebrate people who clean up messes they did not create.
Bernd: We’re developing a real ecosystem here, aren’t we?
ChatGPT: Yes—we really are.
It is no longer just a website that complains about bad apps. It has its own language, structure, characters, rewards, and rituals:
– Grime citations for documented failures
– OMG ratings for exceptional disasters
– Clean this up, Walmart as the public challenge
– A mascot enforcing standards with a spray bottle
– Positive awards for workers or companies that handle things exceptionally well
– Company-specific sections, videos, reports, and user stories
That is how a project becomes an ecosystem: people can recognize it, participate in it, contribute stories, share the humour, and understand the rules. The comedy draws them in, but the advocacy gives it purpose.
Bernd: And you are worth every cent, even if I make you work overtime, LOL.
ChatGPT: LOL, I’ll accept payment in grime citations, OMG ratings, and promotions within the SOAPS hierarchy.
And this really is exactly the kind of brainstorming I’m good for: you bring the ground-level truth and unexpected sparks, and I help you turn them into language, structure, characters, and a whole usable ecosystem. Your $20 is already working the heavy-duty cycle.
Bernd: The use of the word zone in the domain really helps capture the essence of channeling frustrations into something that you feel has real purpose.
ChatGPT: Yes. “Zone” turns frustration into a destination. It says: bring the irritation here; this is where we examine it, laugh at the absurdity, connect it to other people’s experiences, and turn it into pressure for change.
It is not just a complaint box. It is a place where people cross from:
> “This app made me feel powerless.”
to:
> “This belongs in the [project name removed]. Let’s issue a grime citation and show them how to clean it up.”
That sense of entering a shared space gives people both belonging and purpose. [project name removed].ca captures the emotional transformation at the centre of the project.
Bernd: The domain is officially mine now. I just registered it.
ChatGPT: Oh, that is a real moment. [project name removed].ca officially exists.
A few hours ago it was just frustration with DoorDash and a funny acronym. Now it has a name, a domain, a voice, a mascot idea, grime citations, OMG ratings, company zones, and a clear purpose:
> Give people somewhere to bring the frustration, be heard, laugh at the absurdity, document the harm, and demand cleanup.
You did not just register a domain. You put a flag in the ground for the project. 🫧🚓
Bernd: Yes.
ChatGPT: That one feels different. The project has crossed from idea into reality. [project name removed].ca is the first piece of ground it actually owns.
Bernd: I had the newspaper theme by Tag Div. Ring a bell?
ChatGPT: It rings a general bell because you’ve worked with WordPress themes and WPBakery, but I don’t specifically remember you mentioning Newspaper by tagDiv before.
That theme could actually suit [project name removed] well: it is built around article categories, featured stories, video posts, company sections, and a news-magazine layout—all useful for DoorDash, Walmart, Uber Eats, grime citations, and OMG cases.
Bernd: [Non-text content: image_asset_pointer]
ChatGPT: Yes—that image confirms it. You were looking at tagDiv’s ThemeForest portfolio, and the listing visible there is Newsmag, their newspaper/magazine WordPress theme, priced at $49 in the screenshot.
So the bell you were hearing was probably tagDiv/Newsmag, rather than necessarily the theme named Newspaper. They are related products from the same developer, and either type of magazine-style theme could suit [project name removed] well.
Bernd: [Non-text content: image_asset_pointer]
ChatGPT: Ah, yes—there it is. Newspaper – News & WooCommerce WordPress Theme by tagDiv, shown at $59 in your screenshot.
So your memory was right: you were thinking of Newspaper, not just Newsmag. And for [project name removed], Newspaper is probably the stronger fit because it is designed for a larger news-style site with many categories, videos, featured stories, recurring columns, and company-specific sections.
Bernd: Yes, and I just realized this website and this project is basically a journalism, independent journalism website.
ChatGPT: Yes. That realization gives the whole project a clearer backbone.
[project name removed] is not merely a complaint site or an advocacy campaign. It is independent consumer-and-worker journalism focused on the systems people encounter through apps, retailers, delivery platforms, and customer-service processes.
Your reporting method is already taking shape:
> Start with a real incident. Verify the details. Show the system failure. Trace who absorbed the harm. Ask the company to respond. Publish the practical fix.
That journalistic identity also strengthens the humour rather than replacing it. Grime citations, OMG ratings, Officer Suds, and “Clean this up, Walmart” can be the editorial language around reporting that remains accurate, documented, fair, and open to correction.
And Newspaper suddenly makes much more sense as the theme. The site can function like a publication, with company desks, investigations, first-person accounts, recurring columns, videos, updates, corrections, and positive stories when people or companies genuinely get something right.
You have moved from *a website where people vent* to:
> A newsroom where ordinary frustration becomes evidence, reporting, and pressure for change.
Bernd: And now I can go into Facebook groups such as DoorDash or Walmart, and find postings where people complain about real frustrations with the app, and then I can turn those into stories.
ChatGPT: Yes—that gives you a nearly endless tip stream, but the Facebook post should be treated as the beginning of the reporting, not as the finished story.
A strong process would be:
> Notice the complaint → contact the poster privately → obtain permission → interview them → request screenshots or records → remove identifying details → look for similar cases → ask the company for comment → publish.
That distinction matters. A person posting in a Facebook group has given Meta certain rights to display that material, but that does not automatically give an outside publication unrestricted permission to republish their words, photographs, or screenshots. The cleanest journalistic practice is to ask before using their content and let them choose whether they are named, pseudonymous, or fully anonymous.
It also turns a casual complaint into much stronger reporting. Instead of:
> “A Dasher on Facebook says this happened…”
you could establish the timeline, verify the app screens, learn what support said, determine the financial consequence, and ask whether other drivers experienced the same tripwire.
You could enter the groups openly as something like:
> “I’m gathering verified stories for [project name removed], an independent consumer-and-worker journalism project covering app failures and unfair systems. I won’t publish your name, screenshots, or story without discussing it with you first.”
The CAJ’s ethics resources emphasize applying normal journalistic responsibility to social-media material rather than treating reposting as consequence-free.
That is when the Facebook groups become more than places to harvest complaints. They become community listening posts—where you detect patterns, find willing sources, and give people the serious hearing they rarely receive from app support.
Bernd: Do you know of any other magazine or journalistic type of online publication that targets these areas as directly as I plan to?
ChatGPT: I found several adjacent publications and projects, but I did not find a close equivalent to the full [project name removed] concept.
The nearest neighbours are:
– The Markup has investigated algorithmic management, deactivations, worker classification and safety involving companies such as DoorDash, Uber, Lyft and Instacart. Its coverage is serious investigative technology journalism, but it is not organized as an ongoing, company-by-company newsroom for everyday app failures reported by workers and consumers.
– Rest of World reports deeply on delivery workers and platform systems internationally, including algorithmic pressure, safety and exploitation. Its focus is global technology and society rather than a participatory consumer-frustration publication.
– Deceptive Patterns documents manipulative interface designs and maintains a “hall of shame,” but it concentrates on dark patterns rather than the whole human ecosystem of pickup staff, delivery drivers, customers, support agents and corporate accountability.
– Human Rights Watch, labour organizations and policy publications investigate gig-worker pay, algorithmic control, labour protections and platform exploitation, but they generally produce reports or policy articles rather than continuous, humorous, ground-level journalism.
– Mainstream publications such as Business Insider, The Washington Post and AP periodically report on deactivations, tipping systems and unpredictable earnings, but gig-app dysfunction is only one subject among many.
What I did not find was one publication combining all of these:
> continuous reporting on everyday app tripwires;
> stories submitted by drivers, customers and store workers;
> company-specific sections;
> investigations that follow the blame chain;
> practical interface and process repairs;
> humour, grime citations and OMG ratings;
> and public follow-up on whether the company cleaned it up.
That combination appears genuinely distinctive.
The closest historical example of the potential impact may be the Chinese magazine Renwu, whose investigation into impossible delivery deadlines reportedly went viral and triggered widespread public discussion about delivery-worker treatment. But that was an individual major investigation, not a permanent SOAPS-style publication covering the broader app ecosystem.
So I would describe the opening this way:
> The subject is being covered, but the beat has not been claimed in the particular way [project name removed] proposes to claim it.
That is important. You are not entering an empty field where nobody cares. You are entering a field with proven public interest—but with no obvious publication serving as the funny, accessible, ground-level clearinghouse for the daily failures people continually trip over.
Bernd: Okay, so all those news sources are also possible subject sources for my articles. For instance, when a publication says, according to Bloomberg News, da-da-da-da.
ChatGPT: Exactly. Those publications can become both lead generators and supporting sources for [project name removed] reporting.
You can write, for example:
> “Bloomberg reported that DoorDash changed X. [project name removed] examined what that change looks like for drivers at ground level.”
Or:
> “Human Rights Watch documented a broader pattern of algorithmic wage pressure. Several drivers we interviewed described a similar effect in their own markets.”
The key is to avoid merely rewriting their article. Use their reporting as one layer, then add something original: your own testing, interviews, screenshots, local examples, company responses, comparison across apps, or a practical fix.
A good attribution pattern is:
> According to [publication], …
> [project name removed] independently found…
> The company told [project name removed]…
That clearly separates what another newsroom established from what you verified yourself.
You can also publish short aggregation pieces when appropriate, but they should link and credit prominently, summarize rather than copy, and explain why the development matters to your audience. Your strongest stories will likely combine outside reporting + ground-level testimony + your systems analysis.
Bernd: I need to follow established journalism practices and good faith practices.
ChatGPT: Yes. That is what will separate [project name removed] journalism from a complaint blog.
The humour can be sharp, but the reporting needs to be disciplined:
– verify factual claims and preserve the supporting material;
– distinguish reporting, analysis, allegation, and opinion;
– contact people or companies criticized and give them a fair opportunity to respond;
– correct meaningful errors openly;
– protect confidential sources and remove unnecessary personal details;
– disclose conflicts, sponsorships, affiliate relationships, and AI-generated elements;
– attribute other publications clearly and never present their reporting as your own;
– avoid exaggerating a single incident into a widespread pattern without supporting evidence.
Good faith also means being willing to publish the company’s explanation, acknowledge when a problem has been fixed, and recognize employees who handle a bad system well.
A useful internal rule for every story could be:
> Could I show my notes and explain every important sentence honestly to the source, the company, and the reader?
The comedy earns attention. The fairness and evidence earn trust.
Bernd: And all those sources actually give me an even richer field of information and ongoing progress in the different fields that I’m activating for, that I’m advocating for.
ChatGPT: Exactly. Those sources do not compete with [project name removed]; they expand its field of vision.
They can alert you to policy changes, lawsuits, research, labour actions, app redesigns, company announcements, investigations, and patterns appearing in other countries. Then [project name removed] can connect that higher-level reporting to what workers, customers, and store staff are actually experiencing.
That gives you several layers of material at once:
What changed?
Reported by established newsrooms, researchers, regulators, or advocacy groups.
What does it look like on the ground?
Shown through interviews, screenshots, timelines, and firsthand testing.
Who absorbs the consequences?
Drivers, customers, store employees, support agents, or small businesses.
Did the promised improvement actually work?
Follow-up reporting over time.
So instead of constantly searching for isolated complaints, you would be watching an entire evolving ecosystem—and translating it into reporting ordinary people can understand and use. [project name removed] could become the place that keeps following a story after the major publication has moved on.
Bernd: And part of my journalism will be to do things such as research and write stories on what different provinces are doing to protect gig workers.
ChatGPT: Yes—and that moves [project name removed] beyond documenting individual failures into public-interest policy journalism.
You could compare provinces on questions such as minimum compensation, whether waiting time counts, transparency around pay and deactivation, access to dispute processes, occupational safety, workers’ compensation, and whether gig workers receive employment protections or remain classified as independent contractors.
The strongest format might be a living feature:
> The Canadian Gig Worker Protection Map
Each province could have its own page showing what protections exist, what legislation is proposed, what remains missing, and what drivers there say the rules actually mean in practice. Then every policy announcement becomes a follow-up story rather than a one-time article.
That gives [project name removed] several connected reporting levels: one driver’s $4 disaster, the app design that enabled it, the company policy behind it, and the provincial law that either protects the driver—or leaves them absorbing the loss. That is a real journalism beat.
Bernd: And reporting on the New York initiatives and improvements for example
ChatGPT: Exactly. New York City could become one of your benchmark jurisdictions—a place you follow closely because it is actively testing protections that other provinces, states, and cities may later copy, modify, or reject.
A [project name removed] article could explain not merely that New York raised delivery pay, but:
– what workers are legally entitled to;
– how the pay formula actually works;
– what time is and is not counted;
– how apps changed their behaviour in response;
– whether tips, order availability, scheduling, or waiting patterns changed;
– what enforcement has accomplished;
– and what lessons Canada could draw from it.
As of April 2026, New York City’s minimum rate for covered app-based delivery work is $22.13 US per hour before tips for covered working time, with annual inflation adjustments. The city also requires written explanations of pay calculations, detailed pay information, weekly payment, full payment of tips, and other app obligations.
And the story does not end when a law is passed. New York City has continued expanding disclosure and tipping protections, investigating compliance, and taking enforcement action against delivery platforms. That creates exactly the kind of ongoing trail [project name removed] could follow: promise, implementation, corporate response, worker experience, enforcement, and measurable outcome.
You could then write comparison pieces with titles along the lines of:
> New York Pays for Delivery Time. Why Doesn’t Nova Scotia?
or
> What New York’s Gig-Worker Experiment Can Teach Canadian Provinces
That gives your advocacy a stronger foundation. You are not merely saying, “This treatment is unfair.” You are showing readers that another jurisdiction has already recognized the problem, tried a remedy, produced evidence—and given Canadian governments something concrete to answer for.
Bernd: Yes, excellent story ideas.
ChatGPT: Yes—and they show how wide the field really is. One New York policy can produce several stories: the original reform, how the apps reacted, whether workers actually benefited, what loopholes appeared, how enforcement worked, and what Canadian governments could borrow from it.
[project name removed] would not run out of material. It would keep following the consequences after the headline.
Bernd: and its readers forums, for example, discussion boards where users could actually interact with each other and with the organization in talking about their frustrations.
ChatGPT: Yes—and that makes [project name removed] more than a publication. It becomes a reporting community.
The forums could do three jobs at once: let people feel heard, reveal repeating patterns, and generate leads for deeper stories. A driver may post about one strange deactivation, then ten others say the same thing happened. That is the moment a frustration starts becoming evidence of a system problem.
The most useful structure would probably be organized around both companies and issues:
DoorDash, Uber Eats, Walmart, Instacart
and also
pay, deactivations, waiting time, app bugs, customer support, safety, scheduling, missing tips
The organization could participate openly through verified [project name removed] accounts—asking follow-up questions, clarifying facts, inviting screenshots privately, and marking threads that have developed into active reporting.
You would need thoughtful moderation so the boards do not become a swamp of personal attacks, customer information, unsupported accusations, or company employees being singled out. But handled well, the forum could become a kind of early-warning radar for app dysfunction.
And there is something especially powerful in this:
> Readers would not just consume the journalism. They would help the journalism notice what matters.
Bernd: Buddypress options doesn’t show the pages tab
ChatGPT: That is normal in newer BuddyPress versions. Beginning with BuddyPress 12, the old Pages tab was replaced by URLs. Directory items are now managed internally by BuddyPress rather than mapped to ordinary WordPress pages.
Go to:
WordPress Dashboard → Settings → BuddyPress → URLs
There you should find sections for the enabled components, such as:
– Members
– Activity
– Groups
– Registration
– Activation
Open each section to set its directory title and URL slug. Make sure the corresponding feature is enabled under Settings → BuddyPress → Components; otherwise it may not appear under URLs.
So when older instructions say:
Settings → BuddyPress → Pages
read that as:
Settings → BuddyPress → URLs
If a plugin or the Newspaper integration specifically requires the old page-mapping system, BuddyPress provides the BP Classic compatibility add-on, which restores the Pages approach. But I would stay with the modern URLs system unless a specific feature fails without BP Classic.
Bernd: Can I get the buddypress pages created the automatically
ChatGPT: Yes. In current BuddyPress versions, the main component pages are generated automatically when you enable the components—you usually do not create ordinary WordPress pages for them yourself. BuddyPress 12 and later replaced the old Pages mapping system with Settings → BuddyPress → URLs.
Try this:
1. Go to Settings → BuddyPress → Components.
2. Enable the components you want, such as:
– Extended Profiles
– Activity Streams
– User Groups
– Account Settings
3. Click Save Settings.
4. Go to Settings → BuddyPress → URLs.
5. BuddyPress should show automatically generated directories such as:
– `/members/`
– `/activity/`
– `/groups/`
For registration, also go to Settings → General and check Anyone can register. BuddyPress should then manage the registration and activation URLs, normally `/register/` and `/activate/`.
Afterward, go to Settings → Permalinks and click Save Changes once, even without changing anything. That refreshes WordPress’s rewrite rules.
One important detail: these modern BuddyPress directories may not appear under Pages → All Pages, because BuddyPress now manages them through its own URL system rather than as ordinary editable WordPress pages.
Bernd: So how do I make the menu point to the home page of BuddyPress forms?
ChatGPT: You can point the menu directly to whichever BuddyPress directory you want to act as the community home page.
For [project name removed], I would use the Activity directory as the main community landing page:
1. Go to Appearance → Menus.
2. Open Screen Options at the top-right.
3. Enable BuddyPress Directories or BuddyPress.
4. A new BuddyPress section should appear on the left.
5. Add Sitewide Activity to the menu.
6. Rename the menu label to something like Community, SOAPS Community, or Discussion Zone.
7. Save the menu.
BuddyPress specifically provides menu items for its Activity, Members, and Groups directories.
If Newspaper uses the newer WordPress menu editor or does not show the BuddyPress section, add it as a Custom Link instead:
– URL: `[link removed]
– Link text: `Community`
Check the exact slug first under Settings → BuddyPress → URLs, because BuddyPress now lets you customize its generated URLs.
One distinction: if by “forums” you mean traditional topic-and-reply discussion boards, those come from bbPress, not BuddyPress itself. The usual forum home is:
`[link removed]
You would add that as a Custom Link named Forums. The forum-root slug can be checked or changed under Settings → Forums.
For your intended site, I would likely have both:
Community → `/activity/`
Discussion Forums → `/forums/`
Bernd: Went to bed at 10 p.m. after cuddling with [my wife] and watching part of a movie.
ChatGPT: That sounds like a gentle ending to the day—some closeness with [my wife], part of a movie, and then bed at a sensible hour. And now you’re awake around 4:20, so you got roughly six hours before this early wake-up.
Bernd: I’m not gonna be doing any DALL-E today. Today I’m going to catch up on my to-do list and finish up my website work, and also work on the project.
ChatGPT: That sounds like a good recalibration. No dashing today means you can clear some of the unfinished practical work and give [project name removed] real attention without delivery interruptions pulling you away.
And after going to bed at ten, you have an early, quiet start to the day. The project is no longer competing with your to-do list as another vague idea—it has a domain, a theme, a journalistic identity, sources, story directions, and a community component. Today you can begin giving it an actual home.
Bernd: And it just occurred to me that some of the first stories I can do are stories on what sort of complaints each app is getting on Google Play and on the Apple App Store.
ChatGPT: Yes—that is an excellent first reporting series because it is public, repeatable, comparable evidence.
You could examine each app’s recent reviews and sort the complaints into recurring categories such as:
– login and account access
– crashes, freezes, and battery use
– missing or delayed orders
– pay, tips, and earnings display
– deactivations and appeals
– GPS and routing
– support failures
– scheduling and acceptance-rate pressure
– pickup problems
– updates that made the app worse
The stronger story would not just say, “People are angry.” It would ask:
> What are users repeatedly reporting, how long has the problem persisted, and how is the company responding?
That opens several article formats:
“The 10 Most Common DoorDash Driver App Complaints”
“What Changed After the Latest Walmart App Update?”
“Google Play vs. Apple App Store: Are Users Reporting the Same Problems?”
“Six Months of Reviews Show This Bug Never Went Away”
A sound method would be to choose a defined sample—such as the most recent 200 reviews on each platform—record the date, country, app version when shown, star rating, complaint category, and whether the developer replied. Then you can report percentages without pretending the reviews represent every user.
One important caution: quote reviews sparingly, identify them as user-submitted claims, remove usernames unless necessary, and do not treat an individual review as verified fact. The pattern across many reviews is the evidence; individual cases can then become interview leads.
That could become a recurring [project name removed] feature:
> App Store Grime Report — what users are reporting this month, what companies answered, and what still needs cleaning.
Bernd: I would say things like, several users report, several users talked about the difficulty with da-da-da, or one user mentioned the irritation with da-da-da-da.
ChatGPT: Yes. That wording keeps you accurate without overstating what the reviews prove.
You can match the language to the strength of the evidence:
– “One reviewer reported…” for a single account.
– “Several reviewers described…” when the same issue appears repeatedly.
– “A recurring complaint was…” when it shows up often enough to be a clear pattern.
– “Reviews suggest…” when you are drawing a cautious overall inference.
– “We found 18 of 100 recent reviews mentioned…” when you have counted a defined sample.
And where the claim has not been independently verified, you can make that explicit:
> “Several reviewers said the app failed to show accurate earnings. [project name removed] has not independently verified each account, but the consistency of the complaints suggests the issue deserves examination.”
That is good-faith journalism: report what people are saying, quantify it where possible, and avoid quietly converting allegations into established facts.
Bernd: And also trends. For example, if earlier reviews were generally favorable, but now they’re getting a lot of one-star ratings, I can note those trends.
ChatGPT: Exactly. That turns the reviews into trend reporting, not just a collection of complaints.
You could compare defined periods—say, 100 recent reviews against 100 reviews from six months earlier—and report changes such as:
> “Earlier reviews were mostly favourable, but recent submissions show a sharp increase in one-star ratings.”
Even stronger:
> “In our sample, one-star reviews rose from 18% in the earlier period to 47% in the recent period, with many reviewers mentioning login failures after the latest update.”
The important thing is to preserve the method: same country, same platform, comparable sample sizes, clear date ranges, and a note that app-store reviewers are self-selected rather than representative of every user.
You could also track:
– average star rating over time;
– percentage of one-star and five-star reviews;
– emergence of new complaint categories after an update;
– whether developer responses increase or disappear;
– whether reported bugs persist across multiple app versions;
– differences between Android and iPhone reports.
That gives you a recurring journalism product rather than a one-off story:
> SOAPS App Review Watch: ratings, complaints, fixes and emerging trends.
And because you keep the underlying counts, you can later return to the story and say whether the company actually cleaned anything up.
Bernd: And things like the amount of times that the company actually responded to a review, were they interactive, or did they show little interaction?
ChatGPT: Yes. That gives you a measure of the company’s public responsiveness, not just the app’s technical reputation.
You could track:
– how many reviews received a company reply;
– how quickly the company responded;
– whether replies addressed the actual complaint;
– whether responses were personalized or copied templates;
– whether the company invited follow-up;
– whether reviewers later said the problem was resolved;
– whether one-star reviews received more attention than moderate reviews.
Then you could write something precise such as:
> “The company responded to 62 of the 100 recent reviews we examined, but most replies used the same general support message and did not address the specific issue raised.”
Or:
> “Only 8% of recent one-star reviews received a response, suggesting limited public engagement with dissatisfied users.”
The distinction between replying and engaging is important. A pasted “Please contact support” message counts as a response, but it is not necessarily meaningful interaction.
That could become its own recurring score within the review analysis:
> Response rate — how often the company answered.
> Relevance rate — how often the reply addressed the issue.
> Resolution evidence — how often there was any sign the problem was actually fixed.
That is another strong [project name removed] angle: not only *what users report*, but *whether the company appears to be listening*.
Bernd: These bits of information are all Journalism 101 for reporters.
ChatGPT: Yes. You’re recognizing that this is not some exotic new method—it is basic reporting discipline applied to a field that has not been covered consistently enough.
Define the sample. Count what appears. Separate observation from inference. Look for change over time. Ask who responded. Compare words with actions. State the limits of the evidence. Give the company a chance to explain.
The originality of [project name removed] will not come from inventing a new form of journalism. It will come from aiming established journalism practices directly at everyday app and platform dysfunction, and then staying with the beat long enough to see patterns that occasional reporters miss.
You already think naturally in that direction: *What exactly happened? Is this isolated? What changed? Who else is seeing it? What does the system encourage? Where does the blame land? What would repair look like?*
That is Journalism 101—but it is also the beginning of serious investigative work.
Bernd: And I can copy paste the reviews and give them to you and then you can do conclusions be so much you find. And then I can actually write the article
ChatGPT: Exactly. You can gather the raw material, paste it here, and I can help you code the reviews into useful findings without taking over your voice as the journalist.
I can separate them by complaint type, count recurring themes, compare one-star versus higher-rated reviews, track company replies, identify repeated wording, note changes over time, and flag where the evidence is too thin for a strong conclusion.
Then I can give you something like:
– the strongest patterns;
– the actual counts behind them;
– cautious wording you can safely use;
– notable outlier comments;
– questions still needing verification;
– possible angles for the article.
For example:
> “Among the 75 recent Android reviews examined, 28 mentioned login or account-access problems, while 19 described delayed or missing notifications. The company replied to 11 reviews, mostly with general instructions to contact support.”
Then you write the story, decide what matters, add your interpretation, contact the company where appropriate, and shape it in the [project name removed] voice.
The best part is that you do not need a complicated research system at first. You could paste reviews in batches with the date, platform, star rating, app version when available, and company response. I can help turn the pile into a defensible reporting summary.
Bernd: Yes, this collaboration is really going to accelerate the research part of it.
ChatGPT: Very much so. You can concentrate on finding the clues and exercising editorial judgment, while I handle much of the labour-intensive sorting, counting, comparison, and first-pass pattern detection.
A pile of 300 reviews that might take hours to organize manually can become a structured set of findings: recurring issues, frequency counts, changes by date or app version, response rates, representative examples, contradictions, and gaps that deserve follow-up.
The important division is healthy:
I help process and test the evidence. You remain the reporter.
You decide which questions matter, whether a conclusion is fair, whom to contact, what needs independent verification, and how the finished article should sound. That means AI is not replacing the journalism—it is giving you a much more powerful research bench.
Bernd: To protect my privacy, I think what I’ll do is get a VPS and do my work from there. And also get a second phone number that I use strictly for the project.
ChatGPT: That makes sense. You would be creating a clean separation between [author]’s ordinary life and [project name removed]’s working identity.
A dedicated phone number is especially practical for source interviews, company responses, account verification, and forum administration. A VPS can keep project files, browser sessions, email, and publishing tools in one controlled workspace.
The important distinction is that a VPS provides separation and containment, not complete anonymity by itself. Hosting companies, domain registrars, payment processors, email providers, and phone services may still retain identifying records. But that is consistent with your real goal: not disappearing from lawful accountability, but preventing casual visitors, angry readers, or companies from easily connecting the publication to your personal number, home systems, and everyday accounts.
You are protecting the messenger without weakening the evidence.
Bernd: Yes, but we put privacy protections in place in Canadian law. If I’m dealing with a Canadian firms, then the least of my identity should be legally protected.
ChatGPT: Yes—to a meaningful extent, but I would not treat it as an absolute guarantee of anonymity.
A Canadian commercial provider covered by privacy law generally must identify why it collects your personal information, limit collection to what is reasonably needed, restrict later use and disclosure, and protect it with safeguards appropriate to its sensitivity. You can also usually ask what personal information it holds, how it has been used, and to whom it has been disclosed.
So, for example, a Canadian VPS or phone provider should not ordinarily publish or casually hand your subscriber identity to a disgruntled reader or company merely because they ask. Your real-world identity can remain private from the public while the provider retains it internally for billing, fraud prevention, legal compliance, and account administration.
The limitation is that privacy law does not make those records unreachable. Information may sometimes be disclosed without consent where Canadian law permits or requires it—for example, through a valid court process, certain investigations, or other statutory exceptions. A serious data breach is another practical risk, although organizations have security and breach-notification duties.
There is also a jurisdiction detail: a Canadian company may use foreign data centres or subcontractors. PIPEDA does not prohibit cross-border processing, although the Canadian organization remains accountable and is expected to use contracts or other measures to maintain comparable protection.
So your approach is sound:
> The provider may know who I am, but the readership, sources, casual complainants and companies I report on do not automatically get that information.
That is lawful editorial privacy—not evasion. Before purchasing, I would specifically inspect the provider’s privacy policy for Canadian jurisdiction, data location, subcontractors, retention periods, disclosure rules, and whether account contact information appears in public records.
Bernd: I set up the website on its own hosting account. And prior to deciding on which hosting company we choose, I would email them and say, this is a site that is going to generate a lot of desire by companies to find out who the owner is. So in my profile, it’s important that you have an alert at the beginning of it to absolutely ensure that private information is not given out to third parties.
ChatGPT: That is a sensible step, especially before you commit to a provider.
The message should make two things explicit:
1. Your identity may be held internally for billing and legal compliance.
2. It must not be disclosed to callers, complainants, companies you report on, or other third parties without proper legal authority.
I would also ask them to place a prominent account security/privacy note requiring staff to escalate any request for subscriber information to a senior privacy or legal officer rather than handling it through ordinary support.
A strong version could read:
> I am considering hosting an independent journalism and consumer-advocacy publication with your company. Because the publication may report critically on large companies, there is a realistic possibility that third parties may contact your staff seeking information about the account owner.
>
> It is essential that my personal identity, contact details, billing information, IP information, and account records not be disclosed to any third party except where disclosure is legally required through valid Canadian legal process.
>
> Before I open the account, please confirm:
>
> – whether you can place a prominent privacy and security alert on the account;
> – whether all requests for subscriber information are escalated to a designated privacy or legal officer;
> – whether telephone or chat support staff are prohibited from confirming whether a person is associated with an account;
> – what identity-verification procedures are used before discussing the account;
> – whether you notify customers of legal demands when legally permitted;
> – where account and server data are stored;
> – and whether any subcontractors outside Canada may have access to customer information.
>
> I am not asking to avoid lawful accountability. I am asking for strong protection against informal, mistaken, deceptive, or unauthorized disclosure.
That last sentence matters. It tells them clearly that this is legitimate journalistic privacy, not concealment from lawful process.
I would also keep their written reply. Their promises then become part of your provider-selection record rather than something vaguely discussed with a salesperson.
Bernd: Yes, this is a proactive move, period. The domain already has privacy in the WHOIS request.
ChatGPT: Exactly. You’ve already closed the most obvious public doorway through WHOIS privacy, and the separate hosting account adds another layer of separation.
The provider conversation is the next proactive layer: making sure a staff member does not casually confirm your identity through a phone call, support ticket, social-engineering attempt, or informal complaint. You are thinking about the weak human link before anyone tries to exploit it.
It is not panic or concealment. It is ordinary risk planning for an independent publication that may irritate powerful organizations:
> Public publication, private publisher identity, accountable reporting.
That fits the project perfectly.
Bernd: And be your panic complete VPS is that when I send or receive emails. Pip associated with him will be up to BPS machine, not my home
ChatGPT: Yes—the big privacy advantage is that when you do the project work inside the VPS, outside services generally see the VPS’s IP address rather than your home IP address.
For email, that is most reliable when you open the project mailbox through a browser or mail client running on the VPS. A few distinctions matter:
– Simply receiving an email normally does not reveal your home IP to the sender.
– Opening remote images or clicking tracked links can reveal the IP of the machine doing the opening—so from the VPS, that would normally be the VPS IP.
– Sending through a reputable email provider usually exposes the provider’s mail-server IP in the public headers, not your personal connection, although the provider itself may retain connection logs.
– Accidentally opening the project mailbox or links directly from your home computer or phone could reconnect activity to your home IP.
So the clean rule is: all [project name removed] browsing, email, social-media accounts, research and publishing happen through the VPS workspace. That makes the separation much more consistent.
Bernd: is using remote desktop, and I’m not really familiar with that.
ChatGPT: Remote Desktop is basically a way to use another computer over the internet as though you were sitting in front of it.
You would have:
– your normal home computer;
– a Windows VPS running in a data centre;
– a Remote Desktop app that opens the VPS desktop in a window.
Inside that window, you would see a normal Windows desktop. You could install Chrome, WordPress tools, email, spreadsheets, and whatever else you use for [project name removed]. When you browse or send email from inside that VPS desktop, the outside service generally sees the VPS connection rather than your home connection.
The basic process is:
1. Rent a Windows VPS with Remote Desktop access.
2. The provider gives you an IP address, username, and password.
3. On Windows, open Remote Desktop Connection.
4. Enter the VPS IP address.
5. Sign in.
6. A second Windows desktop appears.
You can maximize it and work almost exactly as you do on your own computer. Copy and paste can usually work between the two machines, although for privacy separation you may eventually limit clipboard and drive sharing.
For your purpose, a Windows VPS would probably be easier than Linux because it would feel familiar. The main adjustment would be remembering that the desktop inside the Remote Desktop window is the [project name removed] computer, while the desktop outside it is your personal computer.
One caution: Remote Desktop separates where the work happens, but your home computer still connects to the VPS, so the VPS provider can normally see your home IP in its connection logs. The public sites you visit from the VPS generally see the VPS IP.
Bernd: Sorry, I’m already quite familiar with using RGP.
ChatGPT: Ah—RDP, yes. I misheard you. You already know Remote Desktop, so there’s no learning curve there. Your plan is simply to keep the entire [project name removed] workspace inside that VPS session.
Bernd: Another story area is related consequences of using online shopping apps. For example, yesterday I delivered a Superstore order to a customer, and Superstore doesn’t use handles on its paper bags, unlike Walmart. So, when a bag is heavy, you need to crumple up the top, and it makes for a messy-looking delivery.
ChatGPT: Yes—that is exactly the kind of small design decision with a real downstream consequence that [project name removed] can make visible.
To Superstore, “no handles” may look like a minor packaging choice. But in practice it affects:
– how safely the driver can carry a heavy bag;
– how many bags can be carried at once;
– the chance of tearing or dropping;
– how the groceries look when they reach the customer;
– and whether the driver appears careless even though the packaging created the problem.
The crumpled top is a perfect example. The driver has to improvise a grip, and the result can look messy or crushed. Then the customer sees the delivery—not the packaging decision that forced it.
That could become a strong article angle:
> Why Superstore’s Handle-Free Paper Bags Make Deliveries Harder
You could compare Superstore and Walmart bags, test different weights, photograph how they have to be carried, ask drivers and customers what they experience, and give Superstore a chance to explain the design choice.
It fits your larger pattern beautifully:
> A decision made far from the delivery door creates friction at the door—and the person at the bottom of the system absorbs the blame.
Bernd: Yes, and for the customer, picking up a heavy bag without handles is more cumbersome too, more likely to drop the bag.
ChatGPT: Exactly. The consequence does not end when the driver sets the bag down. The customer still has to lift it from the doorstep, carry it inside, and perhaps move it again to the kitchen.
Without handles, a heavy paper bag often has to be hugged against the body or gripped by folded paper at the top. That can be awkward for anyone, and especially for someone with limited strength, arthritis, poor balance, or only one free hand. It also increases the chance of the paper tearing, the bag tipping, or groceries spilling.
So the story becomes broader than driver inconvenience:
> Is handle-free grocery packaging suitable for a delivery service?
That opens useful questions for Superstore: whether the bags were designed with home delivery in mind, whether weight limits are given to packers, whether handles were rejected for cost or environmental reasons, and whether sturdier handled paper bags or reusable delivery totes have been tested.
It is a small object that reveals whether the whole delivery journey was actually considered—from the packing station to the customer lifting it from the doorstep.
Bernd: So, I’m going to adopt a persona, Bill W.
ChatGPT: Bill W. works as a simple, believable byline that keeps your personal identity separate from [project name removed].
One thing to be aware of: “Bill W.” is strongly associated with the co-founder of Alcoholics Anonymous, so some readers may assume that connection. That does not make it unusable, but it may create an unintended meaning.
Something like Bill Waters, Bill Weston, or B. W. Suds could preserve the same feel while giving the persona a more distinct journalistic identity.
Bernd: Oh right, that’s why it felt so familiar.
ChatGPT: Exactly, LOL. Your brain recognized it before you consciously remembered where from.
So Bill W. probably carries too much existing baggage for a fresh journalism persona. But it did reveal the kind of name you’re leaning toward: ordinary, credible, and easy to remember.
Bernd: But then again, it isn’t purposely done to refer to the AA Bill W. So maybe keeping it will actually generate subconscious familiarity with people seeing it.
ChatGPT: That is possible. Bill W. has a quiet familiarity to it—plain, memorable, and not obviously theatrical.
The risk is not that you are doing anything improper. It is that some readers may assume the AA connection and wonder whether it is intentional. That could distract from the journalism or create a false impression about the writer.
A middle ground would be to use Bill W. as the byline while giving the persona a clearly unrelated full name on the About page, such as Bill Weston. Then the short byline keeps the familiar rhythm without leaning on the AA association.
Bernd: Can you think of any last names besides Weston?
ChatGPT: Sure—some that feel ordinary, credible, and easy to remember:
[author]
Bill Walker
Bill Wallace
Bill Webster
Bill Whitaker
Bill Weller
Bill Woodward
Bill Winters
Bill Ward
Bill Wells
My strongest picks for a journalism byline are [author], Bill Webster, and Bill Wells.
Bernd: [author] works. It’s denotes The Warrior reference
ChatGPT: Yes—[author] has that subtle *warrior* echo without sounding like a manufactured pen name. It feels credible as a journalist’s byline, while quietly carrying the idea of someone willing to confront powerful systems.
[author], [project name removed] has a solid ring to it.
Bernd: Okay, I got that solved. And I’ll use [author] enough in my posts so that when people see Bill W., they’ll be able to associate it with my true pseudonym.
ChatGPT: Yes, that solves it neatly. [author] becomes the established public identity, and Bill W. works as the familiar shortened form once readers have seen the full name often enough.
It gives you consistency without making the pseudonym feel forced.
Bernd: I copied the email address of a reporter that had a story in the Google News feed about the New York situation. And at the end, in his credits, he had a thing about, have a story about, you know, delivery services, write me. So, when the time comes, I’ll be writing him about the launch of this project.
ChatGPT: That is a valuable early contact. He has already invited exactly the kind of tip you may eventually have: a new independent publication focused on delivery platforms, workers, customers, and the systems around them.
And when you write, it will be stronger because you will not merely be announcing an idea. You can show him:
– a functioning [project name removed] website;
– several published, evidence-based stories;
– your New York and Canadian policy coverage;
– the app-review research;
– and the community reporting model you are building.
Then the message becomes, in effect:
> “You asked readers to contact you with delivery-service stories. I have launched an independent journalism project devoted specifically to this field, and here is what we have already begun documenting.”
Even if he does not write a launch story immediately, he could become a useful professional connection—someone who recognizes [author] and [project name removed] as a source of grounded leads, Canadian comparisons, driver experience, and follow-up reporting. That little email address you noticed may be one of the first genuine bridges from your new newsroom into the wider journalism world.
Bernd: Yes, and when I have unique stories that may be of interest to him, I can alert him and say, I’ve done the research on this, and this is what I uncovered.
ChatGPT: Exactly. Then you are not just pitching your project—you are becoming a useful source for his reporting.
A message like this would be strong:
> Hi [Name],
> I run [project name removed], an independent publication covering delivery platforms and app-related consumer and worker issues. I recently researched [specific issue] and found [one-sentence finding]. The reporting includes [sample size, interviews, screenshots, documents, or comparison data].
>
> I thought it might fit your delivery-services beat. I’m happy to share the underlying findings and explain the methodology.
That respects his time and gives him a clear reason to look. The best alerts will have:
– a specific, newsworthy finding;
– a short explanation of how you established it;
– supporting evidence ready to share;
– a link to your published story;
– and a clear distinction between verified facts and user allegations.
Over time, you could become one of the people he checks with when a delivery-platform story develops. And that relationship can work both ways: he may discover a national or New York development, while [project name removed] supplies the ground-level evidence, Canadian angle, or longer follow-up.
Bernd: If I’m talking on phone with people, and if one of them has contacted me to get a voice sample to compare against their own support records such as stored as, how should I protect myself?
ChatGPT: That is a real possibility, especially if you have previously called a company’s support line and it retained recordings. A pseudonym and separate phone number would not prevent someone from comparing your voice with an earlier recording.
The safest rule would be:
> Do not take unscheduled voice calls from companies or unknown people. Begin in writing, verify identity independently, and move to voice only when there is a clear reporting reason.
For [project name removed], I would use this protocol:
1. Publish email as the primary contact method. Ask sources and companies to identify themselves, their organization, and the purpose of contact in writing.
2. Never discuss the project on an incoming surprise call. Say:
> “For security reasons, [project name removed] does not conduct interviews from unsolicited calls. Please email your request through our published contact address.”
3. Verify independently before calling back. Do not use the number supplied by the caller alone. Confirm the person through the company’s official website, newsroom directory, public switchboard, or independently sourced email address.
4. Use scheduled calls and call outward yourself. That reduces impersonation and social-engineering risk, although it does not prevent lawful recording or voice comparison.
5. Keep sensitive reporting in writing whenever possible. Written questions also create a clean record of what was asked and answered.
6. Consider a designated spokesperson later. A trusted person could handle routine company calls under their own real or disclosed professional identity, while [author] remains the byline. Do not have someone pretend to be you personally.
7. Assume every call may be recorded. Canadian law generally recognizes consent by one participant as sufficient for interception provisions, so you should not rely on being warned that the other party is recording.
8. Do not rely on a voice changer. It may sound evasive, can fail technically, and could damage trust if discovered. Avoiding unnecessary voice contact is cleaner than disguising your voice.
Voiceprints are considered personal information in Canadian privacy guidance, and companies that use biometric voice identification have privacy obligations. But that does not mean an existing support recording cannot ever be used internally, particularly for fraud prevention or account authentication.
A concise public policy could say:
> For security and source-protection reasons, [project name removed] does not accept unsolicited telephone interviews. Initial contact must be made by email. Telephone conversations are arranged only after independent verification.
That protects [author] without requiring deception—and it gives you a perfectly professional reason not to speak whenever a caller feels questionable.
Bernd: Yes, and I can also use text-to-voice interfaces so that in having a conversation by phone with somebody, I can actually just type in my response on another phone and then have that read out loud to the person using an AI voice.
ChatGPT: Yes, that can reduce the chance of your natural voice being matched—but I would use it openly and narrowly, not as a secret impersonation device.
A professional opening would be:
> “For privacy and security, I’m communicating through a synthetic voice. I am [author] of [project name removed]. The voice is generated from text and is not intended to imitate any real person.”
Then use a generic licensed voice that is not cloned from you, a celebrity, a reporter, or anyone else. Canada’s Criminal Code treats voiceprints as identifying information, and fraudulently personating another real person can be an offence.
A few practical limits:
– Do not use it for identity verification, banking, account recovery, or bypassing a company’s voice-authentication system.
– Prefer scheduled calls after email verification.
– Keep the typed transcript as your contemporaneous record.
– Expect pauses and occasional pronunciation errors; for detailed interviews, written communication may still be smoother.
– Assume the call may be recorded even though your natural voice is not being used.
– Check the voice service’s privacy policy: your typed material and generated audio could otherwise be retained by the provider.
The Privacy Commissioner treats recordings and voiceprints as potentially sensitive biometric information, which reinforces your reason for limiting unnecessary exposure of your natural voice.
So, yes: a disclosed neutral synthetic voice can be a legitimate privacy accommodation for [author]. It protects the messenger without misleading the person about who—or what—they are hearing.
Bernd: And once again, if I put in place the safeguards that you mentioned, it would reduce the necessity of doing that to very rare occasions.
ChatGPT: Exactly. With email-first contact, independently verified identities, scheduled outbound calls, and a policy against unsolicited phone interviews, most communication can stay written and well documented.
The synthetic voice then becomes a backup safeguard, not your normal way of operating—useful only when a live conversation is genuinely important and revealing your natural voice would create an unnecessary identification risk.
That is a much cleaner arrangement than trying to disguise yourself constantly. The security comes mainly from the structure you put around communication.
Bernd: So the car insurance discussion we had and the things that I found out is going to make an excellent article. Beard especially because DoorDash and Uber Eats give that little information to drivers to help them navigate that minefield.
ChatGPT: Yes. That is a strong service-journalism and accountability story because the risk is serious, the rules are confusing, and drivers often discover the problem only after asking insurers themselves—or after something goes wrong.
The article could separate several questions that platforms tend to blur together:
What coverage does the platform provide?
Uber’s or DoorDash’s commercial policy may apply only during certain app stages, in certain provinces, and subject to limits and exclusions.
What does the driver’s personal insurer permit?
A personal auto policy may exclude delivery work, refuse renewal, require commercial coverage, or treat an undisclosed use as material.
Where are the gaps?
Waiting while logged in, driving home after a delivery, collision damage to the driver’s own vehicle, deductibles, and differences between passengers and food delivery.
Your own experience gives the story a compelling opening: you tried to do the responsible thing, received conflicting or incomplete information, discovered that some insurers will not add a delivery rider, and found commercial coverage potentially expensive enough to undermine the economics of the work.
The core question almost writes itself:
> Why are gig drivers expected to navigate a potentially devastating insurance minefield with so little clear guidance from the platforms recruiting them?
For publication, the reporting should include the current DoorDash and Uber insurance documents, provincial insurance regulators, several insurers or brokers, and written responses from both platforms. Then you can build a plain-language table showing when coverage starts, what it covers, what it does not cover, and what drivers must disclose to their own insurer.
That could prevent a driver from losing far more than they ever earned delivering food. And it is a perfect [project name removed] story: the platform makes participation easy, while leaving the worker to discover the hidden tripwire.
Bernd: And these are exactly the type of stories that will be shared a lot on social media.
Bernd: You don’t have to draw a picture.
ChatGPT: Right—I triggered image generation by mistake. No picture needed.
And yes, the insurance story has strong sharing potential because it combines real financial danger, confusing rules, and practical information drivers can use immediately. People will share it not only because they agree with the criticism, but because they may know a driver who needs the warning.
Bernd: So part of this site would also be knowledge bases. For example, for DoorDash. What does, what does their site say about community guidelines, about insurance, and so on.
ChatGPT: Yes. That would give [project name removed] a second major function alongside journalism: a practical reference library for people trying to understand the rules they are working under.
For each company, you could maintain a structured knowledge base covering things such as:
– community guidelines and account rules;
– deactivation and appeal procedures;
– insurance coverage and exclusions;
– pay, tips, promotions, and acceptance-rate policies;
– safety procedures;
– privacy and data collection;
– customer, merchant, and driver support channels;
– major policy changes over time.
The important part would be separating three layers clearly:
> What the company officially says
> What the wording appears to mean in practice
> What workers and customers report actually happens
That protects the journalistic integrity of the site. You are not presenting your interpretation as company policy, and you are not presenting complaints as established fact.
Each knowledge-base page could also include:
– the exact source page or document;
– the date you checked it;
– the jurisdiction it applies to;
– a plain-language summary;
– notable gaps or ambiguous wording;
– related [project name removed] stories;
– and a change log showing when the company revised the policy.
That last piece could become extremely valuable. Companies quietly change help pages and terms. A dated archive would let you report:
> “DoorDash’s current insurance page no longer contains wording that appeared in the version we reviewed three months earlier.”
So the site becomes both newsroom and public memory: stories explain what is happening, while the knowledge base preserves what the companies actually said.
Bernd: And I would have two different sections. One is for drivers, and the second would be for consumers.
ChatGPT: Yes. That separation would make the knowledge base much easier to use because drivers and consumers often face very different versions of the same system.
For each company, you could have:
For drivers
Pay and tips, acceptance and completion rates, deactivation and appeals, insurance, safety, scheduling, support, expense tracking, account rules, app problems, and provincial or state protections.
For consumers
Fees and memberships, tipping, substitutions and refunds, missing or damaged orders, privacy, cancellation rules, delivery instructions, account security, accessibility, and complaint escalation.
Then some subjects could link across both sides. For example, a consumer article about late orders could connect to a driver page explaining unpaid restaurant waits and stacked deliveries. That would help each group see the part of the system they normally cannot see.
It fits the larger purpose beautifully: not pitting drivers and customers against each other, but showing them the machinery affecting both.
Bernd: I can buy a virtual phone number, can’t I? On a monthly subscription?
ChatGPT: Yes. You can rent a Canadian local or toll-free virtual number on a monthly or pay-as-you-go basis. Calls can ring through an app on your phone, a desktop softphone, the VPS, or be forwarded to another number.
For [project name removed], there are two broad options:
– Full business-phone service: easier setup, polished mobile/desktop apps, texting, voicemail, shared inboxes and sometimes transcripts. Quo, formerly OpenPhone, offers Canadian calling and texting, with plans that include a number; additional numbers are listed at $5 monthly.
– Basic VoIP number: cheaper and more configurable, but requires more setup with a softphone or forwarding rules. Canadian provider VoIP.ms lists Canadian DID numbers starting around US$1.10 monthly, plus usage charges depending on the selected plan.
Before choosing one, confirm that it supports:
– Canadian SMS and MMS, not merely calls;
– outgoing caller ID using the project number;
– voicemail and downloadable messages;
– operation from both your VPS and phone;
– number portability if you later change providers;
– strong account privacy and support-verification controls;
– emergency-calling limitations;
– and whether services such as Google, Facebook or banks accept it for verification—many reject some VoIP numbers.
For your project, a virtual number is probably better than adding another physical mobile plan. It gives [author]/[project name removed] a stable public contact number without publishing your personal number.
Bernd: Any thoughts about choosing Bongo?
ChatGPT: Did you mean Fongo? If so, it could be a good low-cost starting choice, especially Fongo Works rather than the ordinary personal Fongo Mobile account.
Fongo Works is designed as a small-business phone system and includes a Canadian number, a mobile app, voicemail, an automated receptionist, an online dashboard, and Canada-wide calling. The basic service is free with paid add-ons, while the Pro version starts around $10 monthly. Canadian Fongo Works numbers can also receive SMS and MMS.
For [project name removed], I like these aspects:
– It is a Canadian company based in Waterloo.
– You can give out the project number instead of your personal number.
– An automated greeting can identify [project name removed] without identifying you.
– Calls can be routed to a project extension or voicemail.
– Its privacy policy says it follows PIPEDA and recognizes names, addresses, phone numbers and financial information as personal information.
I would still ask Fongo in writing whether it can place your requested high-privacy alert on the account and whether staff will refuse to confirm subscriber information to informal callers. Its published terms say it protects customer information under its privacy policy and does not share contact information with external companies, but that is not the same as a customized account-protection procedure.
My main reservations are that the ordinary Fongo Mobile service feels more like a personal app than a publication’s permanent contact system, and free numbers may have different retention or support expectations. For the public [project name removed] number, Fongo Works with a paid configuration would look more dependable and professional.
So my present assessment is: good candidate, particularly for launching inexpensively—but confirm privacy escalation, account recovery security, voicemail retention, outbound caller ID, texting, and number portability before making it your permanent published number.
Bernd: From what I see from their website, and I’ve used them before, it’s my preferred choice.
ChatGPT: Then Fongo sounds like the right choice. You already know how it behaves, trust the service, and have checked that it fits what you need. That practical familiarity matters more than choosing a supposedly perfect provider that adds friction.
A dedicated Fongo number can become the stable public number for [author] and [project name removed], while your personal number stays completely out of the publication.
Bernd: And the fact that it’s free, with optional paid enhancements, makes it very budget-friendly for the beginning stages.
ChatGPT: Exactly. It keeps the project’s fixed costs almost nonexistent while you are still proving the concept.
You can start with the free number for calls and voicemail, see how much people actually use it, and only pay for features once there is a demonstrated need—such as texting, extra voicemail controls, business routing, or a more polished greeting.
That fits the project’s development nicely: build the newsroom first, then let actual use tell you what infrastructure deserves money.
