AI can write my entire app but it can't get Microsoft to approve it

My Windows build has been in Microsoft's certification queue for three weeks. Three new models shipped in that time, two of them can supposedly do my whole job, and not one of them can get a human at Microsoft to press approve.

What's the longest you've been stuck on something you couldn't code your way out of?

65 views

Add a comment

Replies

Best
Getting 15 testers for the Google Play Store is a pain when trying to find people with Android devices willing to give a stranger their email and download an unverified app.

The silence already lied to you once, so it can't be read as pending.

That's the part Gal's visibility point and Asad's two clocks both route around. Setting it down works when no news means no news. Yours failed silently on attempt one, which means no news and dead look identical from where you sit, and that's why you keep opening the dashboard.

The thing you control isn't the waiting, it's the sampling. A fixed cadence you decide in advance, once a day at a set time, turns silence into a checked state instead of an open question. Same information, but the ambiguity stops living in your head all day.

Did the first rejection show up in the dashboard eventually, or did you only catch it because you happened to look?

 Only because I looked. The status flipped to failed in the dashboard, no reason attached, and no email ever arrived. It could have sat there another month and I'd have had no idea.

Fixed cadence is a good call. Right now I check whenever it crosses my mind, which is the worst of both.

Though a checked state has limits here. I know attempt one failed. I still don't know why, and that's what makes attempt two feel like a coin flip rather than a wait.

 The three weeks is the wait. The coin flip is a different problem.

Right now Microsoft is the only thing that can tell you whether the build passes, so running one test costs three weeks. That is what makes attempt two feel like chance instead of work. Most of what certification checks is mechanical though, and the Windows App Certification Kit runs that same class of check on your own machine in minutes.

Did that failed submission ever produce a certification report in Partner Center, or was the status flip genuinely the only artifact you got?

 Different process than you're picturing, worth clarifying for anyone else hitting this. My blocker isn't Store certification, so WACK doesn't apply. It's Azure Trusted Signing identity validation, Microsoft verifying my company is a real entity before they'll issue a code-signing cert. No Partner Center submission, no cert kit step, nothing runs on my machine.

To your question: no report. The only artifact was the status going to failed in the Azure validation dashboard. No reason, no email.

The WACK point is solid for an actual Store submission though. Mine is one step earlier, they won't sign until they've verified the company, and that verification is the black box.

app store review is the one that gets me every time. spent two weeks once waiting on Apple to re-review a payments-related update, got rejected for a guideline that wasn't even the reason they cited, resubmitted, waited again. no model helps with that because the bottleneck isn't code quality, it's a human on the other side reading a policy doc differently than you did. the annoying part is you can't even parallelize it, one submission at a time and the clock resets on every rejection.

 Same, and the reset is what actually costs you. Mine failed on the first attempt with no reason given and no email. I only found out by checking the dashboard. Resubmitted, back to zero, and I still can't tell if it's moving or silently dead again.

Everything downstream got faster and the approval step didn't move at all.

 no email on a rejection is brutal, at least give people something to act on. silently going back to zero with no diagnostic is worse than a slow yes. makes me think the real fix isn't a faster model, it's just better status visibility from Microsoft's side, a dashboard that actually tells you where in the queue you sit instead of pass/fail with no timestamp.

The Microsoft queue is a particular kind of stuck, because the skill that got you this far stops working. Everything else in a build responds to effort. You push, it moves. A certification queue does not care how good you are, and that is the part that itches. Not the three weeks, but that your competence suddenly buys you nothing.

The longest I have been stuck was never on a hard problem, it was always a dependency on someone who had no reason to hurry. What helped was separating the two clocks. There is work time, where effort maps to progress, and there is queue time, where the only honest move is to set it down and go build the next thing, so the wait runs in the background of a good day instead of the foreground of a stalled one. Three new models shipping while you sit there stings, but none of them are stuck where you are either, and that gap is usually a moat rather than a threat. The approval lands on its own schedule. The only thing you control is whether it gets to sour the weeks around it.

 Two clocks is right, and I've been doing it by accident. Shipped a scrolling capture feature and wrote a few posts while this sat there, so the wait has been background noise rather than the main event.

The part that still gets me is the silence. Work time gives you feedback even when it's bad news. The queue gives you nothing, so you can't tell whether you're waiting or whether it already failed weeks ago and nobody sent the email. Mine did exactly that on the first attempt. That's why I keep checking instead of leaving it alone.