When a user asks for a new button or a custom dashboard, it's super tempting to just build it and make them happy. But half the time what they re asking for is just a band aid over a much bigger issue they re having with your app. If you just keep adding every requested feature your product becomes a bloated mess.

On the flip side if you question everything they say you risk sounding like you aren't listening to them. How do you decide when to just build what they asked for versus spending time investigating what's actually broken in their workflow?

I run a 3-level classifier (0 normal, 1 distress, 2 crisis) in front of a text generator. At Level 1 or 2 the app skips the generation, shows a support page, and pauses the feature. The final level is the max of three signals: a phrase filter, the classifier, and a level returned by the generation call itself. I tested it across several sets of typed messages: crisis, sarcasm, jokes, other languages.

Then, while testing on production, I tapped one of the three preset answer buttons ( I failed again ) and got the support page and a lockout. At the time, a Level 1 flag meant a 72-hour pause.

The cause: the preset labels went through the classifier exactly like typed text. Run alone, the classifier returned Level 1 on that label in 5 of 5 runs, while the grading call returned 0. Max of signals meant Level 1 won.

Why my evals missed it: every test row was something a person might type. None of them were the strings the app itself sends.

About a week ago, there was a case in my country where a woman took a Bolt taxi, which, instead of taking her to the designated place, started driving her to the Austrian border. In the end, the woman jumped out of the car to save herself.

[So the police were not very active, and Bolt just used only AI-generated answers, the company then announced it was a mistake in the GPS... but to be honest, I would say it was a kidnapping attempt].

Hey everyone,

I m curious to know: What are some of the apps, tools, or services you ve discovered that have truly transformed the way you work or build your startup/product, or basically helped you in day-to-day work productivity? It could be anything from project management, productivity hacks, automation, communication, or even wellness tools that help you stay focused.

In the last few days, I have seen many AI tools for creating audio-visual, and even people in my feed shared short movies created by AI (Hollywood will probably cut costs quite a bit in the next few years).

What is your experience with AI video generators, and which ones do you find the best?
(In terms of which AI tools have given you the best video results.)

Trusting user reports is so hard when your local tests and staging environment show a completely green checkmark.

For the longest time I used to assume the user was doing something weird on their end only to find out weeks later it was a subtle race condition.

Once you finally spot one of those ghost bugs, it completely changes how you look at support tickets. Do you usually ask users for video recordings right away, or do you try reproducing it blind first?

I'm a marketing tech builder. I've sold a startup, angel invested in a few, and am now building another.

Some things I believed 10 years ago turned out to be completely wrong. Others took me years (and quite a bit of money) to understand.

I've been building for developers for four years, and the pattern I keep seeing in my own product and in every dev tool I've watched succeed or die is this: developers don't switch tools because the new one has more. They switch because the old one wasted their time in a specific, annoying way, and the new one removed that specific annoyance.

"More features" is what vendors think developers want. "Less friction in the exact place I'm already annoyed" is what actually drives adoption.

The corollary: every feature you add is a new thing a developer has to evaluate before trusting you. The best dev tools I use are the ones that resisted adding things.

So my claim is: in developer tools, feature count is a cost, not a benefit. Curious whether people agree or think I'm overstating it. And if you build dev tools. what's the one feature you deliberately refused to build, and did it help or hurt?

Every launch guide says the same thing: go live at 12:01 AM Pacific so you get the full 24-hour window. But I've now watched enough launches to suspect that's a leftover from an older version of the platform, and that the algorithm and the audience have both changed.

What I actually observe: the first few hours are dominated by people in a similar timezone, most upvotes come from a few concentrated windows during the day, and a late-afternoon launch with a strong push can outperform a midnight launch with a weak one.

But I don't have real data I have vibes and a handful of launch pages I've stared at.

So: if you've launched more than once at different times, did the time actually move the outcome, or did the audience and prep dominate everything? And has anyone seen a launch succeed that went live at, say, 9 AM Pacific? I'm looking for counterexamples specifically.