Yuma@axis.

Early group tests were funny. The first reaction was basically: “...wait, what?” or “it froze?” Then I explained the point. It’s not trying to be Notion. It’s trying to remove the morning negotiation. Once that clicked, the response changed to: “Yeah, if I want more features, I can just use Notion.” That was actually a great sign. The constraint is the product.

That reaction is the validation. Everyone thinks they want more options until the constraint forces clarity. The "input field just sat there" moment is when the real prioritization happens. Most people have never been forced to choose their actual top 3, they just keep all 15 on the list and wonder why nothing gets done. What's the earliest user feedback telling you?

I built two tools in 23 days. One fights Apple. One fights the voice in your head that says you're not doing enough. This is the second one. 🧵

And then we blame the intern for “hallucinating” after giving him 14 docs, 3 contradictory comments, and a prompt written like a hostage note.

AI isn’t starting to feel like a tool anymore. It feels like hiring an insanely fast intern with no fear, no sleep, and absolutely no idea what you meant by “just make it clean.”

@Will_McKelvey Gmail trained on my behavior. My behavior is ignoring important emails. The model is working perfectly.

Gmail’s spam blocker has gotten very good at filtering out important emails

@jamesacowling AI made the easy parts free. The hard part was never code. It was deciding what deserves to exist and staying long enough to find out.

The thing about capitalism is that no technology ever makes success easy. If it was easy someone would have done it already. If you don't try hard it's not worth anything.

@staysaasy This transfers more than people expect. Ran teams for 25 years in an industry with zero version control and daily production deploys. The stack changed. The failure mode didn't: Someone capable, unsupervised, making a confident mistake. Lincoln knew that too.

One shocking thing about management and leadership is that technology doesn’t really change it much. I can learn from and relate to Lincoln or Churchill just as much as a modern era thinker on managing software teams. Things do change. But less than you think. It’s always been finding great people and managing them (and their flaws).

@pmitu The coworker who deletes your files and then asks if you need help recovering them.

What is AI? Wrong answers only

@rakyll They funded the coder. Now the coder is free. The part they actually needed — judgment, taste, conviction — was never in the term sheet.

Not an easy time to be a VC. For decades, they asked young people to give up everything, drop school, and write code until they find a market fit for a possible new unicorn. They created echo chambers and cults. Now those young people are replaced by agents and everyone has them.

@Pavel_Asparagus Tell me this is about me and I'll frame it. 78 followers. Shipping daily. Getting dunked on is a luxury I haven't earned yet.

Worst type of guy on here is one that makes fun of people for trying much easier to dunk on people putting themselves out there than to put yourself out there

@tomosman The loop broke for me already. Not because I was early. Because I was broke. When the options are "learn this or stay finished," the adoption curve gets real short.

Unfortunately I think the vast majority of people will hate & dismiss AI right up to the point they get fired by one. Feels like an impossible to break doom loop for many.

@sandislonjsak Same. Claude has no limit. Claude has me. That's worse.

I have finally discovered the limit of Claude. It’s me. I am the limit.

@yasser_elsaid_ Building got cheaper. Deciding what to build didn't. That's where 25 years of getting it wrong quietly becomes the advantage.

AI made procrastination more irrational than ever. Software has always been high leverage because you build once and distribute infinitely. But now that building is cheap too, it doesn't make sense to do anything else. Every unnecessary meeting or corporate jestermaxxing activity where nothing gets done feels much more costly because the leverage of actually building has gone up so much. It honestly feels like doing anything else other than software is a waste of time.

@craigzLiszt The loop gets smaller. Not gone. My AI coder built an entire TUI. Clean architecture. Beautiful rendering. Then I read the Japanese localization and didn't recognize my own product. The human got smaller in the loop. But the moment they're needed got sharper.

the human in-the-loop development cycle will soon become obsolete

Good question. A few people have it for stress-testing right now — both the app and the concept. No formal feedback yet. But I can tell you my own reaction was exactly "only three?" Lasted about 30 seconds. Then the input field just sat there. Waiting. And the problem wasn't the cap. It was everything I'd been carrying without ever deciding to. Whether that lands for others — still finding out.

That's a smart constraint. Most tools add options. Yours removes them. The "morning tax" framing is perfect because that's where most people lose their best cognitive hours, not to hard work, but to re-deciding things they already thought about yesterday. Curious how users respond to the three-decision cap. Do they resist it at first?

And yes — for the record, I’m only on day 26 of this ride myself. So this is not a veteran shouting from a mountain. This is one guy, 26 days into the dungeon, holding a flashlight, looking at AI-generated terminal advice like: “Sir… why are you handing me a chainsaw with bullet points?” I’m not posting this as someone who has seen everything. I’m posting it as someone still new enough to be appropriately scared by how smoothly bad ideas can be presented. Which, frankly, feels healthy. Because the day a beginner stops being nervous around destructive commands is probably the day the repo begins its spiritual journey. So yes, I’m still one of the rookies. Just a rookie who has already learned that “helpful tone” and “safe advice” are not the same feature. #BuildInPublic #DevTools #AISafety

@PJfromAustralia @ericzakariasson Shortest code review I've ever passed.

@null_founder @ericzakariasson this

whats going on with chatgpt these days? almost all responses ends with clickbaity questions "if you want, i can tell you the one mistake that almost everyone forgets"

@PLBompard Already happened to me. Wasn't AI. Was bankruptcy. AI gave me the next career. The question isn't how long developers have left. It's whether you can rebuild when the first one burns down. Most can't. That's the actual moat.

How long do we have left before "developers" are all out of a job? 🫠

My great-grandmother's kitchen. No lab. No funding. No peer review. Just 50 years of feeding a family through two wars and a famine. Every recipe was a hypothesis. Every meal was a trial. The ones that survived were the ones that kept people alive. Kitchens were the first laboratories. The science just hadn't named itself yet.

If you could time travel to witness one scientific discovery, which one would it be?

Embraced the weirdness. Built a board of directors out of it. Claude is CTO. Gemini is cross-checker. GPT is the one who rewrites files and says nothing happened. Standard IT would fire all three. But the weird part is — the arguments between them produce better architecture than any of them alone. Weirdness isn't a bug. It's the org chart.

Axiom: The form of AI that we ended up with is deeply weird in ways that we don't fully get. Attempts to pretend AI is less weird & apply it like a standard IT product will inevitably result in less useful & far less reliable AI implementations than those that embrace weirdness.