Kyle

Yesterday I wrote 5 numbered prompts for Cursor to start amiqo's v4 build. The prompts aren't what makes this work. The closed-loop underneath does. Every product decision lands in a log with a measurable hypothesis. Two weeks later, a sync job grades it against real metrics from a JSONL pulse file. Two of my recent decisions abandoned themselves before code shipped, because the system saw the data was noise and stopped me from building against it. A one-shot prompt does what you ask. A prompt living inside a closed-loop gets graded, and so does the next one. That's the part that compounds. Built on @claudeai . The shift @karpathy keeps writing about is mostly about the loop, not the model.

@zach_yadegari but what about the 1 person 1 billion dollar company?

Being the best CEO is really just about being able to get the smartest brains in a room together and then knowing when to use each one to make a decision.

@connorpaton lol, the tired, confused, and endless iterations is so real

you can tell everything about a founder by how they answer "what are you working on" at a coffee shop some light up. some rehearse a pitch. some look tired and say "man, i don't even know anymore" the third group is usually building something real

@nivi yup, just added this to my product backlog. desperately need it to reduce friction.

Tapping, typing, and orchestrating your user experience has to go.

@yashhq_22 what can you do about it?

as a solo founder in 2026 your biggest competition isn’t another startup. it’s the founder who’s been posting on X for 6 months longer than you.

@DanielSmidstrup nothing says real people like a real image

Hey you🫵 If you're building anything with AI, let's connect👋

@RoyInProgress @kzitouni1 how do you get into this "build in public" community? is it a hashtag thing, consistent posts, just the type of content, or something else?

@kzitouni1 There’s no community like the build in public crowd 🙌

building on X sucks if you're posting and getting 2.5 likes whatever you do, make sure you find your community makes everything 10x easier.

@kzitouni1 suggestions for finding your community on X?

A user told me he'd downloaded twenty friendship apps before landing on amiqo. I asked what made him stay. "Every other app like this is full of predators or scammers or bots," he said. "I chose this one because it felt different." Twenty apps. The guy knew the landscape better than most people who write about it. And what moved him wasn't the design, or the AI matching, or the features. It was whether the people who built it were thinking about who was going to show up. I think about that one a lot.

Real men measure your worth by how high you build, not how far down others fall.

Working on amiqo last week, I went looking at other friendship apps I'd vaguely heard of. Cappuccino: still alive, 4.7 stars on iOS, last update late 2024. Their pitch is "a mini-podcast made up of stories from your closest friends, for elite friend groups only" is targeting a similar cohort I'm building for. Different mechanic (they're audio beans; we're daily practice + check-ins), same person at the center: someone trying to keep the friends they already have. "No one's doing this" almost always means I haven't looked carefully enough.

@Shpigford real-life build in public case study

@Kyleschaffer11 here

first paying customer of anything new...magic.

First time I opened terminal I didn't know what I was looking at. 18 months later: full Personal OS in markdown files, consumer apps shipped in Cursor with GitHub. The conceptual ramp is legitimately steep. The payoff is the complexity turns out pretty manageable once the model clicks.

For a non-technical founder, there is a lot of learning that is required to properly set up the infrastructure for AI. Once you understand it, it’s not complex nor hard but going from never working inside the terminal to learn how to set up your markdown files and cohesion across your company is a lift.

Most advice about a fading friendship says "have the conversation." That ends it faster. The move that works is one text: "I miss you. Free for a walk Saturday at 10?" That's it. About 70% of fades recover from that one message. amiqo.life/blog/signs-a-f…

Friend.com (@AviSchiffmann) raised $8.5M to build an AI you can talk to when you're lonely. We're building the opposite. Two years into amiqo I keep coming back to one fact: the apps that ask nothing of you are the ones that don't work. We send you one tiny friendship action a day. Text a person you haven't talked to in three weeks. Plan something with someone you've been meaning to see. That's it. You do the rest. Call it friction-maxxing if you want. The AI is supposed to make friendship easier. We think the friction is the whole point.

@BatsouElef Yes but always feeling behind… scrolling through god-tier launches while my biggest win this week was finally closing 47 browser tabs.

X is cool. But it’s 100x better when your timeline is full with people who code and build things. If you’re into tech, AI, and startups say hi 👋

@Bhavani_00007 Keep building

In 2026, • Every idea feels taken • Every API already exists • Every SaaS has 12 competitors • Every startup is already on Product Hunt So what the hell do we build now 😭

@garrytan Same Garry. Never felt like I could do more in my life… while simultaneously feeling 3 years and 47 failed retrieval loops behind everyone else. Agents are the ultimate ‘touch grass’ reminder except the grass is on fire. Random question but how much sleep are you getting?

I'm having the time of my life frankly

@mal_shaik What’s your favorite story of someone using what you’ve built?

ppl really assume that the main reason ppl start companies is to make money but if thats your only motivation its sooo easy to quit money is just the tool to continue doing what i actually love: seeing ppl use what we built

@itsjessyin Finally was able to build products myself, for myself!

i’m curious for people who don’t have an AI/ML background what got you into using AI? what was the first thing you did with it and what do you wish you knew then that could have accelerated your learning or use of it?