Developer Advocate @monad | merc @buidlguidl |馃憖 address.vision | abi.ninja
I get this question a lot, and I ask this to myself almost every 3 weeks. I ask it so often that I built shouldirunthis.xyz to settle the money side of it (buy a rig, or keep your sub?). The
I put together a lightning talk for ClawCon Ankara around one question I keep getting asked: should I run AI locally? People want a yes or a no. The honest answer is that it depends on the job, and I
A few days ago I published the Gemma 4 12B test, where a free local model wrote a dapp and found zero of its own bugs. The obvious follow-up was to run the same test on a frontier model and keep the
A free model that fits on a laptop wrote my entire dapp, contract and frontend, and then couldn't find a single one of its own bugs. I had a question in mind: can a free, open model you run on your
Any problem yields to enough complexity. Most of puddleswap.org was built by an AI agent, and that is most of why I want to talk about it. An agent reaches for the textbook answer by default,
The hardest part of building anything is picking the right thing to build. There is exactly one heuristic I trust: If I have a problem with the tools I use every day, other people probably have it
I always wanted a local coding agent that could work inside a real repo without costing me 100 bucks a month. Here is what I landed at after a lot of trial and error: Unsloth Studio as the local
Source: OpenAI. Prompting GPT 5.5 is A LOT different than how you prompted any model before. And GPT 5.5 itself can't write good prompts for itself! See the screenshot below from @VictorTaelin So,
In my previous article, I argued that skills don鈥檛 work the way most people expect. The data from SkillBench supports this. Attaching skills doesn鈥檛 automatically guarantee better performance. So
I just finished reading SkillBench paper : https://arxiv.org/pdf/2602.12670 And the results are definitely not what most people expect. What researchers did They did 86 real-work tasks across 11 domains
It's 2026. We have AGI (or at least the ability to code almost anything thanks to models like Opus 4.5 from Anthropic and GPT 5.2 from OpenAI). But there's one problem. What you create in minutes
Thinking about games to build, one common pattern we can observe is that there usually is a game and a separate mini-game that uses some kind of randomness built around it. Most mobile games work like
