I Still Built It — Ewan's Corner

I Still Built It - Ewan’s Blog

if an LLM wrote a significant amount of the implementation, apparently the human directing it no longer counts.

David Hendrickson on Twitter / X

🌞This is big Local AI news! A new open-source Computer-Use LLM has just launched. Holo 3.1 is H Company’s (🇫🇷) new local computer-use agent model that beats Qwen3.5-397B, Kimi-K2.5, and Sonnet 4.6!Since it is built for local deployment → ⬩ Runs fully on your machine… https://t.co/CpOEsuWN2k pic.twitter.com/w39iOh7cO1— David Hendrickson (@TeksEdge) June 2, 2026

AI Built It in 1 Hour. Lighthouse Gave Me 43.

I let Claude generate our entire design system. Then I looked at the cost.

One year as an AI Engineer: The 5 biggest misconceptions about LLM reliability I've encountered

535 votes, 59 comments. After spending a year building evaluation frameworks and debugging production LLM systems, I've noticed the same…

Lisp Against the (LL)Machine

Every Lisp, Scheme, and Lisp-adjacent project listed is a non-toy implementation that is at least somewhat active. When I first wrote this article (2026-03-12), most Lisps did not have a policy or even public stance on LLM contributions. In such cases I posted to their bug trackers or discussion lists to ask. Most maintainers kindly responded.

I rebuilt Karpathy's LLM Wiki gist: what's missing

Karpathy's original LLM Wiki gist, explained and rebuilt as a working system. 44 open-source commands, plus the five things the gist leaves out.

I accidentally turned LLM memory into program analysis :: pwning.systems

Why I stopped trying to give LLM agents a better memory and instead built Lemmalog, a Datalog engine that maintains an agent's knowledge as analysis state, with provenance, retractions and incremental evaluation, plus what happened when I benchmarked it on LongMemEval and LoCoMo.

Solving a Million-Step LLM Task with Zero Errors

LLMs have achieved remarkable breakthroughs in reasoning, insights, and tool use, but chaining these abilities into extended processes at the scale of those routinely executed by humans,...

"Useful" is not sufficient

So Linus Torvalds, head of the Linux kernel development, put his foot down on the Linux Kernel development mailing list when someone was bringing up criticism of LLMs: “Linux is not one of those anti-AI projects, and if somebody has issueswith that, they can do the open-source thing and fork it. Or just walk away. […]

The Bitter Lesson of LLM Extensions

From ChatGPT Plugins to Agent Skills, a look at how we've been trying (and failing) to extend LLMs for the last three years.

Proposal Assembly 2026: ToU extension to prohibit LLM-extrusions

Here is the proposed addition for the ToU that @Gusted and I agreed on post-assembly for the upcoming membership vote. 🙏 The discussion for this proposal is available for Codeberg e. V. members in the forum: https://forum.codeberg.org/d/139-resolutions-for-assembly-on-codeberg-taking-a-stand-a...

How LLMs Actually Work

A from-the-ground-up walkthrough of how modern LLMs work, from tokens to transformer blocks to the next-token loop

Mia (@MiaAI_lab) on X

Building with AI & LLMs | Insights, recipes, tools & honest experiments

crawshaw - 2026-02-08

I wrote up my experiences programming with LLMs a bit over a year ago, and updated it for the world of agents eight months ago. A lot has changed since then, so here is an update.