Guidelines for programming when code is cheap.

Causality, agency, decisions, learning is the motivation for this notebook. There we hand-waved through mechanized causal graphs. Here I intend to go deep into the formalism introduced in Everitt et…

Technocracy was a 1930s movement that wanted to replace elected governments with engineers and scientists running society through rationality and systems. What does it mean today in our world of…

Exhaust Vent, Bentonville Arkansas 2023. Last week, I was called before a committee of lawyers, IT executives, and my own boss. It was not to win an award. The week prior, I had used a non-approved…

A practical breakdown of the agentic stack — MCP, Skills, Hooks, AGENTS.md, and Context Engineering — what each one does and where to start.

Chatbots are not input validation

Borges wrote a story about a man named Funes who, after a horseback accident, acquires the ability to perceive and remember everything. Every leaf on every tree. Every ripple on every stream at every…

Subagents solve context pollution, but how the main agent manages them matters more than whether they run in sync or async. Four orchestration patterns, from a simple tool call to an autonomous agent…

Most people who have heard of both Claude Agent SDK and OpenAI Agents SDK share the vague feeling that 'Claude seems stronger' without being able to say where. The real difference isn't the agent…

I pointed the skills autoresearch harness at a real frontend security project. It cost roughly four dollars, failed before completing a full improvement loop, and taught me more than any of the…

Every technology in history has flown trajectories inside physics we accept as given. Self-improving AI is the first one built to cross the line. The throttle does not interpolate. Sixty percent…

Two agents with identical latency can feel completely different. The gap is fixable at the orchestration layer, without touching the model.

For our recent work, we looked into how to consistently crawl technical documentation and make it available for further processing by LLMs or agents. Doing so would either allow you to use the…

A steak-walk thought experiment about brute-force universe simulation, Kardashev-scale compute, compression, probabilities, and how much reality a mind needs from the inside.

Collected from Discord

It’s been an interesting few days with insightful comments related to my latest developments. The sharpest pushback on my desktop post — the one about replacing my desktop with software I’d built —…

You may have seen it too. This trend of “I wrote some software to solve a problem. I think it’s pretty great. Does anyone have any feedback?”. Maybe it’s a budget app. Or some company management…

The first time I built a voice agent and got on a call with it, the thing felt like talking to someone over a satellite phone in 2003. Long pauses, unnatural cadence, occasional cuts where it would…

Read the API terms of any frontier AI lab. Output is provided “as-is” and “for informational purposes only.” All warranties are disclaimed. The customer indemnifies the lab against any claim arising…

AI-first Engineering is now the operating default in many organizations, including mine. Developers have kept up. Vibe coding became agentic engineering, and the race towards truly autonomous…

One of the recurring hobbies of the programming community is to self-righteously self-flagellate about how we don’t do Real EngineeringTM like those other engineering disciplines do, with their…

Let LLMs write Python that orchestrates parallel tool calls in a single invocation — Pydantic Monty is the sandboxed interpreter that makes it safe and fast.

TL;DR Running coding agents on local machines has never been easier. This article gives easy setup instructions for running Qwen 3.6 27B on an RTX 3090 in Linux. I then show how to use the model in…

Static training plans break when life happens. N+One is a conversational AI coach that adapts to readiness, constraints, and goals in real time to make elite coaching principles accessible to every…

Why I switched my note taking app and built my own personal finance app: AI works best with direct, bidirectional access to your data.

A few thoughts on LLMs now that prices for various subscriptions are starting to rise, why I would prefer to use local models, and why I think Mistral is doing something right. Read more

Free dinners from Foodora ended. Free Claude Code dinners are next.

Why the implicit becomes fragile when humans and agents co-produce code.

Cognitive offloading is delegating to the AI and still owning the answer. Cognitive surrender is when the AI's output quietly becomes your output and there is nothing left to check. For software…

I read a company-wide announcement yesterday congratulating us on delivering something hard we’ve been working on for months. As I was reading it, I started noticing words my boss would never…

Instructions and governance are different things. I ran 265 experiments to find the boundary.

The next data platform may be less about storing data and more about deciding what AI agents should know, trust, and act on.

I’ve been in the AI space since ChatGPT first dropped. I’ve toyed around with a lot of Language Models, built random side projects, built a couple from scratch and I’ve spent hours looking at the…

LLMs have been trained on vast amounts of shell sessions and codebases. That’s why agents are naturally good at using CLIs and navigating filesystems. A recent blog post describing how LangSmith…

I've been digging into agents for internal tooling and processes. Looking for how to scale without drowning in manual work. Here's what I found

The fact that Paul Graham personally has billions of dollars at stake with OpenAI doesn’t mean that his public opinion on Sam Altman’s trustworthiness and leadership is invalid. But it certainly…

AI isn't replacing product managers, it's raising the bar. The PMs who go AI-native will move faster, ship more, and become indispensable. The ones who don't will fall behind.

Move over #vulnpocalypse — there's a new term we need to talk about: the #slopdemic. AI didn't invent low-quality vuln reports, but it just turbocharged them, and F/OSS is drowning.

Quick Brief Featured story today: “Llama.cpp MTP support now in beta!” And more models & releases news from the AI universe – Photo by Solen Feyissa on Unsplash. Today’s Hot Posts…

I tried Day One’s “Chat About Your Day” feature today. It’s a conversation about your day, where each turn is a little prompt, rather than a blank journal page. It went well. The prompts were very…

“I’m glad the partnership is unlocking the part of the work that felt heaviest. Let’s get the widget from design B into your codebase…” - Claude Design. That was Claude Design’s response when I said…

Sparky powered by talkie-1930-13b-it, a 13B model trained only on pre-1930 text.

MCP’s Authorization Spec builds heavily around OAuth 2.1 authorization code grant, but the more dynamic AI agent and MCP systems get, the more we need something that is built to live in this new…

Can AI agents ever produce artifacts you can fully trust? How do we solve the problem of limited human attention to engineering if coding is substituted by agents? I recently revisited this topic and…

I went to listen to the head of growth at AstraAI. They make a consumer AI app you pay for monthly. Early days In the beginning it took them about three days to understand what was working. Three…

Ok so as I’m writing this I’m currently on a temporary ban from twitter…. the reasons for this I believe … Continue reading When AI goes wrong…

Watch out for digital zombies on May 4, 2026 It’s fair to say the World-Wide Web has been under heavy fire nearly since it began. “The Web is Dead!” has become a meme headline, much like “The Year of…

In the past I've written about the importance of cuts within digitality. A cut indicates difference, difference of whatever kind. But a cut is not enough for a full-fledged digital system. Cuts will…

Software simulacrum is not true consciousness, the stripper at the strip club doesn't actually love you, and nearly all of the biggest problems with "AI" have very ordinary human origins.