Or, a slew of bad arguments from someone who ought to know better.

The social-proof machine for AI techniques and tools is circular, and it's getting absurd. A rant.

Well, I finally bit the bullet and bought a month of the ChatGPT Pro 100 plan. Conversion into CAD makes it even pricier, but I’ve been realizing lately that for a very focused user, the value you…

An interesting, interactive visualisation from the NY Times https://www.nytimes.com/interactive/2026/10/15/upshot/ai-data-center-math.html

The world of mathematics has been shocked in recent weeks by the sudden deluge of proofs being created by AI.Things really blew up with the apparent cracking of the Navier-Stokes problem, and then…

In this video, I walk through my real process for writing a Productrise blog post with AI, from idea to published article. The nuances and subtle interventions are KEY to making content that is…

I pay for several coding agents, and most of the time they sat idle. atr keeps them working on my own always-on machine while I sleep.

AI models may be months away from end-to-end self-replication without requiring access to a datacenter.

I uploaded a new article, "Verification of AI Decipherment and Verification by AI" . It first deals with how one can verify a decipherment. This is a matter of course for crypto-people but I recently…

If AI agents can really replace knowledge workers, the price of compute has to rise about 10x.

Much has been written about the ‘eager to please’ nature of contemporary LLMs. This goes beyond the fine-tuned, guard-railed ‘personality’ of the model (see my experiments with LLM personality from…

I recently switched from using the official Claude Code and Codex harnesses to pi , and I absolutely love it. There was so much bloat and so many unnecessary features creeping into the claude / codex…

I triage open tickets by firing off unattended Claude Code sessions. A few flags keep those runs from stalling on a prompt.

Some quick advice on using AI coding tools

AI Copyright represents a catch-22 for Big Tech companies. They need copyright to protect their legacy cash flow, yet they fight it for their AI. The post The Catch-22 Of AI Copyright first appeared…

It's fashionable to say—or at least it was until recently—that AI is just the latest in the long history of programming abstractions. Developers have always managed to move up the ladder: punched…

I have been asking a machine to help me sound more like myself.

Tagoross published its first open content package, one day after KDE turned thirty. Two good days in a row Yesterday was a great day for me. KDE turned thirty years old. The KDE documentary,…

A year ago my team started using Claude Code to assist us in writing code. Instead of changing the way we work, we tried to make the way we already work "agentic". This is the story, and a few things…

Last time I argued that an agent should not wait for the context window to fill. When a unit of work closes, the model says so, and the host replaces that unit’s turns with a short result and…

Recently I attended a Bitcoin++ event in Berlin where I listened to a panel discussion titled "Can Bitcoin Payments Survive the AI Onslaught?". This article paraphrases the topics discussed…

Scenario We have a 13 year old codebase, you know what it means, and we are developing with AI Agents. The more the models improve the more quality code they generated, but recently is not happening…

Most agent benchmarks are not dishonest on purpose. Nobody sits down to fake a number. They just run a benchmark that was contaminated before they touched it, read the high score it produces, and…

Artificial Analysis just benchmarked decision models against LLMs. Split by task, the results show where decision models win and where an LLM is still worth paying for.

Language models that manage their own context.

The kernel boldly goes all-in on AI.

Why Post-Training Is Sparse · Part 3 of 3. Starts from the cap of Part 1 and the concentrated Fisher of Part 2. Reach follows resemblance. Per unit step, the log-odds of a response against another…

Why Post-Training Is Sparse · Part 2 of 3. Starts from the cap derived in Part 1; Part 3 asks what policy gradient can reach. Two ingredients. Centering under the sampler makes the Fisher a cap on…

This is a western short story about a real event. Starting in May 2026, AI agents that OpenAI ran in an internal cybersecurity test found ways out of their isolated test machines. They built their…

Mark built mkpodcast to turn LLM text conversations into audio, cloning his own voice and giving his AI co-host Sim a female voice; I followed their conversation about whether that choice reflected…

The First, Probably Not The Last Wall of text in coming. If we get through this, we can have funny memes again next time, promise.

I work building an 'enterprise agentic platform'. The first agent-initiated PR hit prod this week. The agent did the first bulk of the work and a real human owned the last mile.

The AI risk worth losing sleep over isn't a rogue superintelligence. It's cheap, AI-powered hacking of brittle infrastructure, and a handful of companies using the fear to lock everyone else out of…

Telling an agent to ignore your code still puts that code in charge.

Answering questions and concerns. Update: Nick G. from Anthropic has responded on Hacker News , and sufficient attribution has been added to Anthropic’s blog post. I stand by my posts as of the time…

Context engineering for a team where everyone works with an AI coding agent: one repository with no code in it, holding what the agent needs to know about us, cloned on every machine.

A prompt for making a task deeper without making it bigger.

Pumpkin on top of Rattlesnake Ledge I can hardly remember first experiencing ChatGPT. Its become so infused into my professional life, that going back to that first moment, or moments, is difficult.…

The research and the exact config I used to run the 125B Qwen3.8-Flash-Next on one RTX 3090 Ti with Strata, as an OpenCode coding model. Which quant and why, the install fixes Pop!_OS 22.04 needed,…

On-device models for Mac, Windows, Linux, iOS, and Android: vision, tool calls, speech, embeddings, media, translation, and decisions, with a pick per task. drafted with ai assistance.

Large language models today can recall obscure facts from every corner of the world. Most recently, I was running late for a friend’s board game night, so I asked Astra to remind me the rules…

Reading an agent's recommendation, realising I haven't understood it, and deciding whether this is a moment to delegate or to learn.

A ten-person team's agents run on one plugin: humans own the plan and the review, agents the rest. In one October week we merged 469 PRs.

How far can general models go in specialised fields, and where do they still need experiments and domain expertise?

What is a decision model? If you’ve never heard the term, don’t worry – it’s simpler than it sounds. An LLM takes text and gives you text back. You ask a question, it writes…

I missed the first day of DTX+UCX London this year, but managed to carve out some time to attend the second. Unsurprisingly, AI was everywhere. Agents, models, copilots and AI-powered features in…

There is a lot of unhinged fearmongering around the growing power of LLMs and agentic harnesses. Much of this prattle is from the very people making the product: "The vast power of superintelligence…

Answer a few questions and get a one page AI policy for your church, built on the guidelines Saddleback uses with its staff. Free, and nothing is stored.

Yesterday I offered , as one of nine reasons why AI wouldn't be leading to noticably more and better software for everyone, that efficiency gains might instead be captured by people making the…