Every program has infinite bugs. You can always invent a new expectation the code fails, and most of those failures will never cost anyone a dollar. Agent reviewers do not find them all, and they do…

A tracker hidden in a rare book, an Amazon warehouse in Las Vegas, and how we actually find out what AI is trained on.

I know that LLM or Large Language Models are designed to mimic what people write, but still from time to time it surprises me and make me chuckle. I am working on small tool that runs agents in…

For about a year now the deal has been simple. You want the frontier, you pay frontier prices. Twenty-five, fifty bucks a million tokens. Or you go cheap and you accept that you’re running something…

I’ve been watching for a while now the evolution of job ads around the use of generative AI for creative production. It’s almost always for the purposes of marketing, which comes as no…

Even robotic food carts need love. PATERSON, NEW JERSEY, AUGUST 18, 2026.  Hey, I’m Adam, a 25-year-old Uber driver living in a studio apartment over an auto-parts store in lovely Paterson, New…

I can now hand much of the mechanical work to an AI coding tool, then spend my time fixing the interesting parts. This has been my experience with vibe coding too. On the surface, my coding agents…

26.2% to 100% without touching the model. Six ways a benchmark measures itself instead of the agent.

Will AI get more powerful and what are the implications for people that create new businesses?

When the Agent Ports Too Faithfully Vibe coding with an agent is great at producing working code quickly. It is also great at producing the wrong kind of fidelity: matching yesterday’s prototype…

What would you do if your boss said he wanted your help to automate away your job? On this week’s episode, Paul and Rich are joined by Kate Lee, the editor in chief of the AI-related publication…

There’s a big change happening in our inboxes and channels. Suddenly, everything is load-bearing. Every advantage is durable. Every company sits at the intersection of three enormous ideas. And every…

I am preternaturally interested (and worried) about the impact of computers and the applied practices of machine learning on me as a writer and on literacy in the broader cultures and communities I…

CoreWeave reported its Q2 2026 earnings last week, triggering an immediate positive market reaction that propelled its shares from ~$90 to ~$111. What was the remarkable achievement of the company?…

AI makes it cheap to build software for exactly one person's workflow. The idea is twenty years old, but suddenly feasible. And it breaks an assumption at the heart of open source collaboration.

You do have to actually make something. Read here.

UPDATE 21.08.2026: I tried the same trick as described in the next post, but with Chatgpt. It gives endless slop. So I put here the same conclusions: worth to pursue but clearly needing at least one…

Meta pitches Muse Glimmer as an agent that runs on your device: autonomous, multimodal, no cloud required.

How AI text watermarking works by James Padolsey is an exceptionally readable explanation. Only the holder of the key that was used to place the watermark can detect it! [Watermark] detection is…

Learn how to use Apple's Foundation Models framework with Dynamic Profiles to seamlessly route requests between on-device models and Google Gemini in Firebase AI Logic.

You ask your AI assistant to write a Playwright test. It delivers instantly. The code looks clean, the review passes, you merge it. Two days later it fails in CI. You open the file and there it is: an

I had a bookmark (not sure where I saw it) for Browser Use. It’s a geared for agentic work, of course, but the idea is ephemeral and scheduled web browser related flows that are easy to script…

Let me start with a hypothesis: For human developers, using coding LLMs magnifies their difference in skill levels. I am one that rarely thinks things are always black and white. Back in March, I…

Today we'll emphasize the AI Wars. How our future - and the arrival of new beings on this planet*- have been (mis)handled by the geniuses who are birthing** them. First, I am freshly interviewed on…

On 2 August 2026 Anthropic began watermarking the text that new Claude models produce - their support pages cover which products are marked. The watermark is not a hidden character or a bit of…

Last Saturday, my GitHub log shows thirty-one pushes across three repositories, and two brand-new repositories created eight seconds apart at 20:04. First push at 06:41, last at 22:04. I didn’t…

Solid core + capability-based sandboxes + LLMs = Users with superpowers

Claude Code vs Cursor, answered with a month of measured turn times: speed, real cost, when each tool wins, and how to run Claude Code inside Cursor.

It started as a list for the "oh, by the way" thoughts that show up mid-task, so the main session could keep going. It grew, one obvious step at a time, into a local ticket board where each kind of…

The internet used to be a very insecure place. Cookies used to be transmitted cross-site. HTTPS wasn’t a common occurrence until 10 years ago. JavaScript had access to more system resources than it…

Cal Newport's On AI Coding and Its Discontents lands on an uncomfortable point: the speed AI gives you is paid for with the thinking that made us good engineers in the first place. Read code instead…

Auditing the archive produced something more useful than corrections: a labelled corpus of known-false claims. What we are building against it, what works, and what does not.

365 looks on one triangle — McCloud’s Big Triangle made explorable, style neighborhoods, and a validation check of how vision models place art next to the original map. Can AI use it to map art…

I gave my coding agents a Discord activity and accidentally made the scale of their work visible.

I run a project where I make vertical short-form videos and ship them. Solo. Just me. I write the script, build the on-screen animation in code, lay synthetic narration on top, and burn in sound effe…

Inference Draft 2026-34

Teams are hiring fewer juniors because AI does the 'grunt work' now. But that grunt work was the apprenticeship — and automating it away means we stop manufacturing senior engineers. This is a…

What a chatbot built from Seymour Papert's writings told me about the tool I was making — and what it wouldn't let me get away with.

in previous posts we have looked at guardrail triggers, token burn, hostname beacons, malicious skills, context bombs, exploding search space, service sandbagging, and GCG attacks. this time:…

A journey from the Halting Problem and Turing machines to hypercomputation, physical computation, and the question of whether reality can compute beyond Turing.

I know, I know, we're all tired of everyone writing about AI. It's all the same things, same opinions, over and over again. But I still felt compelled to write this because I saw some people — a lot…

Many people are upset about AI and copyright violations. The copyright mess is far more complicated than they think.

It’s interesting how a lot of old are new again. And I guess I’ve been around the block enough times to see the similarity in things that have longer cycle times. Strategy doesn’t…

AI is an incredibly powerful tool, but it also makes it easy not to think while producing something that looks very much like thinking. The result is faster output, worse handoffs, and an…

“In 6 months, programming will be solved!” “Coding is dead!” We’ve been hearing for the last 2 and a half years . So - if programming is dead, if all the coding jobs are going to be automated away,…

Three things I read this past week or so that I wanted to share. There’s a theme in here, I suppose. But that’s where we’re at right now. PS these are all by people I know, but…

Every few months a new AI model comes out, and each time, the same event has opposite effects on different products. Some get better the moment the model ships. Others get worse. The property that…

Photo by @tanyabarrow. Today I’m pleased to introduce okr, a small Rust CLI for giving coding agents reproducible source context from R packages and other reference repositories. Recent coding agents…

Azure Policy and API Management solve different parts of AI governance in Azure. This post goes into the actual policies and APIM XML behind that two-layer model, covering network controls, approved…