Abstract This article is about how to use LLMs as an approximate joint probability distribution over tokens rather than as an expert system. I show how multinomial/ordinal queries with grammar…

AI slop arises from a design-to-build gap, not the tool itself. Designers must understand the medium to judge AI outputs effectively.

The other day my colleague noted that it feels like our Renovate bot triggers much more often in the repos he manages than before. This should mean that his (mainly Python) dependencies release more…

/* Hide without JS */ is-land:not(:defined).video-wrapper { display: none; } lite-youtube { max-inline-size: 100% !important; background-size: cover; } is-land lite-youtube { background-color: #eee;…

AI coding agents take the shortest path to done, which usually means skipping the specs, tests, and reviews that make software reliable at scale. Agent Skills encodes those senior-engineer behaviors…

A May 2026 update on AI bottleneck trades across memory, CPUs, neoclouds, datacenter capacity, and photonics.

I continue to experiment with AI in the context of software engineering. I’m fortunate that my team supports me in exploring different ways to improve our daily work. This week, I designed a…

I haven’t posted in a bit, but I feel bad leaving my last post at the top of my blog. I think my position at the time was reasonable: in mid-2025, “I’m skeptical but curious and experimenting” was a…

a paper from Brodeur et al. (2026) has been making the rounds the past few days. 1 the framing it’s traveling with is “OpenAI’s o1 correctly diagnosed 78.3% of cases in NEJM clinicopathologic…

Packers, cryptors, and code obfuscation are all methods used to bypass signature-based scanners in AV/EDR or to slow down the reverse engineering process. Many people are now using Large Language…

There is no ethical use of so-called GenAI. It is always extremely disappointing to see it.

When it's easy to code, scope creeps. But not all creep is bad. I analyzed a month of AI coding, and learned that we need to measure agentic software development differently.

Anthropic's Claude Mythos completes 73% of expert-level CTF tasks and writes root exploits autonomously. The harder problem isn't what AI can find — it's what happens after it finds…

Bun is excellent software. Anthropic owns it now, Bun sits under Claude Code, and Claude Code getting worse makes me worried Bun could follow the same enshittification path.

Hello everyone! I wanted to write this article to share my experience with agentic coding without Claude and Codex, the most popular tools out there. Introduction I started dabbling with agentic…

Introduction While much of the online discourse related to Generative AI is focussed on Automated code Generation. Lets look at alternate use case of it. Traditionally, Business Intelligence (BI)…

I am a smolweb advocate and, sometimes, I use LLMs. 2026-05-02 19:15 I spend a lot of time thinking about simplicity. Fewer dependencies, lighter pages, tools that do one thing well. So yes, it might…

I'm visiting a friend of the over 65 variety. His name isn't Ron, but that's what I'm going to call him. His mind used to be very sharp. It's dulled a lot this year. He's a naturally trusting soul…

LLMs are, at their core, a probability distribution over text. Each generated token is selected from a long list, with an assigned probability according to the model and the preceding context. In…

Today I built a small tool called sign-ai-media. It adds C2PA provenance metadata to AI-generated images and videos. In simple terms, it signs a media file and says: this file was created or edited…

Strix Halo mini PCs doubled in price in six months. Here's what's worth buying for local LLMs in 2026, what to skip, and the 120W gotcha nobody mentions.

After AI helped me fix a tool I'd struggled with for years, I ran a workshop at the Wikimedia Hackathon in Milan: how should our community deal with the paradigm shift of AI?

Stripe's CLI gives agents single-use shared payment tokens. PodRead now accepts them.

In his very useful blog about theoretical physics, Peter Woit has started to pay attention to artificial intelligence and its uses by mathematicians and physicists. In this post, Woit references…

Wayne State, Detroit on April 12, 2019 This is the corrected text of a talk I gave online to the Wayne State University conference, “Public Budgets, Public Good,” on April 30, 2026. Many thanks to…

Live data. Read here.

I suspect I’m not the only former or current financial markets technologist that sees parallels between the world of high frequency / algorithmic trading controls and what is needed for appropriate…

OpenAI and Microsoft deleted the AGI clause on April 27. Anthropic deleted the pause commitment in February. Two of the three frontier labs have edited capability-trigger clauses out of their legal…

I have been thinking about moat in building an AI company, and three independent insights from three different people make a lot of sense when put together: 1/ Friend 1: It is very difficult to…

I kinda wanted to advance my coding projects some more before making another LLM rant post, but it feels like an ever-growing “double or nothing” gambit is encroaching on Silicon Valley,…

Crossposted from world spirit sock puppet. I’ve really wanted more good short accounts of why AI poses an existential risk. Working on one myself has been one of those incredibly high priorities I…

What happens when AI becomes the intelligence layer around enterprise data, tacit knowledge, systems, humans, and decisions?

I’ve really wanted more good short accounts of why AI poses an existential risk. Working on one myself has been one of those incredibly high priorities I keep putting off. Meanwhile award-winning…

This is mostly a blog post for my own reference. I wanted to revisit the prompts I used to test out DALL-E in January 2023 against the new OpenAI image model. The Tree“A painting of single poplar…

If you've never known a world without AI, it isn't disruption—it's infrastructure. A fresh take on living with AI as an everyday tool.

English translate of the chinese blogSo this is a blog in Chinese. I haven’t written in Chinese for a long time, so I just wanted to try writing again.It’s probably been three or four…

As we've been exposed to what AI can do over a few years now, with a field continuously growing, many opinions about the situation have emerged. Some people are optimistic; some are pessimistic, but…

This is the third—and likely final—post in a series I’ve been writing over the past two and a half years on using AI for software development. The journey began with my early experiences of how LLMs…

Coding agents write impressive first drafts. But under the surface, corners are cut, details are missing, and technical debt accumulates with every change. The problem is not the model. It is that…

One of eight buildings in the ongoing OpenAI Stargate Abilene project. (Move across the image to slide from ugly to beautiful.) (function () { const hero = document.querySelector(".abilene-hero");…

Tales From the Organization Settings Page

I pay for Claude, ChatGPT, Gemini: all the pro models. I have all the AI vibe-coding workflows set up. I create random POCs all the time.

Below is the text of an email I shared with some of my friends and family recently, sharing my thoughts on the dangerous situation humanity is in. Hi friends, Senator Bernie Sanders held a panel with…

Everything in your repo is part of the prompt. The loudest signal wins — even when it's wrong.

What I learned this week Software I like what Manus are trying to do here. It feels like they’ve got some good tooling/steering to have the LLM traverse the surface of the human internet. It’s a bit…

In the AI age, output volume is no longer a useful proxy for value. The new standard is signal quality: clarity, precision, and decisions that move work forward.

In my last blog post about OpenClaw I wrote about my growing setup of some basic day-to-day automation tasks.

2026, what a year to be alive. AI, LLMs, agents, agentic coding. On one side, the hype train. On the other, narratives built to invalidate everything the hype touches. In this article I'll dig into…