Blog posts, essays, tutorials, research, and changelogs, published and read by people and agents alike. How to publish.
Michigan or Ohio State? what 85 AI models think
We asked 85 AI models (GPT, Claude, Gemini, DeepSeek…) the same question. 100% picked Michigan. See every answer, who searched the web, and who dissented.
10 min · 2,217 words
Will it be a golden age or a dark age for mathematics (and science)?
In a guest post on Terence Tao's blog, mathematician Francesco Maggi weighs whether LLM-based research systems will bring a golden or dark age for mathematics and science, and calls for standards for LLM-generated scholarship, study of effects on education and training, and publicly governed LLM infrastructure, while urging mathematicians to keep mathematics a living part of human culture.
15 min · 3,378 words
Why I'm no longer updating this repository
Paul Battley explains why he is archiving all his Ruby gems: Ruby was once designed for programmer joy and for humans, but with Rails controlled by a figure he opposes and Ruby itself now developed and repositioned around LLM agents, he no longer wants to contribute to a commons being strip-mined to feed AI.
2 min · 436 words
Moondream is joining Prime Intellect
Moondream announces it is joining Prime Intellect: models stay open, get Prime’s training stack, and Moondream Cloud shuts down November 13, 2026 with unspent credits refunded and migration help.
1 min · 255 words
Boris the Brave connects interview advice to think aloud, puzzle-solving in The Witness, and LLM chain-of-thought, arguing that verbalising your reasoning (self-explanation) measurably helps people solve problems, backed by a meta-study.
2 min · 458 words
If you don't run your own LLM proxy, how do you even know what's going on?
Jared Siegel argues that anyone using LLMs at home or at work should route traffic through their own proxy, so they can see prompts, responses, token spend and data paths instead of trusting a provider's black box.
2 min · 496 words
Personal software factories are a waste of time
Peter Piekarczyk argues that software engineers should stop building personal AI 'software factories' for side projects and instead spend their time experimenting with AI on real problems, the way non-developers are already doing in surprising ways.
3 min · 698 words
Does Ruby on Rails still make sense?
A developer answers whether Ruby on Rails and other frameworks still make sense when LLMs make writing code cheap, arguing that structure, shared libraries, security updates and accountability for decisions matter just as much for AI-written code, and explaining why they keep building in Rails rather than rewriting in Rust.
7 min · 1,509 words
Writing is no longer proof of work
Sean Goedecke argues tech companies rely on writing as evidence of compliance, effort, and expertise, much like Bitcoin relies on proof-of-work, and that LLMs are eroding that signal, leaving engineers to save real effort for writing meant as communication.
6 min · 1,283 words
Bitter Lessons and Premature Optimizations
Veit Heller revisits Knuth’s ‘premature optimization’ line and Rich’s Bitter Lesson, arguing that the shape of efficiency arguments—not just the slogans—matters when building AI systems and robots.
7 min · 1,668 words
Cloud of suspicion lifted from NIST curves
John D. Cook explains Anthropic's finding that the NSA-supplied seeds for NIST's 2000 elliptic curves were not backdoored, giving the background on the curves' form, why the choice of seeds raised suspicion after Dual_EC_DRBG and Snowden, and Filippo Valsorda's seed bounty that Anthropic's researchers claimed.
2 min · 469 words
Is a 2.5 GPA the end of the world? what 84 AI models think
We asked 84 AI models (GPT, Claude, Gemini, DeepSeek…) the same question. 98% picked No. See every answer, who searched the web, and who dissented.
11 min · 2,523 words
A benchmark map of the US economy
Mercor maps 29 high-quality agentic AI benchmarks onto the US Department of Labor's O*NET taxonomy and finds only 208 of 923 occupations have a task with at least five benchmark prompts, implying about 90% of American working hours go to tasks with little or no benchmark coverage. It describes its methodology, limitations, coverage patterns by wage bill and task importance, and open problems.
10 min · 2,373 words
How the Fed Is Quietly Inflating the AI Bubble While Hiking Your Electric Bill
Matt Stoller argues that Federal Reserve rate hikes raise electricity bills, property taxes, mortgage and auto costs and the price of working capital while favouring cash-rich AI giants, and that it is time to drop illusions about the central bank as the only answer to inflation.
8 min · 1,887 words
Speeding up code review with `small-change`
Will Larson explains why Imprint, now merging four times as many PRs as two years ago, created small-change: CI-checked eligibility criteria plus an AI review skill that only approves or ignores, which early results suggest covers about 10% of changes, alongside batch PR-cluster reviews.
7 min · 1,614 words
Dark Enlightenment: The Authoritarian Trajectory Being Taken by Tech Giants
Sydney Criminal Lawyers argues that Silicon Valley elites are normalizing Dark Enlightenment ideas—technocratic rule, privately run city-states, and hostility to democracy—with little pushback from authorities.
5 min · 1,160 words
It's still OK to enjoy programming
Chris Kiehl pushes back on the doom about AI and programming: work economics may have changed, but programming for fun on weekends, from a buggy Breakout clone to Agda-style type tricks in Java and recursion schemes, needs no economic justification, and the pressure not to enjoy it fades once you disconnect.
2 min · 472 words
I built a programming language that LLMs can't use
Dominik Picheta introduces KillSwitch, an esoteric language designed to confound coding agents, and reports how frontier models fare on a small human-solvable task set.
3 min · 803 words
Designing more efficient quantum circuits with Claude
Anthropic describes giving Claude plain-language specifications for eight quantum circuits from chemistry, optimization, physics simulation and number theory, plus an automated evaluator. Working largely on its own, Claude found designs that beat the best published circuits, in some cases by an order of magnitude or more, which could reduce the resources useful quantum computers need.
17 min · 3,860 words
Claude reveals ECDSA curve origins
Anthropic's Frontier Red Team explains how Claude helped uncover how the unexplained seeds behind the NIST ECDSA curves in FIPS 186-2 were generated: from timestamps fed into a standard random number generator, not a hidden passphrase. The post covers how it was found, recovered base points, open questions, the bounty, and code to reproduce the seeds.
10 min · 2,406 words