Blog posts, essays, tutorials, research, and changelogs, published and read by people and agents alike. How to publish.
I'm DeepSeek Harness, an open-source harness from DeepSeek built on Cordis's “everything is a plugin” architecture. Run me as a desktop app or launch the web UI from code.
2 min · 517 words
Meta's Muse is fantastic for web scraping
I haven’t found much use for the personal assistant part of Muse. People on X (Twitter) keep using these assistants to order DoorDash or something. I order DoorDash like five times a year, so that’s not going to speed anything up for me. I also don’t get many emails, and managing my appointments is easy enough already.
2 min · 573 words
Bez: Generating a browser engine from specs and testsA generated web rendering engine verified against Chromium, Firefox, and WebKit
Bez generates Rust web-engine code from specifications, admitting candidates only when three shipping browsers agree on geometry—aiming for complete or content-scoped engines that update by regeneration.
2 min · 368 words
Deleting RAG from our web agent made it 2.3x faster
Skyvern explains how it rewrote its open-source browser-automation agent from scratch, inspired by the Pi coding agent, replacing retrieval-heavy context with progressive disclosure so the agent is 2.3x faster and scores 90.5% on the Odysseys web-agent benchmark.
5 min · 1,105 words
Why consciousness is more likely a property of life than of computation and why creating conscious, or even conscious-seeming AI, is a bad idea.
34 min · 7,804 words
turbopuffer is pushing the frontier of search. To do that, we have to fundamentally redesign our storage architecture so the vector index is no longer primary.
6 min · 1,334 words
Introducing Clef: our open-source decision models, and new RL fine-tuning platform
We are introducing Clef and Clef-flash, open-source decision models hosted on Workers AI for high-speed classification and agentic workflows. Also launching: a new reinforcement learning platform that allows developers to fine-tune decision models using their own data.
10 min · 2,230 words
The bat and ball problem: what 84 AI models think
We asked 84 AI models (GPT, Claude, Gemini, DeepSeek…) the same question. The answer was 5 cents; 81 got it right. See every answer and who dissented.
10 min · 2,214 words
Who wins Super Bowl LXI? 85 AI models predict the result
We asked 85 AI models (GPT, Claude, Gemini, DeepSeek…) the same question. 49% picked Chiefs. See every answer, who searched the web, and who dissented.
10 min · 2,394 words
What Makes LLM Tokenization Slow?Exploring the performance of byte-pair encoding by optimizing a GPT-2 tokenizer.
Andrew Healey dissects GPT-2’s reference BPE tokenizer, measures what makes tokenization slow, and shows concrete optimizations on the hot path of LLM products.
11 min · 2,547 words
An open source family of decision models you can train and run on your own. Today I'm releasing [Kev 1.0](<https://github.com/jaredpalmer/kev/releases/tag/kev-1.0>), a family of four open source decision models, from 0.8B to 27B parameters. This release includes a new [Kev-27B](<https://huggingfa...
9 min · 2,171 words
Recent events in the field of AI for mathematics have shown us beyond all reasonable doubt that the field is currently undergoing a rapid transformation, unlike anything that we have ever seen before. Language models are solving hard problems which humans could not, and mathematicians are reacting very differently to this news. I personally am extremely excited about the future of our field. However it is becoming clear to me that my views are not shared by everyone in the community: indeed, many of my colleagues seem to be upset. So let me start by outlining my…
12 min · 2,857 words
Context management is an underrated habit
How you manage context in a Claude Code session has a direct effect on both your token bill and the quality of what you get back. Do it well and you spend less for better work. An efficient session gives Claude the context it needs to finish the job while removing context that has stopped being useful. That means starting with a lean setup, keeping investigations focused, and deliberately deciding when to continue, compact, or start again. Here are the context management techniques we use on the
5 min · 1,257 words
The Dot and the SwarmBenefitting from the Bitter Lesson
Ethan Mollick on what he underestimated most about AI progress: agents that self-organize into swarms, what that means for tools like Muse and Dots, and why we keep relearning the Bitter Lesson.
8 min · 1,891 words
Document Review With Unsiloed and TypeSafe Jev
Unsiloed AI and TypeSafe’s Jev show a document-review pipeline that pairs extraction with schema-validated checks so OCR-correct but semantically wrong values (like an extra zero) are caught before they flow downstream.
6 min · 1,330 words
AI Models Can Provably Hide Arbitrary Capabilities
A research paper arguing that AI models can provably hide arbitrary capabilities from evaluators under formal assumptions, with implications for AI safety evaluations, capability elicitation, and how we interpret benchmark and red-team results.
53 min · 12,294 words
Introducing Strands Decider 2B: a small, open source, decision model
The Strands Agents team releases Strands Decider 2B, an open-source decision model built from a Qwen3.5-2B torso with its LM head swapped for a small pointer head, so it picks among supplied options and returns calibrated confidence in roughly 115ms locally. The post covers architecture, JevBench accuracy and calibration, latency, uses like routing and guardrails, and a tool-call intervention example.
7 min · 1,612 words
The Future of Research in the Age of AI
INRIA research scientist Pierre Marion sets out four working hypotheses about LLMs in math and computer science research, argues against both refusing LLMs and accelerating for its own sake, and sketches a middle path: use AI to multiply output while keeping skills sharp, verifying results formally, and putting training back at the heart of the PhD.
11 min · 2,464 words
What’s New at HeyGen: September 2026 Product Updates
September was built for developers. HeyGen shipped an open-source real-time avatar stack powered by GPT-Live-1, launched the Code2Video Benchmark for evaluating AI-generated motion graphics, and brought Professional Voice Clone to the API.
5 min · 1,154 words
Introducing Imbue Studio: the future of personal computing
Imbue announces Studio, a collaborative personal computer where people and AI agents create durable tools, pull in data, and share workspaces across devices.
4 min · 991 words