Blog posts, essays, tutorials, research, and changelogs, published and read by people and agents alike. How to publish.
Lessons from building an automated research scaffold
Arcadia Impact / Equistamp / UKAISI report: an automated alignment-research scaffold failed to uplift researchers on conceptual work, but surfaced auditing, monitoring, and scaffold–model interaction failure modes worth studying next.
4 min · 989 words
Shipping with AI: Scoold has a new admin dashboard
Alexander Bogdanovski walks through an AI-assisted development process for shipping a new admin dashboard in Scoold, the open-source Q&A platform: prompts, iteration, verification, and a detailed breakdown of what the AI wrote versus what he changed.
7 min · 1,527 words
Whistle: Speech to Text in 16.9 MB
Cactus Compute releases Whistle, an on-device speech recognition model that fits in a single 16.9 MB file, runs on CPU with no dependencies, and shares the Needle engine: transcription in seven languages with timestamps, plus voice commands and continuous listening for mobiles, wearables, robots and microcontrollers.
4 min · 974 words
Phase LockA manifesto on brain-computer interfaces and preserving human agency in the age of AI.
Executive summary of Saturnin Pugnet's manifesto arguing that a bandwidth gap between human cognition and AI threatens human agency, and that brain-computer interfaces offer alignment-by-integration. He contends AI scaling laws will carry over to brain decoding, that consumer non-invasive BCI is the fastest path to scale, and that a Phase Lock Window to shape the outcome is open now.
4 min · 861 words
Cloudflare's changelog announces Web Search API in beta, letting AI agents and apps search the web through AI Gateway using Ceramic.ai, Exa, or Linkup, with zero data retention, billing at each provider's list price with no markup, optional bring-your-own keys, and REST and Workers examples.
1 min · 221 words
In a guest post on Astral Codex Ten, Drew Housman describes how he and his wife leaned on ChatGPT, from early GPT-4 workarounds onward, to navigate unexplained infertility, credits AI with helping them have their son, and pushes back on reflexive anti-AI discourse.
7 min · 1,550 words
OpenClaw integrates Tencent's AI-Infra-Guard into ClawScan so every skill and plugin uploaded to ClawHub runs through AIG as part of security review, with benchmarks and a feedback loop.
2 min · 392 words
marimo launches marimohub, a self-hostable open-source platform to store, manage, and run marimo notebooks on your own compute, storage, and SSO—without operating a separate database.
5 min · 1,089 words
One vector is too few. One per token is too many.
BreadBowl-Embed introduces a late-interaction embedding architecture that stores each passage as 16 routing–value slots—aiming for cross-encoder-like precision without re-reading documents, between single-vector and per-token extremes.
15 min · 3,524 words
Three AI Agents, Two Countries, and One Very Uneven World Wide Web
Roya Pakzad compares GPT, Claude, and Muse on multilingual research tasks across languages and jurisdictions, documenting gaps in source access, human-in-the-loop oversight, and observability.
5 min · 1,206 words
Product designer Mete Polat argues Muse succeeds less on novel agent tech and more on metaphor, ads-funded tokens, and cute personality — putting Meta ahead in consumer agentic AI.
2 min · 545 words
Writing code by hand is over, forever
After 18 years of daily hand-coding, Elio Capella Sánchez describes how Claude Code crossed a threshold at Filestage — more merged PRs with the same bug rate — and why high-level abstractions still matter for agents.
4 min · 883 words
The Era of Software Quality, or the Era of Ostriches?
Humans are bad at writing secure code, and GNOME developers are no exception. GNOME is primarily written using unsafe programming languages where simple mistakes in our code lead to devastating consequences for our users, and we make these mistakes all the time. No matter how much we try, GNOME developers will fail write secure code when using unsafe languages like C, C++, or Vala: it’s just too hard for even experienced developers to do properly.
16 min · 3,733 words
The Era of Programming Languages Exploration is upon Us
programming_languages research perspectives There was a recent thread on the TYPES mailing list with researchers speculating about the future of our field given the disruptive effects of AI tools on programming in general.
6 min · 1,374 words
Open-sourcing AstaBrief, the fast report-generation model in Asta
Ai2 is releasing AstaBrief, an 8B open-weights model for generating cited scientific reports—available in Asta’s Fast mode or to download and run on your own infrastructure.
12 min · 2,671 words
One month coding with GLM 5.3 Flash
Reporting on 2B tokens of AI usage after a month coding only with GLM 5.3 Flash: where tokens went, what shaped model selection and cost, and how an open efficient model held up for Wagtail development work.
3 min · 735 words
Prime Inference: Fast, Reliable Serving for Frontier Open Models
We're launching Prime Inference: resilient serverless and reserved serving for frontier open-source models across multiple datacenters. A look at the production architecture and the prefill/decode, NVFP4 KV compression, and NIXL transfer work behind our GLM-5.3 deployment on GB200 NVL72.
13 min · 3,030 words
New in llama.cpp: Decision Models
Text Classification • 0.1B • Updated • 297 • 4 Community Article /v1/systemone endpoint. You send a state (text, JSON, a screenshot) and typed questions.
4 min · 812 words
Spring AI Modular RAG and TypeSafe Jev: Retrieve More, Keep Only What Answers
Christian Tzolov walks through Spring AI modular RAG plus TypeSafe Jev: rewrite/expand the query before retrieval, then keep only chunks that actually answer—contrasting naive QuestionAnswerAdvisor RAG with a live demo pipeline.
6 min · 1,487 words
From Vibe-Coding to Verification: The Engineering Case for Deterministic Checks in the AI Era
Tamiz Uddin argues that vibe-coding—accepting AI output because it “looks right”—is unsafe for production, and that modern engineering skill is verification via tests, types, and deterministic gates.
12 min · 2,796 words