ProseMirror
ProseMirror is a toolkit for building rich-text editors on the web, created by Marijn Haverbeke and released under the MIT license.
News 📰
Short notes, quick updates, and things I find worth sharing. For the full graph of my thinking, browse my public notes. Subscribe via RSS.
Qwen Image 3.0
Qwen Image 3.0 is the third generation of Alibaba's Qwen image generation model, released 21 July 2026. The tagline: "Rich Content, Authentic Details, Deep Knowledge".
Kimi K3
Kimi K3 is a 2.8 trillion parameter open-weight LLM) from Moonshot AI, announced on 16 July 2026 with weights published on Hugging Face on 27 July. It is the successor to Kimi K2.6 and the first open-weight model to be argued, credibly and by people with no stake in it, as sitting inside the closed
Qwen 3.8
Qwen 3.8 is a 2.4 trillion parameter multimodal LLM) from Alibaba's Qwen team, announced on 19 July 2026. Alibaba's claim is that it is "one of the most powerful models available today, compatible to leading frontier AI models, second only to Fable 5".
Gemini 3.6 Flash
Gemini 3.6 Flash is Google's workhorse Gemini model, released on 21 July 2026 alongside Gemini 3.5 Flash-Lite and Gemini 3.5 Flash Cyber. Its pitch is not intelligence, it is cost per completed task: cheaper than the model it replaces, meaningfully better at coding and tool use, and 17% more token-e
Bonsai 27B
Bonsai 27B is an extremely low-bit compression of Qwen3.6-27B from PrismML, released on 14 July 2026 under Apache 2.0. The 1-bit variant is 3.9 GB. That is a 27B-class model, with reasoning and vision intact, running on an iPhone.
LM Studio Bionic
LM Studio Bionic is an agent application from the LM Studio team, announced on 16 July 2026 and positioned as "the AI agent made for open models". It is a separate app, not a mode inside LM Studio: coding, research, and document work, driven by open-weight models running locally, over LM Link, or in
Codex Micro
Codex Micro is a $230 macro pad from OpenAI's Supply Co., made with Work Louder and announced on 15 July 2026. It is a physical control surface for OpenAI Codex: thirteen mechanical keys, a rotary dial, a planar joystick, and per-key RGB that reflects live agent status.
DSLs Make LLM Output Reliable
Give a model a domain-specific language) instead of a general-purpose one, and its output gets dramatically more dependable. Unmesh Joshi (Thoughtworks) made the argument on Martin Fowler's site in July 2026, and it is one of those ideas that is obvious in hindsight and almost nobody applies deliber
GitHub Copilot Vision
Copilot Vision became generally available on 1 July 2026. You can attach images and PDFs to a GitHub Copilot chat prompt, and Copilot reasons about them alongside your code. Available on every plan including Free, enabled by default, no admin action needed.
Cursor Agent Swarms
In July 2026 Cursor published the numbers behind its agent swarm work, and it is the most useful public document on agent swarm economics so far. Not because the idea is new, but because they ran the same task through several model combinations and published what each one cost.
JetBrains Context
JetBrains Context is a repository intelligence layer for coding agents, announced in July 2026 and in early access at no extra cost with a JetBrains AI subscription. It builds a semantic index of your repositories and lets agents query it by concept instead of by keyword, so they stop burning turns
FFASR Leaderboard
The Far-Field ASR Leaderboard is an open benchmark from Treble Technologies and Hugging Face, measuring speech recognition under realistic acoustic conditions rather than clean studio audio. It is the first open far-field ASR benchmark maintained as a continuously updated leaderboard.
OpenAI Open-Sourced Its Security Scanner
Last week OpenAI pushed a repository called Codex Security to GitHub. No announcement, no blog post, no tweet. Hacker News found it anyway and sent it to 560 points before OpenAI said a word.
Kimi K3, Qwen 3.8, and a Week That Reset the Open-Weight Frontier
July 2026 was a LOT. Two open-weight models at frontier scale announced within three days of each other, a 27B model that runs on a phone, and a pile of tooling news that will affect your daily work more than either of them.
Benefits of Obsidian
Obsidian is a powerful note-taking and knowledge-management app that focuses on local, plain-text files and interconnected ideas. Unlike many other tools, it prioritizes longevity, privacy, and linking your notes together.
Kanban Action Planner 1.8.0 - GTD contexts
Last time, version 1.7.0 of the Kanban Action Planner brought WBS mode, automation rules, and embedded boards. 1.8.0 is smaller and about one idea I've wanted in there for a while: contexts.
Kanban Action Planner 1.7.0 - WBS mode, automation rules, and embedded boards
Two weeks ago, version 1.1.0 of the Kanban Action Planner shipped timeline mode. I haven't slowed down since. Versions 1.2.0 through 1.7.0 add a WBS mode, automation rules, embedded boards, a real estimates system, and a performance pass that touched almost every interaction. Here's what's new.
Riddle (reMarkable)
Riddle turns a reMarkable Paper Pro into Tom Riddle's diary. You write with the pen, pause for a few seconds, and the answer appears on the page as flowing handwriting. No screen glow, no keyboard, no chat window. Built by Maxime Rivest, MIT licensed, around 1.6k stars on GitHub.
Obsidian Keyboard Shortcuts
Obsidian has great support for keyboard shortcuts. Most commands can be invoked using shortcuts, and most of those can be customized.
Concepts Website
A curated, cross-linked collection of concepts — designed to learn from first principles. Built from concept cards and published as a website.
OpenMontage
OpenMontage is an open source (AGPLv3) agentic video production system. You describe the video you want in plain language, and an AI coding assistant (Claude Code, Cursor, Copilot, Windsurf) handles the whole chain: research, scripting, asset generation, editing, and composition. No manual editing w
Zencastr
Zencastr is an all-in-one remote recording platform for podcasters, video creators, and interviewers. Founded in 2016, endorsed by people like Seth Godin and Sam Harris. The pitch: spend less time juggling tools, more time making content.
AgentBox
AgentBox is Rui Carmo's Docker-based sandbox for running coding agents in isolation. The README is refreshingly honest about the state of the art: "There's no perfect way to sandbox agents (yet), but containers are a practical start." MIT licensed, ~141 stars, mostly Dockerfile and Makefile.
Kurt Vonnegut - Enough
True story, Word of Honor: Joseph Heller, an important and funny writer now dead, and I were at a party given by a billionaire on Shelter Island.
Kanban Action Planner 1.1.0 - timeline mode
I just shipped version 1.1.0 of the Kanban Action Planner plugin. It adds board, calendar, and now timeline views to Obsidian Bases, all driven by your own note properties.
Claude Sonnet 5
Claude Sonnet 5 is Anthropic's mid-tier Claude model, released early July 2026. Anthropic calls it "our most agentic Sonnet yet": it plans, drives browsers and terminals, and runs autonomously at a level that required Opus-class models a few months earlier. The pitch is simple: near Claude Opus 4.8
Claude Tag
Claude Tag is Anthropic's way for teams to work with Claude as a teammate inside Slack. Announced June 23, 2026 (beta for Claude Team and Enterprise), it lets Claude join Slack as a team member: admins grant it access to selected channels, tools, data, and codebases, and anyone in a channel can tag
The strongest version of you is buried under the habits you're afraid to break
> The strongest version of you is buried under the habits you're afraid to break
AI and Context Engineering Glossary
Unified terminology reference for AI, Context Engineering, and Knowledge Management. Organized by domain. Each entry links to its dedicated vault note.
Comprehension Debt
Comprehension debt is the gap between what a system does and what its owners actually understand about it. Every line of code you ship without understanding adds to it. You still own the behavior; you just can't explain it, debug it, or safely change it.
Loop Engineering Went Mainstream
Mid-2026, and "loop engineering" is now the phrase you keep seeing in AI conversations. Boris Cherny from Anthropic (creator of Claude Code) said it directly: "I don't prompt Claude anymore. My job is to create loops." Peter Steinberger made the same point in his own words: stop prompting coding age
Markr
Markr is a tiny desktop timer that lets you bookmark moments while you record, so editing means jumping to marks instead of scrubbing the whole clip. Mark the moment as it happens, not after. Electron/React/TypeScript, MIT.
Obsidian AI Skills by Steph Ango
Obsidian AI Skills (the `obsidian-skills` repository) is an open set of Agent Skills published by Steph Ango (kepano), Obsidian's CEO, that teach AI agents to read and write Obsidian's file formats natively instead of guessing at them. It went up in June 2026 and blew up fast; it was already past 37
Docker Agent
Docker Agent is Docker's open-source framework for running teams of specialized AI agents that delegate to each other, configured entirely in YAML and shipped as OCI artifacts. Built on the `cagent` engine; runs with `docker agent run config.yaml`.
Agentic Resource Discovery Server plugin for Obsidian
Obsidian community plugin that turns your vault into a local-first Agentic Resource Discovery (ARD) publisher and Agent Registry. It scans your AI Skills, builds a rich `ai-catalog.json`, and runs a localhost server so AI agents on your machine can discover and fetch exactly the skill they need by n
Kanban Action Planner plugin for Obsidian
Plan, track, and schedule any set of notes visually, straight from a Obsidian Base. This plugin adds a Kanban board view to Bases, so your tasks, projects, content ideas, or anything else you keep as notes can move across columns instead of sitting in a flat list.
ADMX Viewer
ADMX Viewer is a small web app I built to browse, search, and configure Windows Group Policy templates without fighting the built-in tooling. You point it at a folder of ADMX files, pick a language, navigate the policies, flip the settings you care about, and export them as `.reg` files, PowerShell
GLM-5.2
GLM-5.2 is Z.ai's) flagship open-weight LLM), released as full open weights on June 16, 2026 (coding subscribers got access on June 13). It is the successor to GLM-5.1 and, at release, the leading open-weight model on the Artificial Analysis Intelligence Index. The headline change is a jump from a 2
Vercel Eve
Vercel Eve is an open-source TypeScript framework for building, running, and scaling durable agents, launched in public preview on June 17, 2026. The premise: every team kept rebuilding the same plumbing (sessions, sandboxing, channels, approvals) before their agent could do anything useful. Eve pro
Odysseus (AI)
Odysseus is a self-hosted AI workspace that puts chat, autonomous agents, research, documents, email, notes, calendar, and local model serving behind one local-first interface, with no telemetry. Python + JavaScript, AGPL-3.0-or-later.
Vercel Open Agents
Vercel Open Agents is an open-source, fork-able reference app for running coding agents that live in the cloud and keep working after you close the tab. Built on Vercel's own stack. MIT.
Ponytail (AI)
Ponytail is an open-source (MIT License) AI-agent plugin/skill that enforces minimalist code generation, "He says nothing. He writes one line. It works." It counters the tendency of coding agents to over-engineer (extra dependencies, boilerplate, building when something already exists) by injecting
Agentic Resource Discovery Specification (ARD)
Agentic Resource Discovery (ARD) is an open, draft specification (Apache 2.0 License) for publishing, discovering, and verifying AI capabilities across the web. It's a collaborative effort, not one vendor's project: contributors come from Google, Microsoft, Hugging Face, GoDaddy and others. It answe
Code Mode MCP Pattern
Code Mode is a pattern for using MCP) tools where, instead of exposing each tool directly to the model as a callable function, you give the model a typed API (TypeScript or Python) and ask it to write code that calls those tools. The code runs in a sandbox, and only the final result comes back to th
TypedMark
TypedMark is an open specification for typed Markdown note systems. It adds explicit structure (schemas, field definitions, property sets, note-type inheritance, and validation) while keeping notes as plain Markdown files with YAML frontmatter. Authored by Sébastien Dubois under the MIT license (202
Agent2Agent Protocol (A2A)
Agent2Agent (A2A) is an open protocol for agent-to-agent interoperability, introduced by Google in 2025 and later donated to the Linux Foundation. It lets independent AI Agents, built by different teams on different frameworks, discover each other, exchange messages, and delegate tasks over a standa
Mellum2
Mellum2 is JetBrains' open-weight LLM) for AI workflows, released open source under Apache 2.0 in June 2026. It is the successor to Mellum: where the original focused on code completion, Mellum2 handles both natural language and code. JetBrains positions it as engineered for the hard parts of produc
OpenSRE
OpenSRE is an open-source (Apache 2.0 License) framework from Tracer Cloud for building AI agents that investigate and respond to production incidents. The problem it tackles: when something breaks in production, the evidence is scattered across logs, metrics, traces, runbooks, and Slack threads. Op
CocoIndex
CocoIndex is an incremental data framework that keeps the context behind AI agents continuously fresh by reprocessing only what changed. Open source, Rust core, Python on top. Apache 2.0.