mager.co

tmux is a 2007 terminal multiplexer that turns out to be the most native orchestration layer AI agents have: one session per agent, driven by any principal — human or model — with send-keys and capture-pane.

AI coding assistants confidently build the wrong thing when requirements live only in chat.

I'm moving off single-provider AI subscriptions toward a stack of parts — Eve for agents, Vercel AI Gateway as the primary model access and billing layer with no-markup provider pricing, and OpenCode Go kept as the fallback — and the enterprise version of…

OpenCode's CLI is bigger than 'type opencode and start a session.'

Second harness migration in two months.

Blistered cherry tomatoes and garlic collapse into a full pasta sauce in twenty minutes — no peeling, no seeding, finished with butter and pasta water.

Anthropic deleted 80% of Claude Code's system prompt for Opus 5 with no measurable loss.

Block's open-source Nostr workspace puts people and agents on the same cryptographic footing — and lands at the end of a long chain of thinking about where always-on agent infrastructure should actually live.

A Skill is packaged know-how.

OpenRouter lets you pick a different model for every step in a pipeline.

Instead of reading someone else's leaderboard, build a small set of tasks you actually care about and run them yourself every time a new model drops — Simon Willison's SVG pelican test, but for code.

In March I wrote the theory.

Most evals talk is about grading model output.

A small Python voice agent that remembers the thread, streams Claude's reply to the terminal, and speaks it aloud through ElevenLabs — no ffmpeg, just afplay.

Define your agent in a directory, deploy it to Vercel's cloud with one command, and access it from anywhere.

A Skill is reusable know-how Claude reaches for on its own.

How I worked with Claude through five rounds of image generation to design a logo for my Japanese learning app — and ended up inventing a kanji that hides a smile.

Three real ingredients — pecorino, pepper, pasta water — plus a knob of butter for insurance, tossed into a glossy sauce that never breaks.

A plain-English walkthrough for setting up your own always-on AI assistant on a Mac mini — OpenClaw, Google Gemini, and Tailscale — written for a first-timer.

A curated collection of high-quality skills for people who don't code — and an experiment in what actually makes a skill good.

I love OpenClaw.

A month that turned the "agentic turn" from talking point to shipping product.

A five-ingredient Japanese-style spaghetti — butter, tamari, and parmesan tossed with hot pasta and finished with green onion.

Microsoft's SkillOpt is the first paper to treat agent skill files as trainable parameters — propose an edit, evaluate on held-out examples, accept only on strict improvement.

OpenHuman is a desktop-first agentic assistant with persistent memory, 118+ OAuth integrations, and a token compression layer.

Karpathy's four rules for agentic coding are worth reading — having them written down in a shared format is a useful starting point for anyone building with Claude Code.

Hanshin Tigers vs.

阪神タイガース vs 中日ドラゴンズ @甲子園 — ライト側の応援席、売り子さんのビール、7回裏の風船、そしてサヨナラホームランで終わった忘れられない夜。

How I moved magerbot's brain from @-imported markdown files into gbrain's Postgres-native semantic memory layer — what broke, what the gotcha was, and why the context model is fundamentally better.

We missed the original ticket sale, got rescued by a tour, and spent an afternoon learning how much more fun sumo is when someone helps you understand what you're watching.

チケット争奪戦に出遅れ、ツアーに救われ、解説付きで観ると相撲がいかに面白いかを実感した一日。

I built a 200-line harness called conseiller to test Anthropic's new advisor tool — a fast executor model that consults a stronger model mid-generation.

I built a Go Bubble Tea starter for local model servers, used Gemma 4 through llama.cpp, and split the TUI into llocal.

I'd been seeing chatter about Hermes Agent from Nous Research, so I installed it locally and put it to work on this blog.

A practical explainer for both developers and everyday Claude users: what prompt caching is, what gets reused, what breaks it, and how to make long sessions cheaper and faster.

A fennel-forward Italian spice blend that turns any ground meat into proper sausage

A simple set of habits I use to keep long AI coding sessions from getting bloated: better one-shot prompts, matching model and thinking level to the job, understanding cache behavior, and using cheaper orchestrators when it makes sense.

I reverse engineered several of my own sites into DESIGN.md files to see how much of a design system can actually be described, and why writing down design intent might be more reusable than it looks.

A practical tour of Claude Code flags that are easy to miss but genuinely useful once you move past the default interactive loop.

A bright, high-impact rice finished with garlic, lots of cilantro, and fresh lime juice added after cooking.

Anthropic shutting down OAuth-based Claude Code access forced my hand.

The Y Combinator CEO open-sourced his entire Claude Code workflow.

I tested Anthropic's official Claude plugins for knowledge workers.

I used Gemini to write a Loooom skill, installed it in Claude Code, and got a full audio analysis report on a 37-second piano recording of Espresso.

I rebuilt the beatbrain backend in an afternoon.

I built a Japanese learning site in a morning because I wanted something I could pull up on my phone and just look at characters.

Claude Code's new channels feature lets you push messages from Telegram and Discord into a running session.

Dogfooding Karpathy's autoresearch pattern on my own skill marketplace.

Everyone's talking about building a software factory.

12 hours before my bracket was due, I used Gemma-3-27b to generate unique insights for all 32 first-round games.