AI

AI

Make Love Not War

Two rival frontier models, both at maximum reasoning, in one loop that ships unattended for days. It runs $500–$1,000 a day — and measured against what it delivers, that's cheap.

AI

Sol Ships: GPT-5.6 in Your Pro Plan, at Half Fable's Price

The Sol/Terra/Luna preview is now real. GPT-5.6 shipped July 9 — and the story isn't the benchmarks. It's that Sol is bundled into ChatGPT Pro at half of Claude Fable 5's token price, plus one coding benchmark OpenAI quietly skipped.

AI

J-Space: The Silent Workspace Inside Claude

Anthropic found a small, privileged workspace inside Claude — internal patterns, readable with a new Jacobian 'J-lens', holding concepts the model reasons with but never writes down. Not a consciousness claim. Maybe something more useful: a window into the thoughts a model doesn't print.

AI

Sol, Terra, Luna: OpenAI Splits GPT-5.6 Into Three

OpenAI isn't shipping GPT-5.6 as one model. It's shipping a ladder — Sol, Terra, Luna — that turns the frontier into a pricing decision. Public launch Thursday. Here's how to read the three tiers and where each one actually earns its cost.

AI

Advisor and Executor: Using Fable 5 Without the Fable 5 Bill

Anthropic's ClaudeDevs shared the pattern they lean on with Fable 5: don't make the frontier model do the work — make it the advisor. A cheaper executor runs the loop and calls up to Fable 5 only at the decisions that matter. Most tokens bill at the cheap rate; judgment stays near the top.

AI

GPT-Live: OpenAI Bets the Interface Is Voice

OpenAI's GPT-Live rolls out in ChatGPT today: full-duplex voice models that listen and speak at the same time, and quietly hand the hard questions to a frontier model in the background. The interface just stopped being a prompt box.

AI

The Making of Claude Code

Anthropic published the oral history of Claude Code — from a safety-research side project nicknamed clide to the tool that changed how Silicon Valley writes software. The interesting part isn't the launch. It's the four quiet years before it, and the shrug it got after.

AI

The Third Leash

Fable 5 is back, but it's not in my Max plan — using it spends separate credits. So I made one folder run Claude Code on a different account and the Fable model, interactive and headless. The shim, the token, and why the headless loop is the hard part.

AI

The Loop Files Its Own Work

Last time I proved a loop could finish a batch once, with me still in the room. This time it caught its own unfinished work mid-run, filed the ticket, shipped it, and hardened its own machine on the way. Here's the whole system, drawn from zero, and the honest ledger of where it still needed me.

AI

Fable 5 Comes Off the Leash

Anthropic's most capable widely released model was frozen by a US export-control order three days after launch. Eighteen days later the controls came off — and the reason it came back says more than the reason it left. The whole story, in one place.

AI

You Can't Screenshot Your Way to a Pixel-Perfect UI

Hand an agent a screenshot and ask for a pixel-perfect frontend with no human review, and you get a confident guess. The fix isn't a smarter model — it's turning “respect the design” from a vibe into a contract a machine can check.

AI

Proof of Loop

This whole series I've been designing a harness that would let an agent finish work unattended. I finally pointed it at a real batch and walked most of the way out of the room. Eleven tickets, two repos, six hours, zero production incidents — and an honest count of every time it still needed me.