Claude Code, AI-native engineering, and building the future with LLMs.
AI
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
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
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.
AI
There are two ways to use a coding agent: sit in its terminal, or take its face off and drive the engine from your own code. If you're building anything around agents, the second mode is the whole game — and it's full of traps nobody warns you about.
AI
Ralph is a five-line bash loop that forgets everything every iteration — and the amnesia is the whole point. It's also wrapped in more hype than almost anything in AI. Here's what's real, what isn't, and why Anthropic's official plugin gets it backwards.
AI
I built a gate that wouldn't let Claude fake “done.” It worked — and then I spent a month babysitting the gate. The native trio that replaces the whole rig was shipping the entire time, and the hard part was never the loop.
AI
/loop dies the moment you close your laptop, and I argued that leash was a safety feature. /schedule cuts it: your code keeps shipping on Anthropic's cloud while your machine sleeps — and the thing that made the loop safe is gone.
AI
/loop re-runs a prompt on a timer, inside your session — Boris Cherny babysits his own PRs with it. The catch: it's the same conversation, so a forgotten loop can quietly run up a serious bill.
AI
Claude Code's /goal works across turns until a condition holds — judged by a second, blind model. The catch: it forces you to learn the one skill most engineers are bad at, naming a finish line a machine can actually check.
AI
Autonomy isn't a permission you grant an agent — it's a property of the environment you engineer. A worked example: the two-layer verifier I own, the self-driving loop around it, the lessons from weeks of running it, and exactly how to run your own.
AI
Anthropic just put @Claude inside your Slack channels — one shared, persistent, agentic teammate per channel, billed to the org and learning the company over time. The genuinely new part isn't that Claude can read Slack. It's that AI just went multiplayer.
AI
I wanted ultracode as my Claude Code default. I reached for "effortLevel": "ultracode" in settings.json — it silently did nothing. Here's what ultracode actually is, the only three ways to turn it on, and the config that actually sticks.