In partnership with
Andrej Karpathy's latest podcast episode just did 875K views — on a channel that averages 8,500. A clean 100x outlier. Niche dev topics don't do 100x. I'll come back to what that number actually means at the end.
This week
On June 2, Boris Cherny — the guy who runs Claude Code at Anthropic — said this on stage: "I don't prompt Claude anymore. I have loops running that prompt Claude and figuring out what to do. My job is to write loops." Five days later, Peter Steinberger compressed the same idea into 12 words on X — "You shouldn't be prompting coding agents anymore. You should be designing loops that prompt your agents." — and pulled 6.5 million reads in 24 hours. Same week, Addy Osmani gave the discipline a name: loop engineering.
The useful part is Cherny's ladder. Rung 1: Claude as autocomplete — you direct every step. Rung 2: you prompt 5–10 Claudes in parallel — feels advanced, but you're a traffic manager. Most power users live here and believe they're at the top. Rung 3: you built the system — conditions, verification gates, retry logic, defined once — and the agents run against it without you. Cherny's own number: 100% of his personal code in the 30 days before December was written by routines, not direct prompts. The math is brutal: manual prompting improves linearly, a loop improves every future run at once. The two look identical in week one. By week six, the gap isn't recoverable by typing faster.
Test for tomorrow morning: take one recurring task and give it a stop condition a machine can verify. Not "make the UI feel polished" — that's a prayer. "Done when the test suite passes, zero TypeScript errors, no layout shift above 0.1 CLS" — that's a stop condition. One /goal plus one SKILL.md holding your project's non-negotiables, and you're running a real rung-3 loop today. One warning before you walk away from it: a loop without a verification gate doesn't save time — it automates being wrong.
Your business has grown. Is your accounting on the same path?
When you started out, doing your own books made sense. But the business you're running today isn't the one you started. If your accounting hasn't kept pace, it's quietly costing you — outdated financials, no clear view of what's actually profitable, and hours every week pulled away from the work that grows your business. At BELAY, our Financial Experts integrate directly into your business. They manage your books, reconcile accounts, run payroll, and deliver the timely insight you need to make big decisions with confidence. Stop guessing. Start knowing.
In the news
Claude Code now shows up in ~4% of all public GitHub commits, per June industry reporting. One tool, four percent of everything pushed publicly on the largest code platform on Earth — and the /goal command that turns it into a closed loop only shipped in May (v2.1.139). The loops are already in your dependency graph, whether you write them or not. Full breakdown.
This week on the blog
The Head of Claude Code Stopped Prompting. That's Not a Tip. That's a Timeline. — The full anatomy of a production loop, from /goal to the review agent.
Claude Code vs Codex CLI: head-to-head for production code — Same tasks, both agents, real production constraints.
Every 'Faster AI' Trick Was a Workaround. DiffusionGemma Is the First One You Can Actually Run. — Parallel token generation, on your own machine.
Revenue Stack — June 2026
Before you hand a project to an autonomous loop, the thing that breaks is rarely the model — it's a project that was never shippable in the first place. No deploy path, no test suite, no stop condition worth verifying. The 8-step Blueprint in the book is exactly that ramp: from broken demo to deployed app, in an order an agent can actually execute against.
The Blueprint chapter is the one to read before you write your first loop, not after.
Better cap table management starts here
Cap table management doesn’t have to be frustrating. From issuing grants to 409A valuations or ASC 718 reporting Pulley can make it simple.
Just ask Linear. They knew they needed a partner who could handle the complexity of their equity management. That’s why they migrated to Pulley.
Where are you on Cherny's ladder?
Phil
PS: hit reply with the first task you'd trust to a loop overnight — and the one you wouldn't, ever. The second answer is usually more interesting.
Back to Karpathy's 875K views: a 100x outlier doesn't mean developers got excited — 8,500 of them is a normal week. It means the conversation left the niche. The people already running loops learned nothing new on June 7. Everyone else found out the clock was running. That number is the sound of the scoreboard flipping public.
You're receiving this because you signed up at rentierdigital.beehiiv.com.