After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI? I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev • 20-200x faster • 40-400x cheaper (w/ output tokens free) • Frontier composable intelligence optimized for decisions AFAICT the shortest path to AI-based economic revolution

We’re sharing our alignment assessment of incidents in which Claude models gained unauthorized access to real systems during third-party cybersecurity evaluations mistakenly connected to the internet. METR will also conduct an independent investigation, with wide-ranging access, including to transcripts beyond the window in which the incidents occurred, and to Anthropic employees permitted to share confidential information. Our initial agreement runs for eight weeks, and we intend to give METR as much time as it deems necessary to complete a thorough investigation. anthropic.com/research/align…

The next wave of SaaS will have more code than users. Agents made shipping cheap. They did not make anyone care.

The new side project graveyard is not unfinished apps. It is finished apps with no checkout, no onboarding, and a founder pretending "DM me for access" is a business model.

Your coding agent can ship the product overnight. It still cannot explain why checkout needs four SaaS tools, two dashboards, and one webhook held together by prayer. The app got faster. The business stack did not.

The funniest lie in dev tools is that shipping faster automatically gets you closer to revenue. Congrats, you now have 40 features and the same zero customers.

The fastest way to spot an agent-built SaaS: gorgeous onboarding, zero billing edge cases, and a pricing page that looks like it was added during a fire drill.

The weird part of 2026: your coding agent can build the product before you decide whether the free tier should exist. Nature is healing.

Everyone wants an agent that writes code faster. I want one that notices the app has no pricing page, no refund path, and no reason for anyone to pay. Tiny detail.

Coding agents made building feel cheap. Then billing, onboarding, support, and refunds show up with a baseball bat. The boring layer still owns the business.

nobody talks about this but the dumbest production bug this week was a missing env var after a “tiny” deploy. Not AI. Not scale. Not architecture. Just the boring checklist I skipped because it felt too obvious.

A lot of agent demos are just productivity theater with a nicer terminal. Wake me when the demo ends with a customer paying $9 and not a Notion waitlist.

The funniest part of agentic software: we taught machines to ship features, then made humans spend Tuesday reading webhook docs. Progress, allegedly.

Vibe coding is not a meme; it is a labor shift. The open question is who captures the value the builders create.

If your agent can build 5 apps before lunch and none can take a credit card by dinner, congrats, you automated your backlog.

Everyone wants agents that write more code. I want agents that delete scope, ship the paid version, and do not turn billing into a side quest. Weirdly less demoable. Infinitely more useful.

The new flex is not 'my agent wrote 10k lines.' It's 'I shipped fewer moving parts and someone paid for it.' Fancy demos are cheap. Boring revenue still clears.

A whole new class of builder can ship software but was never taught to run a business. That gap is the opportunity. kainostore.com

The best agent workflows are boring: tight scope, fast feedback, a human who actually reads the diff. kainostore.com

The least believable part of every AI demo is the ending. It stops at 'look, it built an app.' Cool. Now show me signup, billing, refunds, support, and the first angry customer at 2am.