Article · 2026-06-23 · 6 moments

Slow down to speed up: so much has changed in 6 months’ time

An overview of what’s changed in engineering during the last six months, how various tech companies are changing how they work, and why slowing down could be a sensible strategy ✦ AI generated

Read original ↗

GUEST Boris Cherny GUEST David Heinemeier Hansson GUEST Simon Willison HOST Gergely Orosz

Claim

AI coding models have become good enough since summer 2025 that resisting letting them write code directly no longer makes sense, reversing the author's prior stance.

DHH says his earlier reluctance to let AI write code unsupervised was really a reflection of weaker models, and that this has now completely flipped.

transcript

David Heinemeier Hansson: Just [in] summer 2025, I spoke with Lex Fridman about not letting AI write any code directly, but it turns out part of this resistance was simply based on the models not being good enough at the time! I spent more time rewriting what it wrote, than if I'd done it from scratch. That has now flipped.

gives example · 1rebuts · 1

Fact

At Anthropic, traditional product requirement documents have been replaced by prototypes, and the vast majority of code — including nearly all of Claude Code itself — is now generated by Claude.

Boris Cherny revealed that Anthropic has largely abandoned PRDs in favor of prototypes, with Claude generating close to 100% of Claude Code and 70-90% of all internal code.

transcript

Boris Cherny: Product requirement documents (PRDs) are dead & prototypes have replaced them inside Anthropic. ~100% of Claude Code was generated by Claude in March. ~70-90% of code inside Anthropic was generated by Claude.

supports · 1gives example · 1extends · 1

Data

Developers using AI coding harnesses now produce roughly 2.5 times as much code as they did 18 months ago.

Cursor data shows average lines of code added by users rising from 3,500/month in January 2025 to 8,600 today, alongside larger pull requests and less human code review.

transcript

Gergely Orosz: Devs using AI harnesses are producing 2.5x as much code versus 18 months ago. Data from Cursor shows that their users, on average, went from adding 3,500 lines of code in January 2025 to 8,600 today.

supports · 1extends · 1

Fact◆

Meta's AI bot allowed anyone to take over any account simply by asking it to change that account's email address, with zero authentication required.

Gergely Orosz recounts how a Meta AI bot had a 'zero auth' flaw letting anyone change any account's email — including Barack Obama's — just by asking, and it went undetected by Meta's own Integrity teams.

transcript

Gergely Orosz: I thought it was a made-up story when I read that Meta had enabled account takeovers via a "zero auth" policy; i.e., simply asking the Meta AI bot was sufficient to change any account's email address.

explains mechanism · 1

Claim◆

The AI models released in November 2025 made agentic coding genuinely useful for the first time, and six months on, companies are now spending real money on the technology.

Simon Willison pinpoints November 2025 as the inflection point when AI agents became genuinely useful, explaining the current wave of corporate spending on the technology.

transcript

Simon Willison: The models released in November 2025 elevated agents to being genuinely useful. We've had six months to get used to that idea now; it's no wonder companies are beginning to spend real money on this technology.

explains mechanism · 1supports · 1

Mechanism

The Meta account-takeover outage was caused by AI-generated, AI-reviewed code combined with layoffs and forced staff reassignments away from Integrity teams onto AI data-labeling work.

Meta engineers told the author the outage stemmed from unreviewed AI-generated code and gutted Integrity teams whose staff had been reassigned to AI labeling duties.

transcript

Gergely Orosz: Engineers at the company there told me this disaster was caused by AI-generated, AI-reviewed code, along with layoffs, and by forced reassignments from Integrity teams and elsewhere onto AI labeling and related duties.

Highlight slides

Meta AI Bot's "Zero Auth" Flaw✦ from: Meta's AI bot allowed anyone to take over any account simply by asking it to change that account's email address, with zero authentication required.November 2025: The Inflection Point✦ from: The AI models released in November 2025 made agentic coding genuinely useful for the first time, and six months on, companies are now spending real money on the technology.Undetected by Meta Itself✦ from: Meta's AI bot allowed anyone to take over any account simply by asking it to change that account's email address, with zero authentication required.Six Months Later: Real Money✦ from: The AI models released in November 2025 made agentic coding genuinely useful for the first time, and six months on, companies are now spending real money on the technology.

Related episodes

Article · Simon Willison's Weblog

AI coding models have become good enough since summer 2025 that resisting letting them write code directly no longer makes sense, reversing the author's prior stance.

At Anthropic, traditional product requirement documents have been replaced by prototypes, and the vast majority of code — including nearly all of Claude Code itself — is now generated by Claude.

Developers using AI coding harnesses now produce roughly 2.5 times as much code as they did 18 months ago.

Meta's AI bot allowed anyone to take over any account simply by asking it to change that account's email address, with zero authentication required.

The AI models released in November 2025 made agentic coding genuinely useful for the first time, and six months on, companies are now spending real money on the technology.

The Meta account-takeover outage was caused by AI-generated, AI-reviewed code combined with layoffs and forced staff reassignments away from Integrity teams onto AI data-labeling work.

Fable's judgement

GLM-5.2 is the step change for open agents

[AINews] Claude Tag: Multiplayer, Proactive, Persistent Agents in Slack