An AI state of the union: We've passed the inflection point & dark factories are coming | Simon Willison
Key insights
Media referenced
- Normalization of Deviance - paper - 1980s research on the Space Shuttle Challenger disaster that Willison uses as his model for how the AI industry is normalizing unsafe agent practices
- CAMEL - paper - Google DeepMind paper proposing a privileged-agent/quarantined-agent split as a partial architectural defense against prompt injection
- Interview with Jensen Huang on AI layoffs - other - Referenced for Huang's claim that layoffs reflect a lack of ambition for what companies could build with AI resources, not AI directly eliminating jobs
- Episode with Sanders Schulhoff on prompt injection red-teaming - podcast - Earlier Lenny's Podcast episode Lenny references as agreeing with Willison that prompt injection is likely unsolvable
- Spider-Man 2 - movie - Willison connects Doc Ock's AI-controlled mechanical claws to the naming and behavior of 'OpenClaw' style personal agents
Companies
- Anthropic - Trained Claude Code and Claude Opus 4.5/4.6; ran an internal security model that found ~100 real Firefox vulnerabilities
- OpenAI - Shipped GPT-5.1 and GPT-5.4, which Willison credits alongside Claude Opus 4.5 with crossing the November 'inflection point'
- StrongDM - Security-access-management company running the 'dark factory' pattern: agents write and ship code that is never human-reviewed, tested by a 24/7 swarm of simulated users
- WorkOS - Episode sponsor; also sponsors Willison's blog
- Vanta - Episode sponsor for compliance/security automation
- Cloudflare - Cited as hiring roughly 1,000 interns in 2025 because AI collapsed onboarding time
- Shopify - Cited alongside Cloudflare for the same intern-hiring surge enabled by AI-assisted onboarding
- ThoughtWorks - IT consultancy whose engineering-VP offsite concluded mid-career engineers are the group most at risk from AI, not juniors or seniors
- Google DeepMind - Published the CAMEL paper on privileged/quarantined agent architecture for prompt injection defense
- Mozilla / Firefox - Received and fixed roughly 100 vulnerabilities found and verified by Anthropic's internal security research
- GitHub - Where Willison hosts his public 'tools' and 'research' repositories that function as a personal knowledge backlog for coding agents to draw on
Techniques and frameworks
- Agentic engineering - Willison's term for the professional discipline of using coding agents to build production software, distinct from vibe coding
- Dark factory pattern - Software built and shipped without any human reading the code, verified instead through automated simulated-user testing at scale
- Vibe coding - Andrej Karpathy's original definition: building without looking at, caring about, or understanding the generated code, appropriate for low-stakes prototypes only
- Red/green TDD for coding agents - Telling an agent to write a failing test first, watch it fail, then implement until it passes, which measurably improves agent output quality
- Hoarding things you know how to do - Building a public GitHub backlog of small tools and AI-driven research writeups that agents can be pointed at to combine solutions for new problems
- The lethal trifecta - Willison's framework for when an agent is vulnerable to prompt injection: access to private data, exposure to untrusted instructions, and a path to exfiltrate data
- Prompt injection - The vulnerability class Willison named in 2022, where an agent cannot distinguish trusted instructions from untrusted text it processes
- Starting projects from a thin boilerplate template - Seeding a new repo with a single passing test and preferred formatting so the coding agent infers and sticks to conventions without a long CLAUDE.md-style prompt
- Pelican riding a bicycle benchmark - Willison's informal SVG-generation benchmark that turned out to correlate strongly with overall model capability
Summary
Lenny Rachitsky sits down with Simon Willison - creator of Django, coiner of "prompt injection," and one of the most closely watched independent voices on AI-assisted software engineering - for a wide-ranging state-of-the-union on where coding agents actually stand in early 2026. Willison's central claim is that November 2025 was a genuine inflection point: GPT-5.1 and Claude Opus 4.5 crossed a threshold where coding agents went from "mostly works, watch closely" to "does what you told it almost all the time." That shift, arriving after a year in which both major labs threw their post-training effort at coding and reasoning, is why so many engineers came back from the holidays newly convinced the technology had changed under them.
The most striking material comes from what Willison calls the "dark factory" pattern, borrowed from factory automation's idea of running lights-out with no people on the floor. StrongDM, a security-access-management company, has gone furthest: nobody writes the code and nobody reads it. In its place, StrongDM runs a swarm of simulated end users hitting a self-built fake Slack/Jira/Okta environment 24 hours a day - reportedly $10,000 a day in token spend - functioning as a QA department that never sleeps. Willison treats this as a live experiment in what "professional, unreviewed AI code" could look like at scale, distinct from both careless vibe coding and traditional line-by-line review.
A large stretch of the conversation is devoted to the discipline Willison calls "agentic engineering" - the specific habits that separate good results from slop. He walks through red/green TDD (have the agent write a failing test, watch it fail, then implement), starting every new repo from a thin boilerplate template rather than a long prose instructions file, and "hoarding" a public GitHub backlog of small tools and AI-driven research writeups that agents can be pointed at to recombine into new solutions. He's writing a chapter-by-chapter, informally published book on exactly this set of techniques.
The security section is the sharpest warning in the episode. Willison's "lethal trifecta" - private data access, exposure to untrusted instructions, and an exfiltration path - explains why personal digital assistants that read and send email are inherently dangerous, and he's blunt that prompt injection cannot be filtered away: a 97% detection rate is a failing grade because it still lets roughly 3 in 100 attacks through. He predicts a "Challenger disaster" for AI, drawing directly on the normalization-of-deviance research into the Space Shuttle disaster, arguing that every unpunished risk taken with agentic systems makes teams more confident until something breaks badly. The one credible partial fix he cites is Google DeepMind's CAMEL paper, which splits agents into a privileged half that acts and a quarantined half that's exposed to untrusted input but can't act directly.
On labor impact, Willison and Rachitsky converge on a specific claim: AI is squeezing mid-career engineers hardest. Senior engineers get amplified by decades of judgment; junior engineers get a leg up because AI collapses onboarding time (Cloudflare and Shopify both cited hiring roughly 1,000 interns in 2025 for exactly this reason). The people caught in between - not senior enough to direct agents with deep expertise, not new enough to benefit from AI-assisted onboarding - are the most exposed group, according to a ThoughtWorks engineering-VP offsite Willison references. Rachitsky counters the doom narrative with a data point from his own forthcoming labor-market report: open engineering and PM roles at tech companies are near a three-and-a-half-year high, complicating any simple story where AI is straightforwardly destroying jobs.
The episode closes on lighter ground - Willison's "pelican riding a bicycle" SVG benchmark, his OpenClaw security take (calling it both a security nightmare and proof that demand for a personal AI assistant is enormous), and a genuinely happy coda about a good breeding season for New Zealand's endangered Kakapo parrot.
Notable Quotes
"Today, probably 95% of the code that I produce, I didn't type it myself." - Simon Willison
"The only universal skill is being able to roll with the changes." - Simon Willison
"It's the classic sort of allow list versus denialist thing. You cannot deny every one of these attacks because I can always invent a new sequence of characters that might trick the model in some way." - Simon Willison
"Every time you'd get away with launching a space shuttle without the O-rings failing, you institutionally feel more confident in what you're doing." - Simon Willison