Intel / podcast

An AI state of the union: We’ve passed the inflection point, dark factories are coming, and automation timelines | Simon Willison

Source: lennysnewsletter.com
Author: Lenny Rachitsky
Date: Published Apr 2, 2026

Context

A coding 's tests are more informative when it first demonstrates that they can fail. Software developer Simon Willison recommends having the agent write a test, run it before implementation, then make it pass. The initial failure checks that the test can detect the missing behavior; the later pass supplies evidence that the implementation addresses it. Keeping those tests also helps detect when subsequent changes break covered behavior. His argument explains how verification can support faster iteration, with the protection limited to what the tests actually exercise.

Terms in this piece · Glossary

AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
prompt injection — An attack that hides instructions in content an AI will read — a webpage, email, or document — tricking it into following the attacker instead of the user.

Key quotes

“Using coding agents well is taking every inch of my 25 years of experience as a software engineer.”
Simon Willison

“I can fire up four agents in parallel and have them work on four different problems. By 11am, I am wiped out.”
Simon Willison

“Today, probably 95% of the code that I produce, I didn't type it myself.”
Simon Willison

“We've been using these systems in increasingly unsafe ways. This is going to catch up with us.”
Simon Willison

“The code is now good. It's good code for my definition of good code at least.”
Simon Willison

Chapters

00:00Introduction to Simon Willison
02:40The November 2025 inflection point
08:01What's possible now with AI coding
10:42Vibe coding vs. agentic engineering
13:57The dark-factory pattern
20:41Where bottlenecks have shifted
23:36Where human brains will continue to be valuable
25:32Defending software engineers
29:12Why experienced engineers get better results
30:48Advice for avoiding the permanent underclass
33:52Leaning into AI to amplify your skills
35:12Why Simon says he's working harder than ever
37:23The market for pre-2022 human-written code
40:01Prediction: 50% of engineers writing 95% AI code by end of 2026
44:34The impact of cheap code
48:27Simon's AI stack
54:08Using AI for research
55:12The pelican-riding-a-bicycle benchmark
59:01The inherent ridiculousness of AI
1:00:52Hoarding things you know how to do
1:08:21Red/green TDD pattern for better AI code
1:14:43Starting projects with good templates
1:16:31The lethal trifecta and prompt injection
1:21:53Why 97% effectiveness is a failing grade
1:25:19The normalization of deviance
1:28:32OpenClaw: the security nightmare
1:34:22What's next for Simon
1:36:47Zero-deliverable consulting
1:38:05Good news about Kakapo parrots

Read the source www.lennysnewsletter.com