Top 5 · 2026-05-26 · source-backed
Nolan Lawson Says Use AI to Write Better Code More Slowly. 662 Hacker News Points Agree.
Story
Nolan Lawson (ex-Microsoft, ex-Salesforce) published an essay that hit 662 points and 247 comments on Hacker News. His argument: stop using LLMs to ship faster. Use them to ship better. His approach runs multiple models to review code, ranks findings by criticality, and filters false positives before generating a final report. Speed isn't the point. Quality is.
This resonates with me because I've been living the same shift. I spent months optimizing for velocity with Claude Code in my personal projects. More subagents, faster task completion, bigger PRs. Then I started noticing patterns. The code worked. It passed tests. But it was mediocre. Not wrong, just bland. Missing the architectural awareness that comes from actually thinking about what you're building.
The timing is perfect because this essay arrives alongside three other signals pointing the same direction. George Hotz's "Eternal Sloptember" essay (462 points, 360 comments) argues AI coding agents will produce "buckets and buckets of slop," creating a golden era for quantity and a dark age for quality. His key insight: high performers can recognize slop while bottom performers produce 10x output without self-checking mechanisms. A new Constraint Decay paper (280 points, 190 comments) demonstrates that LLM agents fail when structural requirements accumulate in backend code. They're great at unconstrained generation but fall apart when they need to respect architectural patterns, ORMs, and database constraints. And a HollandTech essay "Claude Is Not Your Architect" (267 points) makes the case that AI agents are "pathologically agreeable" and can't provide the pushback that real architects deliver.
Four independent signals, all crystallizing around the same thesis: speed without taste produces slop.
What builders should do: Try Nolan's approach. Next time you'd use AI to generate code, use it to review code instead. Have one model write, another review, a third filter false positives. You'll ship slower. The code will be better.
Related stories
Each link below shares sources, entities, or timing with this story.
Shared entities / Shared topic / Earlier coverage / Tension
Stripe Shows Its Cards: Minions Blueprint Architecture Uses 400 Tools, Deploys ~15 Per Run, and Ships 1,300+ PRs Weekly
Both cover Claude Code, LLM, PRs, Then; overlapping topics (agent, code); earlier Claude Code coverage from 2026-03-25.
Shared entities / Same source domain / Shared topic / Earlier coverage / Tension
Your AI-Generated Code Compiles. It Doesn't Work.
Both cover LLM, LLMs, Speed; reported by the same outlet (arxiv.org); overlapping topics (agent, code).
10 actionable skills from today's findings:
Both cover LLM, LLMs, Next; reported by the same outlet (arxiv.org); overlapping topics (agent, code).
Shared entities / Same source domain / What happened next
Uber burned its entire $3.4B AI budget in four months, and the COO can't prove it did anything
Both cover Claude Code, Hacker News, PRs, Then; reported by the same outlet (news.ycombinator.com); picks up the Claude Code thread on 2026-06-02.
Shared entities / Same source domain / Shared topic / What happened next
Local models are good now, and the timing isn't a coincidence
Both cover Claude Code, Hacker News; reported by the same outlet (news.ycombinator.com); overlapping topics (agent, code, comment, point, same).
Shared entities / Same source domain / Shared topic / Earlier coverage
A Cursor Agent Deleted a Production Database in 9 Seconds. Then It Wrote a Confession.
Both cover Claude Code, Hacker News, Then; reported by the same outlet (news.ycombinator.com); overlapping topics (agent, same).
OpenCode: Open-Source AI Coding Agent Hits 802 HN Points, 95K+ Stars — The Real Claude Code Rival Has Arrived
Both cover Claude Code, Hacker News, LLM; reported by the same outlet (news.ycombinator.com); overlapping topics (agent, code).
Shared entities / Shared topic / What happened next
The bottleneck stopped being code generation. It's verification.
Both cover Hacker News, LLMs, Then; overlapping topics (agent, code, point); picks up the Hacker News thread on 2026-06-26.
Source trail
Entities
Provenance
- Canonical issue
- Ramsay Research Agent — May 26, 2026
- AI generated
- no
- Story unit
- 2026-05-26-nolan-lawson-says-use-ai-to-write-better-code-more-slowly-662-hacker-news-points-agree
- Labels
- source-backed, canonical briefing excerpt