Boris Cherny: Building Claude Code
Boris Cherny, creator of Claude Code, discusses how Anthropic's Opus 5 model represents a major leap in AI capabilities, requiring a fundamental rethinking of how to build AI products. Rather than accumulating complexity, successful builders must regularly delete system prompts and code to "unhobble" the model, allowing it to tackle increasingly complex tasks like rewriting entire codebases and maintaining software automatically.
Summary
Boris Cherny shares insights from building Claude Code, Anthropic's most advanced coding harness. He explains that Opus 5, just released, achieves a 30% score on Arc AGI—a massive improvement from previous single-digit performance. The model can now run for weeks or months continuously and demonstrates unprecedented resistance to prompt injection attacks through a combination of alignment research and mechanistic interpretability techniques that detect injections at the neural level.
A counterintuitive finding: Anthropic deleted over 80% of Claude Code's system prompt when upgrading to Opus 5. Each new model generation requires ablation studies—methodically removing every instruction to understand what's actually necessary. This stems from the reality that each model behaves differently with slightly different capabilities and "personality." Rather than treating AI development like traditional software engineering with careful upfront design, Cherny advocates treating models as "living creatures" requiring empirical, scientific approaches.
Cherny introduces the concept of "product overhang"—capabilities that models possess but products fail to elicit. The original Claude Code solved this by giving Opus 3.5 the simplest possible harness to write entire files and features, rather than limiting it to single-line autocomplete like competitors. Examples of current product overhang include language-to-language codebase rewrites (a bun team member used Claude to rewrite a 100,000+ line JavaScript runtime from Zig to Rust in 11 days), image generation via OpenCV, and complex multi-stage orchestration tasks.
For spawning thousands of agents productively, Cherny describes dynamic workflows—an algebra for agents allowing sequential and parallel execution within a sandbox. Anthropic now uses Claude to maintain itself through daily routines: cleaning dead code, unifying duplicate abstractions ("abstraction police"), deleting obsolete tests, and shipping experiments. A notable example: Boris's two-week ongoing prompt to rewrite Claude's Electron desktop app in Swift while pixel-by-pixel comparing outputs.
On skill development, Cherny emphasizes that successful Claude users share an empirical mindset—forget prior knowledge, observe where the model struggles, and iterate. He advises students to learn computer science practically through real problems, combining technical skills with product sense, business understanding, and user empathy.
Key Insights
- Anthropic deletes over 80% of Claude Code's system prompt with each major model release and rebuilds it incrementally through ablation studies, removing lines one-by-one to measure impact, because each new model generation has fundamentally different behaviors and capabilities
- Opus 5 demonstrates unprecedented prompt injection resistance through three-layer defense: well-aligned model training (3 years of research), a mechanistic interpretability-based classifier that detects neurons lighting up during injection attempts, and an auto-mode classifier
- Product overhang describes situations where models possess capabilities that products prevent them from expressing; the original Claude Code solved this by removing scaffolding and giving models write access to write entire files and features, capability that existing products only elicited at single-line autocomplete level
- A bun team member used Claude to rewrite a 100,000+ line JavaScript runtime from Zig to Rust in a single 11-day dynamic workflow run with steering—a task that would have taken traditional engineers over a year
- Anthropic now uses Claude to maintain its own codebases through daily routines spawning hundreds to thousands of agents per day, including dead code deletion, test writing, duplicate abstraction unification (abstraction police), and experiment shipping—work previously requiring dozens or hundreds of engineers
Topics
Transcript
[0:00] All right, Boris, we're so excited to have you here, the creator of Cloud Code. Thank you. >> It's great to be here. >> Fresh off the press, you guys just shipped Opus 5 yesterday. >> Yes. And it seems that model performance [0:31] keeps accelerating. You guys got took Arc AGI 3 to 30%. Which is incredible. >> Yes. >> And for context, before the best score was in in the low single digits or low teens, right? >> What can Opus 5 do now that it couldn't versus a previous version? Yeah, there's um there's a lot that goes into every new model and there's a lot of new capabilities that we teach and uh [1:03] get…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Y Combinator
Self-Maintaining APIs
The speaker argues that API communication is fundamentally broken across the industry, with breaking changes and new features going unnoticed. Since agentic coding tools have proven viable and developers accept external tool access to codebases, API providers should automatically apply changes to customer code rather than just announcing them—similar to how Dependabot works for dependencies.
What Actually Makes A Startup Durable
YC partners discuss what makes startups durable in the AI era, emphasizing that while AI reduces software development costs, founders must focus on harder problems with defensible moats, maintain direct customer contact, and prioritize co-founder relationships for emotional and strategic support. They argue that pure software businesses are no longer durable and that the pace of AI improvements means cost concerns will self-resolve within 6-12 months.
Proving You’re Human
Deepfakes and AI-generated voice clones have made digital fraud increasingly sophisticated and harder to detect, with a recent $25 million wire fraud involving deepfake video call participants exemplifying the problem. The traditional trust signals of seeing faces and hearing voices are no longer reliable, necessitating a fundamental rebuild of the internet's trust layer to verify that humans—not AI—are on the other end of communications.
What Big Tech Missed And How Startups Can Still Win
Alex Lebrun, CEO of Amilabs, discusses his journey building multiple AI startups and explains why world models—AI trained directly on sensory data rather than text—represent the next frontier beyond large language models. He argues that being early in technology requires specific founder strengths, and shares lessons on ambition, fundraising, and the challenge of managing a $1.2 billion seed round.
Why Physical AI Is the Next Platform Shift
Eric Landau, co-founder of Encord, discusses his career trajectory from particle physics to quantitative trading to founding a data infrastructure company for AI. He explains how Encord pivoted from general vision AI to focus on Physical AI, which he believes will be a massive market as robotics proliferates across manufacturing, logistics, and autonomous vehicles.