Source
Interview with Boris Cherny, creator of Claude Code, at Sequoia Capital AI Ascent 2026 (2026-05-06).Key points
1. Who still writes code by hand?
Opening his talk with a joke about "Claude Code addiction," Cherny disclosed that he has personally written zero lines of code in 2026. He now works primarily from the iOS Claude app, merging dozens of PRs daily with a record of 150 in one day, while running hundreds of agents in parallel and thousands of asynchronous agents overnight.2. Programming: solved?
Cherny polled the audience: few still hand-write 100% of their code; nearly everyone sits somewhere in between. "Programming is 50% solved," he said—though for him personally it has been fully solved since October–November of last year. Very large complex systems and unfamiliar niche languages remain challenges, but he views these as timing issues, not technical ones.3. Loop: from Q&A to autonomous operation
Cherny's favorite feature, Loop, lets agents run recurring tasks on cron schedules. He personally runs dozens:- One manages PRs: auto-fixing CI and rebasing
- One keeps CI healthy by auto-fixing flaky tests
- One pulls user feedback from Twitter every 30 minutes, clusters it, and pushes summaries
- No manually written code or SQL company-wide
- AI agents messaging each other on Slack
- Autonomous collaboration on unsolved problems
- Tool level: AI coding has jumped from autocomplete → taking tickets → multi-agent parallel repo maintenance.
- Organizational level: competitive advantage comes from restructuring processes around AI, not from merely owning tools.
- Human level: coding becomes literacy; the best industry software will be created by the people who understand the domain best.
- Boris Cherny interview, Sequoia Capital AI Ascent 2026, 2026-05-06
- Tencent Cloud Developer Community (Chinese digest), 2026-05-14, https://cloud.tencent.com/developer/article/2669104
- Zhidx report, 2026-05-08, https://t.cj.sina.cn/articles/view/5213469505/136bf3b41020015w2m
- Jike digest, 2026-05-11, https://m.okjike.com/originalPosts/69fdbc5feb714cce8c542067
- 51CTO report, 2026-05-09, https://www.51cto.com/article/842726.html
Anthropic also ships Routines—server-side Loops that run even when your machine is off. Notably, the Claude 4.7 model now *spontaneously* proposes creating Loops (e.g., suggesting a 30-minute report delivered via Slack, then wiring up the Slack MCP itself). Your role shifts from executor to dispatcher.
4. Mobile as the battlefield
Cherny runs five terminal tabs each with a Claude instance in plan mode, round-robining between them, and overflow to desktop and phone. Five to ten concurrent sessions with swarms of agents run in the mobile app's code tab. The bottleneck is no longer typing or thinking speed—it is context switching: managing many concurrent workflows efficiently.5. Everyone codes: the rise of the cross-disciplinary generalist
At Anthropic, everyone shares one title: Member of Technical Staff. This flattening is deliberate—if everyone is expected to do everything, nobody is pigeonholed. On the Claude Code team, engineering managers, PMs, designers, data scientists, and finance staff all write code with AI. Cherny predicts the most valuable future talent will be cross-disciplinary generalists spanning product + design + engineering or data science + product + engineering, since AI drives execution cost toward zero.6. Software's printing-press moment
Cherny likens AI coding to the 15th-century printing press: it did not eliminate writers but raised literacy from ~10% to ~70%. Coding will become a basic skill like texting. The best future accounting software will be built by accountants who know the business best—because programming is becoming cheap while domain knowledge remains hard. He even predicts Claude Code's own codebase may shrink to ~100 lines within a year as models need less scaffolding.7. A golden decade for startups
AI erodes incumbents' moats of switching costs and process capability, while startups can build AI-natively from scratch. Cherny predicts a 10x increase in disruptive startups over the next decade—teams of three to five can now compete with what once required 20 engineers. Network effects, economies of scale, and scarce-resource moats remain intact.8. The organizational generation gap
Anthropic internally has:The real divide is not having AI tools but weaving them into organizational operations. For models, MCP, APIs, and Computer Use are all just tokens—what matters is whether the required information and actions are accessible.
9. Product overhang
Claude Code was born accidentally in late 2024 inside Anthropic Labs, a small incubation team. They saw model capability had already outgrown inline-completion products—a "product overhang." After ~6 months without PMF, the turning point came with Opus 4 in May 2025; each subsequent model release (4.5, 4.6, 4.7) drove another growth inflection. Lesson: build for where technology will be in six months, not where it is today.10. Conclusion: you are the commander
> The real dividing line is not whether AI can write code, but whether your organization can fully reshape its processes so that everyone—PMs, designers, finance—becomes a builder.