Boris Cherny, creator and head of Claude Code at Anthropic, has not written a line of code himself in more than 6 months and believes coding as a profession is effectively solved. In an interview with Platformer's Casey Newton, Cherny predicted the title 'software engineer' could start to disappear by the end of 2026 as AI agents automate the core work of programming. The shift would replace traditional engineers with 'builders' — a broader category including designers, product managers, and others who ship code using AI agents.
Claude Code saw explosive internal adoption at Anthropic when Cherny first released it. 20% of the company's engineers started using the terminal-based prototype on day one, and 50% of the engineering team had adopted it within 5 days. The product began life as a simple API experiment in which Cherny built a tool to tell him what music he was listening to, using Sonnet 3.5 to generate AppleScript code he would not have written himself.
Cherny joined Anthropic in September 2024 after 5 years as a principal engineer at Meta. He leads a Labs team of 5 people who built Claude Code, MCP, Skills, and the desktop app over a span of months. The team shipped the products without a clear mandate from leadership — Cherny was simply trying to learn the API and stumbled into building the fastest-growing AI coding tool in the world.
Key facts
- 01Boris Cherny hasn't written a line of code in more than 6 months and says coding is 'solved' for his workflow.
- 0220% of Anthropic engineers used Claude Code on the first day; 50% adopted it within 5 days of internal release.
- 03Cherny predicts the software engineer title could disappear by end of 2026, replaced by 'builder' roles.
- 04Claude Code began as a terminal prototype to control music playback, built in a couple of days.
- 05Cherny spent 5 years as principal engineer at Meta before joining Anthropic in September 2024.
The Claude Code creator's jobs forecast is more optimistic than the 'software engineer is dead' framing suggests. While companies may hire fewer traditional engineers, Cherny expects 100 times more people writing or using agents to write code compared to today. The multiplication comes from non-engineers — designers, managers, analysts — who gain the ability to ship software using tools like Claude Code, expanding the builder population rather than shrinking it.
Cherny's workflow now relies entirely on agents. He recently used Cowork, an AI travel agent, to book 8 flights and 5 hotels with only one mistake — a $5,000-per-night hotel that exceeded his budget. The agent's ability to handle multi-step tasks autonomously, including writing code in languages Cherny does not know, has recalibrated his expectations of what automation can do. He spends every night and weekend iterating Claude Code and says the product still dreams about what to build next.
The terminal-based origin story reveals how far the product overhang stretched when Claude Code launched. Existing coding tools at the time of Sonnet 3.5's release were IDE extensions offering autocomplete, because the model's capabilities had not yet been productized. Cherny's prototype let the model do work it was already capable of — writing full programs, debugging, iterating — but no commercial product exposed. The gap between model capability and product design remains wide today, according to Cherny.
“It solved the problem in a way I wouldn't have as an engineer.”— Boris Cherny, Claude Code creator at Anthropic
Cherny's view contrasts sharply with Box CEO Aaron Levie and Google's James Manyika, both of whom argued in prior Platformer episodes that AI automates tasks but not jobs. Levie emphasized the 'last mile' of human labor that resists automation, while Manyika pointed to tech's historical pattern of improving productivity without eliminating roles. Cherny sees the last mile shrinking rapidly and expects job categories to dissolve as builder tools become the default interface for creating software.
The internal adoption curve at Anthropic suggests the transition is already underway at frontier labs. Engineers who would traditionally resist terminal-based tools adopted Claude Code anyway, indicating the productivity gain outweighed the friction. Cherny's own role exemplifies the shift — he manages a product that automates his prior job, and he has not touched a traditional codebase in over half a year.
Cherny's timeline — software engineer as a distinct title fading by end of 2026 — is aggressive compared to most labor-disruption forecasts. The prediction hinges on agentic coding tools reaching the threshold where non-engineers can ship production-grade software without writing code manually. If that threshold holds, the multiplication effect Cherny describes becomes plausible: far more people building software, far fewer doing it the old way.
The risk is that Cherny's forecast underweights the coordination overhead, edge-case debugging, and long-tail integration work that still defines much of software engineering today. Agents handle greenfield prototyping and scripting tasks well; production systems with legacy constraints, compliance requirements, and cross-team dependencies remain harder. The question is whether those constraints shrink as models improve or whether they define the new last mile that resists automation, preserving some version of the engineering role even as its scope narrows.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
