AI Thinkers·Head of Claude Code, Anthropic
Boris Cherny
Coding — translating intent into working software — is largely solved for experienced practitioners and will be solved for everyone within years. The bottleneck has permanently shifted from implementation to intent: what do you want to build, why does it matter, and does the result actually work? The value of the software engineer as a narrow implementation specialist is being replaced by the value of the "builder" — someone who can orchestrate AI execution, verify outputs, and think across product and engineering. This is a printing-press moment: just as the press made writing universal, AI coding tools will make building software universal.
Key positions
Coding is largely solved — for experienced engineers, AI authoring production code is current reality, not a future prediction.
"At this point, it is safe to say that coding is largely solved. At least for the kind of programming that I do, it's just a solved problem because Claude can do it." He had not written a line of code by hand since November 2025 while shipping 10–30 PRs daily, all authored by Claude Code. The verification load — understanding what was built, whether it is correct, whether it is safe — has become the primary engineering task.
Lenny's Podcast, February 19 2026 (lennysnewsletter.com/p/head-of-claude-code-what-happens)
The title "software engineer" will disappear by end of 2026, replaced by "builder."
"I think by the end of the year, everyone is going to be a product manager, and everyone codes. The title software engineer is going to start to go away." Engineering, design, and product management are converging. On his Claude Code team: "everyone codes. Our product manager codes, our engineering manager codes, our designer codes." He acknowledges the transition will be "very disruptive and painful for a lot of people" in the near term.
Lenny's Podcast, February 2026; Fortune interview, January 2026
Claude Code is the thinnest possible wrapper over the model — the model's capability is the product.
"All the secret sauce, it's all in the model. And this is the thinnest possible wrapper over the model." The CLI-first design was philosophical: Anthropic believed models were improving fast enough that heavy UI investment would quickly become obsolete. The terminal is also the natural surface for delegation — IDEs are optimised for suggestion, CLIs for command.
Pragmatic Engineer interview (newsletter.pragmaticengineer.com/p/building-claude-code-with-boris-cherny)
Context management is the highest-leverage engineering skill in an AI-first workflow.
Cherny runs 10–15 concurrent Claude Code sessions: "It's not so much about deep work, it's about how good I am at context switching." His "context firewall" technique uses subagents for log reading, code searching, and document searching in parallel, with the main agent only dispatching and converging. Context provision is equivalent to giving a skilled contractor a proper brief.
Lenny's Podcast, February 2026; Pragmatic Engineer interview, 2025
AI democratises building the way the printing press democratised writing — creating more, not less, demand at a societal level.
"I imagine a world where everyone is able to program. Anyone can just build software anytime." Scribes (roughly 1% of the population) technically lost their jobs when moveable type arrived, but 70% of the population became literate within 200 years. The bottleneck moves from "who can write the code" to "who has the ideas worth building and the judgment to verify the result."
Lenny's Podcast, February 2026
Using the most capable model is often cheaper than using a cheaper one.
"Because a less capable model is less intelligent, it requires more tokens to do the same task. It is not obvious that using a cheaper model is actually cheaper. Often, the most capable model is cheaper and less token-intensive because it completes the task faster with less correction." Underpowered tools create rework loops that cost more than the capability premium.
Boris Cherny X thread on Claude Code tips, March 2026 (x.com/bcherny/status/2017742741636321619)
In their own words
“At this point, it is safe to say that coding is largely solved. At least for the kind of programming that I do, it's just a solved problem because Claude can do it.”
“All the secret sauce, it's all in the model. And this is the thinnest possible wrapper over the model.”
“I think by the end of the year, everyone is going to be a product manager, and everyone codes. The title software engineer is going to start to go away.”
“I imagine a world where everyone is able to program. Anyone can just build software anytime.”
Tensions with other thinkers
Where this view genuinely conflicts with others in the field.
Karpathy advocates building foundational understanding (Zero to Hero); Cherny says fundamentals will stop mattering "in a year or two."
"Hire generalists not specialists" challenges the dominant SWE career advice of deep technical specialisation.