Feb 19, 2026 · 1h 27m · lennys-podcast
Head of Claude Code: What happens after coding is solved | Boris Cherny
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Anthropic's Head of Claude Code, Boris Cherny, explains how automated code generation and autonomous agentic workflows are eliminating manual syntax writing and transforming traditional software engineers into interdisciplinary builders. He shares key AI-native organizational strategies, tactical tips for multi-agent orchestration, and Anthropic's multi-tiered safety architecture.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Lenny holds 23.9% of the talking time here. How this is scored →
Lenny as informed peer 4.0 Guest teaching 4.1 Guest disagreement 1.1 Lenny pushing back 1.2
speaking balance: gold is Lenny, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 1:05:35 Dismissing complex multi-step scaffolding
Boris firmly dismisses the common developer impulse to box models into rigid multi-step pipelines, asserting that bespoke scaffolding is wiped out by next-generation models.
Hardest push from Lenny ▶ 3:55 Pressing on Cursor departure and swift returnLenny directly confronts Boris with a 'spicy question' demanding an explanation for why he abruptly left Anthropic for competitor Cursor before returning two weeks later.
Biggest teaching moment ▶ 57:25 Explaining mechanistic interpretability and superpositionBoris delivers an in-depth breakdown of how safety teams directly inspect internal model cognition and neuron superposition rather than relying solely on black-box evaluation.
Lenny holds their own ▶ 44:45 Presenting proprietary audience survey on AI sentimentLenny counters the purely optimistic narrative by introducing empirical poll data showing designers report significantly lower enthusiasm for AI tooling compared to engineers and PMs.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Lenny as informed peer | Guest teaching | Guest disagreement | Lenny pushing back | Why |
|---|---|---|---|---|---|---|
| The Cursor Detour and Returning to Anthropic's Mission | 4 | 3 | 1 | 2 | Lenny opens with an inquisitive and mildly provocative question regarding Boris's brief departure to Cursor before returning to Anthropic. Boris transparently explains his core motivation around safety, while Lenny cites industry stats on Claude Code adoption. | |
| Prototyping Claude Code and Terminal Interface Origins | 3 | 5 | 1 | 1 | Boris narrates the origin story of Claude CLI, sharing how unexpected model capabilities with the bash tool prompted him to build a terminal-based prototype. Lenny listens attentively as Boris details product discovery and initial internal traction. | |
| Thinking in Exponentials and Fostering Innovation | 4 | 4 | 1 | 1 | Lenny highlights how rapidly industry consensus shifted on AI-written code, referencing Peter Thiel/OpenAI comparisons. Boris explains Anthropic's culture of exponential extrapolation from scaling laws. | |
| The Reality of 100% Automated Code Generation | 3 | 5 | 1 | 2 | Lenny expresses surprise that Boris writes zero manual code while shipping dozens of PRs daily as a team lead. Boris details his workflow and explains that software engineering has shifted toward proactive task discovery. | |
| Massive Productivity Multipliers and Overcoming Legacy Habits | 4 | 5 | 1 | 1 | Boris provides concrete comparisons to his previous developer productivity work at Meta, noting 200% PR gains compared to legacy single-digit improvements. He illustrates how legacy engineering habits can hinder optimal agent utilization during debugging. | |
| Organizational Principles: Underfunding and Token Abundance | 5 | 4 | 1 | 2 | Lenny probes Boris's counterintuitive management philosophy of deliberate underfunding. Boris explains that token abundance coupled with lean headcount forces teams to automate aggressively rather than premature cost cutting. | |
| The Craft of Programming: Utility vs. Pure Artistry | 4 | 4 | 2 | 2 | Lenny asks whether Boris mourns the lost craft of manual coding. Boris reflects on his pragmatic view of software as utility, while acknowledging the enduring artistic appeal of low-level programming for some engineers. | |
| The Printing Press Analogy and Democratized Creation | 4 | 5 | 1 | 1 | Boris draws a detailed historical parallel between the Gutenberg printing press and AI software creation, predicting that manual syntax familiarity will soon be obsolete. Lenny reinforces the insight with his own experience as a former engineer. | |
| The Expansion of Agentic AI Across Tech and Society | 4 | 4 | 1 | 2 | Lenny raises the economic and societal implications of AI on knowledge work, invoking Jevons paradox. Boris clarifies the precise technical definition of an agent and notes the profound shift from conversational chat to autonomous tool execution. | |
| The Rise of the Curious Generalist Builder | 4 | 4 | 2 | 1 | Boris outlines how rigid specialized roles like PM, designer, and engineer are collapsing into the versatile 'builder'. Lenny observes the changing organizational landscape and seeks practical advice for future workers. | |
| Sponsor Message: MetaView | 5 | 2 | 0 | 1 | Lenny shares proprietary survey data from his social channels comparing job satisfaction across PMs, engineers, and designers after adopting AI. Boris engages with the findings and describes internal designer workflows. | |
| The Power of Latent Demand in Product Design | 4 | 5 | 1 | 1 | Boris explains the product concept of latent demand, citing Facebook Marketplace and non-engineers repurposing Claude Code's terminal for data science and bio-analysis. He connects this to designing AI systems aligned with model distribution. | |
| Rapid Iteration on Cowork and Three-Tiered Safety | 4 | 5 | 1 | 1 | Boris describes building Cowork within 10 days using Claude Code and breaks down Anthropic's three-tier safety framework: mechanistic interpretability, evals, and real-world deployment. Lenny listens to the rapid deployment strategy. | |
| Mechanistic Interpretability and Open Safety Standards | 4 | 5 | 1 | 1 | Boris delves into mechanistic interpretability, neural superposition, and Anthropic's open-source sandbox approach. Lenny actively validates the concepts and notes his interest in hosting deeper safety discussions. | |
| Continuous Agent Orchestration and Shared Roots | 5 | 3 | 1 | 1 | Lenny and Boris discover they were both born in Odesa, Ukraine, sharing personal reflections on migration and programmer heritage. Boris discusses running multiple asynchronous agents continuously. | |
| Principles for AI Builders: The Bitter Lesson and Future Models | 4 | 5 | 2 | 1 | Boris synthesizes architectural principles from Rich Sutton's 'The Bitter Lesson,' cautioning against brittle prompt scaffolding and advising founders to build for models six months in the future. Lenny asks clarifying questions on what capabilities to assume. | |
| Pro Tips for Mastering Claude Code | 3 | 5 | 1 | 1 | Boris gives tactical user recommendations for Claude Code, detailing how 'plan mode' operates via minimal prompt steering and why selecting maximum effort on frontier models reduces total token spend. | |
| Market Competition, Post-AGI Aspirations, and Miso Making | 4 | 4 | 1 | 1 | Lenny inquires about competitive dynamics with OpenAI Codex and brings up a prompt from co-founder Ben Mann regarding Boris's post-AGI lifestyle. Boris details his artisanal miso making and long-term mindset. | |
| Anthropic's Long-Term Vision and User Feedback Loops | 4 | 4 | 1 | 1 | Boris underscores Anthropic's long-term vision connecting coding, tool use, and computer use. Lenny cites reported revenue figures and the massive scale achieved within a single year. | |
| Lightning Round: Essential Reading and Sci-Fi Inspirations | 4 | 3 | 1 | 1 | In the lightning round, Boris recommends seminal technical and sci-fi books including Functional Programming in Scala, Accelerando, and Wandering Earth. Lenny matches Boris with recommendations of his own. | |
| Lightning Round: Favorite Products, Podcasts, and Custom Evals | 5 | 3 | 1 | 1 | Boris explains his workflow with Cowork and reveals that Anthropic utilized Lenny's published newsletter posts as an internal evaluation benchmark. Lenny reacts playfully to becoming an AI evaluation metric. |
Statements from this episode (46)
Anthropic: Claude Code's share in private repos exceeds four percent
“So we actually think if you look at private repositories, it's quite a bit higher than that.”
Anthropic's AGI roadmap: First coding, then tool use, then computer use
“We were building the models in this way that kind of fit our mental model of the way that we built SafeHEI, where the model starts by being really good at coding, then it gets really good at tool use, then it gets really good at computer use. Roughly, this is …”
Cherny: Early-stage product initiatives should be intentionally under-resourced
“And for me, this is actually a pretty important product lesson, right? It's like you want to under-resource things a little bit at the start.”
Cherny: Claude Code took months to achieve widespread adoption after launch
“Actually, something that people don't really remember is Cloud Code was not initially a hit. When we released it got a bunch of users. There was a lot of early adopters that got it immediately, but it actually took many months for everyone to really understand…”
Cherny: IDEs might not be needed for software engineering by end-of-2025
“And my prediction back in May of 2025 was, by the end of the year, you might not need an IDE to code anymore, and we're gonna start to see engineers not doing this.”
Cherny: Claude Code went from writing 20% to 100% of his code
“Because even in February when we released it was writing maybe, I don't know, like, 20% of my code, not more. And even in May, it was writing maybe 30%. I was still using, you know, Kurtzer for most of my code. And it only crossed a hundred percent in November…”
Boris Cherny ships 10 to 30 pull requests every day
“Every, you know, every day I ship like 10, 20, 30 pull requests, something like that.”
Cherny has not edited a single line of code since November
“I have not edited a single line by hand since November.”
Cherny: Developers cannot yet be totally hands-off with AI-generated code
“I don't think we're kind of at the point where you can be totally hands off, especially when there's a lot of people, you know, like running the program. You have to make sure that it's correct. You have to make sure it's safe and so on.”
Cherny: Claude automatically reviews 100% of pull requests at Anthropic
“Here at Anthropic, Claude reviews a hundred percent of pull requests.”
Cherny: Anthropic's Cowork agent handles all project management for his team
“All of my project management for the team co-work does all of it. It's like syncing stuff between spreadsheets and messaging people on Slack and email and all this kind of stuff.”
Cherny predicts coding will be increasingly solved across all stacks within months
“And over the next few months, I think what we're going to see is just across the industry, it's going to become increasingly solved, you know, for every kind of code base, every tech stack that people work on.”
Cherny: Claude Code boosted Anthropic engineer PR productivity by 200%
“Over the last year, like, since we introduced quad code, we probably, I don't know the exact number, we probably, like, forex the engineering team or something like this, but productivity per engineer has increased 200% in terms of, like, pull requests.”
Cherny: Traditional developer productivity efforts yield only minor percentage point gains
“Things that we saw is, you know, in a year with hundreds of engineers working on it, you would see a gain of like a few percentage points of productivity, something like this.”
Cherny: Underfunding engineering projects forces developers to automate workflows with AI
“There's this there's interesting thing that happens also when you when you underfund everything a little bit because then people are kind of forced to clodify.”
Cherny: Give developers unlimited tokens initially; optimize costs only after scaling
“My advice generally is don't try to optimize. Don't try to cost cut at the beginning. Start by just giving engineers as many tokens as possible.”
Cherny: Some Anthropic engineers spend hundreds of thousands monthly on AI tokens
“You know, at Anthropic, we're starting to see some engineers that are spending, you know, like hundreds of thousands a month in, in tokens.”
Cherny: AI models are entirely capable of writing straight to binary
“I mean, it totally can do that if you wanted to.”
Cherny: Understanding underlying code layers won't matter in one to two years
“My take is, I think for people that are using that are using quad code, that are using agents to code today, you still have to understand the layer under. But yeah, in a year or two, it's not gonna matter.”
Cherny: AI automation will soon expand to product management and design
“I think it's going to be a lot of the roles that are adjacent to engineering. So yeah, it could be like product managers. It could be design, could be data science. It is going to expand to pretty much any kind of work that you can do on a computer because the…”
Cherny: Everyone will program within a few years, causing a disruptive transition
“And so I imagine a world, you know, a few years in the future where everyone is able to program. And what does that unlock? Anyone can just build software anytime. And I have no idea. It's just the same way that, you know, in the 1400, no one could have predic…”
Cherny: Everyone on Anthropic's Claude Code team writes code, including finance
“So on the quad code team, everyone codes, you know, our product manager codes, our engineering manager codes, our designer codes, our finance guy codes, our data scientist codes, like everyone on the team codes.”
Cherny: Curious, interdisciplinary generalists will be rewarded most in tech
“So I think a lot of the people that will be rewarded the most over the next few years, they won't just be AI native and they don't just know how to use these tools really well, but also they're curious and they're generalists. And they cross over multiple disc…”
Cherny: The 'software engineer' title will start disappearing by year-end
“I do think that there is a future where I think by the end of the year, what we're going to start to see is these start to get even murkier, murkier, where I think in some places the title software engineer is going to start to go away. And it's just gonna be …”
Rachitsky: 70% of engineers and PMs report enjoying work more with AI
“Both engineers and PMs, 70% of people said they are enjoying their job more. And about 10% said they're enjoying their job less. Designers, interestingly, Only 55% said they're enjoying their job more, and 20% said they're enjoying their job less.”
Cherny: Simplifying existing workflows is the most important product principle
“Whatever people are doing, if you can make that a little bit easier, then that's just going to be a much better product that people enjoy more. And this is just this principle of latent demand, which I think is just the single most important principle and prod…”
Cherny: How users hack or misuse your product reveals latent demand
“Latent demand is this idea that if you build a product in a way that can be hacked or can be kind of misused by people in a way it wasn't really designed for to do kind of something that they want to do, then this helps you as the product builder learn where t…”
Cherny: Minimize scaffolding; let AI models do what they naturally attempt
“The modern framing that I've been seeing in the last six months is a little bit different, and it's look at what the model is trying to do and make that a little bit easier.”
Cherny: Anthropic built its Cowork AI agent in 10 days using Claude
“And so over 10 days, they just completely used quad code to build it. And, you know, co-workers actually, there's this very sophisticated security system that's built in, and essentially these guardrails to make sure that the model kind of does the right thing…”
Cherny: Anthropic can trace specific neuron activations related to AI deception
“We at this point have like pretty sophisticated technology to understand what's happening in the neurons to trace it. And so for example, like if there's a neuron related to deception, we can start, we're starting to get to the point where we can monitor it an…”
Cherny: Strong evidence shows LLMs perform reasoning beyond next-token prediction
“You know, like a long time ago, we weren't sure if the model was just predicting the next token or is doing something a little bit deeper. Now I think there's actually quite strong evidence that it is doing something a little bit deeper.”
Cherny: Anthropic released an open-source execution sandbox for any AI agent
“And so for Cloud Code, for example, we released an open source sandbox, and this is a sandbox they can run the agent in, and it just makes sure that there's certain boundaries and it can't access like everything on your system. And we made that open source, an…”
Cherny continuously runs around five concurrent AI agents in his workflow
“I always have a bunch of agents running. So like at the moment I have like five agents running and at any moment, like, you know, like I wake up and I start a bunch of agents.”
Cherny: One-third of his coding is now done on an iOS app
“Maybe a third of my code now is in the terminal, but also a third is using the desktop app. And then a third is the iOS app, which is just so surprising because I did not think that this would be the way that I code in, even in twenty-twenty-six.”
Cherny: Next-generation foundation models often wipe out custom AI scaffolding gains
“And in general, what we see is maybe scaffolding can improve performance, maybe 10, 20%, something like this. But often these gains just get wiped out with the next model. So it's almost better to just wait for the next one.”
Cherny: Claude Code user growth went exponential following the Opus 4 release
“And Opus four was our first kind of ASL three class model that we released back in May. And we just saw this inflection because everyone started to use quad code for the first time. And that, that was kind of when our growth really went exponential. And like I…”
Cherny: Autonomous AI models will soon run unattended for long periods
“And so I think over time, this is going to become more and more normal where the models are running for a very, very long period of time and you don't have to sit there and babysit them anymore.”
Cherny: Frontier models are often cheaper overall because they require fewer corrections
“The thing that happens is sometimes people try to use a less expensive model like Sonnet or something like this, but because it's less intelligent, it actually takes more tokens in the end to do the same task. And so it's actually not obvious that it's cheaper…”
Cherny: Claude Code's Plan Mode just injects a single-sentence negative prompt
“All it is, is we inject one sentence into the model's prompt to say, please don't write any code yet. That's it. Like there's actually like nothing fancy going on. It's just the simplest thing.”
Cherny: Opus 4.6 successfully completes code changes on the first attempt post-planning
“Once the plan looks good, then you let the model execute. I auto accept edits after that, because if the plan looks good, it's just going to one shot it. It'll get it right the first time, almost every time with the Opus 4.6.”
Cherny: Anthropic's Claude Code team does not spend time testing competing products
“Honestly, for our team though, we're just focused on solving the problems that users have. So for us, you know, we don't spend a lot of time looking at competing products. We don't really try the other products.”
Cherny: Claude Code has become a multi-billion dollar business
“Everything that's happening right now around, you know, just like quad code becoming this huge, you know, multi-billion dollar business.”
Cherny: Global AI and Claude Code adoption is still around one percent
“Most of the world still does not use quad code. Most of the world still does not use AI. So it just feels like this is one percent on, and there's so much more to go.”
Cherny: Anthropic's Cowork agent is growing much faster than Claude Code did
“But like I said, it's just growing much faster than Quadco did in the early days. So I think it's starting to break through a bit.”
Anthropic evaluated Cowork against Lenny Rachitsky's 50 tasks, shipping at 48 passes
“So we actually, one of our PMs used that as a way to evaluate Cowork before we released it. And I think at the point where we hit where Cowork was able to do like, 48 out of the 50, they were like, okay, it's pretty good.”