Aug 7, 2025 · 1h 0m · mad
Anthropic's Surprise Hit: How Claude Code Became an AI Coding Powerhouse
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
In this episode of The MAD Podcast, host Matt Turck interviews Boris Cherny, creator of Claude Code at Anthropic, exploring how a simple CLI prototype evolved into a viral, multi-hundred-million-dollar AI coding tool. They discuss the historical evolution of programming paradigms, agentic architecture, memory management, human-in-the-loop safety, and strategic guidance for building software on frontier AI models.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Matt holds 26.1% of the talking time here. How this is scored →
Matt as informed peer 2.8 Guest teaching 2.2 Guest disagreement 0.0 Matt pushing back 0.2
speaking balance: gold is Matt, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 45:23 Boris's frank self-critique of product limitations
In an exceptionally friendly interview, Boris shows constructive pushback against over-hyping AI by rating Claude Code's code-writing ability as only a 6 out of 10 compared to research.
Hardest push from Matt ▶ 9:15 Matt reframing CLI interaction for non-technical listenersMatt politely intervenes to reframe Boris's technical explanation of CLI tools into an accessible text messaging versus GUI app analogy.
Biggest teaching moment ▶ 29:35 Boris explaining why Claude Code avoids standard RAGBoris educates Matt on why traditional RAG indexing creates security and maintenance issues, explaining how Claude Code instead uses agentic command-line searches.
Matt holds his own ▶ 13:35 Matt citing specific market share figuresMatt demonstrates sharp industry domain knowledge by citing exact market share metrics comparing Anthropic's leading share to OpenAI in code generation.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Matt as informed peer | Guest teaching | Guest disagreement | Matt pushing back | Why |
|---|---|---|---|---|---|---|
| Podcast Opening and Highlights Preview | 1 | 0 | 0 | 0 | Matt introduces the episode, citing reported figures of $400 million in ARR for Claude Code five months post-launch. The tone is welcoming and introductory with no friction. | |
| The Accidental Prototype and Origins of Claude Code | 1 | 1 | 0 | 0 | Boris shares the origin story of Claude Code starting as an accidental CLI tool for personal note-taking and music playback before gaining a bash tool. Matt asks standard background questions. | |
| Historical Evolution of Programming Paradigms | 2 | 3 | 0 | 0 | Boris traces the 70-year history of programming from punch cards to IDEs and agentic coding. Matt contributes by bringing up a story he heard about Boris's grandfather using punch cards in the USSR. | |
| Demystifying the Terminal vs. IDE Experience | 3 | 2 | 0 | 1 | Matt prompts Boris to demystify the terminal versus an IDE for non-technical listeners, offering a concise analogy comparing CLI interactions to texting and IDEs to traditional GUI apps. | |
| Product Overhang and Anthropic's Design Philosophy | 2 | 2 | 0 | 1 | Boris explains the concept of product overhang and Anthropic's philosophy of lightweight UI design. Matt gently presses on whether Claude Code will stay CLI-first or move toward full IDE integration over time. | |
| Claude's Technical Strengths and AGI Alignment | 4 | 1 | 0 | 0 | Matt demonstrates host expertise by citing specific market share statistics comparing Anthropic (over 40%) to OpenAI (21%) in code generation, asking why Claude excels in coding tasks. | |
| Defining Agentic AI: Tools, Actions, and Planning | 3 | 3 | 0 | 0 | Boris details the difference between simple tool use, fixed workflows, and autonomous agents. Matt synthesizes the definition cleanly back to the guest. | |
| System Actions, Safety Controls, and MCP Integration | 3 | 3 | 0 | 0 | Boris explains sub-agents, system actions, and Model Context Protocol integrations, invoking Rich Sutton's 'The Bitter Lesson' to discuss model generality versus specialization. | |
| Context Pollution, Agentic Search, and CLAUDE.md Memory | 4 | 4 | 0 | 0 | Matt raises technical concepts like context pollution and declarative memory. Boris educates on why Claude Code avoids standard RAG in favor of iterative agentic search using glob and grep alongside CLAUDE.md files. | |
| Human-in-the-Loop Autonomy and Control in Claude Code | 3 | 2 | 0 | 0 | Matt asks about human-in-the-loop controls and enterprise deployment in regulated environments. Boris explains how local execution and API-only connectivity satisfy strict security needs. | |
| Designing a Lightweight UI/UX for Terminal-Based Coding Agents | 2 | 2 | 0 | 0 | Matt asks about balancing capability with minimal terminal UX. Boris shares details about rediscovering terminal design, including tweaking loading spinners across dozens of iterations. | |
| Interactive Terminal Delights and Custom Task Status Phrases | 3 | 1 | 0 | 1 | Matt highlights delightful terminal easter eggs like status phrases ('herding', 'schlepping') and gently probes on pricing evolution and rate limits for power users. | |
| Core Developer Use Cases and Workflow Strategies Across the SDLC | 3 | 4 | 0 | 0 | Boris provides a candid breakdown of Claude Code's capabilities, rating codebase research at 10/10 while code writing and debugging are at 6/10, and explains multi-prompt planning strategies. | |
| The AI Coding Ecosystem and Building for Future Model Frontiers | 4 | 3 | 0 | 0 | Matt lists competitor platforms across the AI coding landscape. Boris delivers a strategic framework advising founders to build for model capabilities expected in 6 to 12 months rather than today's limits. | |
| Model Providers vs Application Builders in the Coding Landscape | 4 | 2 | 0 | 0 | Matt asks about the coopetition dynamics between foundation model providers and specialized application builders. Boris predicts most innovation will happen on top of API/SDK platforms. | |
| Essential Advice for Early-Career Software Developers | 2 | 2 | 0 | 0 | Matt asks for career advice for early-stage software developers. Boris emphasizes that modern engineers must master both foundational computer science and agentic AI tooling. |
Statements from this episode (26)
Turck: The Information reports Claude Code hit $400M ARR in six months
“There was an article in the information just a couple of days ago that was saying that Cloud Code, which really came out, what, at the end of February of this year, so five, six months ago, is already generating four hundred million in annualized revenue. That…”
Cherny: Most Anthropic employees became daily active users of Claude Code
“Pretty soon after we launched it, most of Anthropic was a daily active user”
Cherny: Anthropic builds minimal product interfaces to keep up with model evolution
“The way we think about it is the model is evolving so quickly that we build a minimal possible product to keep up with it.”
Cherny: Code is the primary path for AI to reach AGI
“Maybe coding is the way that we get to the next level of intelligence. If you call it like AGI or ASI or whatever, the model needs some way to interact with the world and for a model, the natural way is code.”
Anthropic released Claude Code to learn how to make AI agents safe in the wild
“One of the reasons that we release cloud code is just to learn how people use it, and to learn how to make this thing safe, to learn how it behaves in the wild, so that we know what to do next.”
Cherny: Anthropic data scientists, designers, and PMs all use Claude Code
“The data scientists at Anthropic all use quad code to write their queries, and designers use it to build small prototypes, and product managers use it to manage tasks.”
Cherny: Claude Code can do almost anything a human does on computers
“Cloud Code can do pretty much anything that a person can do on their computer.”
Cherny: Claude Code operates as both an MCP client and server
“Quad code is a MCP client and an MCP server.”
Cherny: General agentic AI systems will outperform specialized systems long term
“If you build a agentic system in the, in this context then the more general agentic system will generally outperform the more specific one in the longterm.”
Cherny: AI models won't need rigid sub-agent roles in 6-12 months
“But I think that the models six or 12 months from now, they probably won't need this anymore because they're all going to be pretty good and you won't have to define very rigidly what each one's responsibilities are anymore.”
Cherny: Claude Code does not use RAG for codebase memory
“And so quad code actually doesn't use this technique called rag. Instead, what it does is it just searches files the same way that a human would.”
Cherny: Anthropic only releases Claude Code features used daily internally
“Generally our bar is if we find ourselves really happy with it and we find ourselves using it every day, then we release it to everyone.”
Cherny: Engineering teams will become more horizontal due to AI memory tools
“And my feeling is that this is something where teams will kind of get more and more horizontal as a result of this because anyone is able to contribute in this way.”
Cherny: Claude Code requires no services beyond the API itself
“It doesn't use any services except the API itself. So that's all it needs. And then everything else you actually don't need. And this is one of the nice side effects of not doing code base indexing or anything like this is it's just very easy to hook up to.”
Cherny: Most code at Anthropic is now written using Claude Code
“You know, at this point, most code at Anthropic is written using quad code and almost everyone at Anthropic is using it every day.”
Cherny: Claude Code offers subscription pricing from $20 to $200 monthly
“For Quadco, there's two pricing models today. One is you can get a subscription. There's Pro, and then there's Max, and this is I think it's 20 bucks a month, hundred bucks a month, or 200 bucks a month, and it has very, very generous rate limits.”
Cherny rates Claude Code 10/10 for codebase research, 6/10 for writing
“I feel like if you look at answering questions about the code base and kind of code base research, I think it's like 10 out of 10 good. It's as good as it can get. When it comes to writing code, it's maybe like a six out of 10. It's pretty good. It won't get e…”
Cherny: Claude Code cut Anthropic engineering onboarding from weeks to days
“And what we saw is adanthropic technical onboarding used to take a few weeks, but now engineers are usually productive within the first few days.”
Cherny: I use both Claude Code and Cursor every day
“I personally use a lot of these products and you know, I use quad code every day, but I also use cursor every day and I use other products every day.”
Cherny: AI startups should build for six-month future model capabilities
“My advice to companies building is definitely build for what the model will be able to do six months from now, not for what the model can do today.”
Cherny: Claude Code hit its stride with Sonnet 4 and Opus 4
“We started building quad code when it was still Sonnet at 3.5 and it was okay. And then with, you know, 3.6 and 3.7, it was still, it was fine. It was pretty good. But then when Sonnet four and Opus four came out, that's when it really hit its stride and we fe…”
Cherny: More AI innovation will happen on platform APIs than in-house
“I think there's probably room for both. And my personal take is probably there's going to be more stuff built on top of the platform than there will be built in-house. Just because there's so many things to build and there's just not enough time and enough peo…”
Boris Cherny: Programming will shift from text manipulation to working with agents
“I think one way it will definitely play out is it's going to change programming where programming is no longer direct text manipulation, but it's more working with agents to get the work done.”
Cherny: Learning to code today is actually harder than in the past
“I think for people that are learning coding today, it's actually harder than it was when I was learning coding, because not only do you have to know coding, because you still have to understand the languages. You still have to understand the frameworks. You st…”