Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

Anthropic rewrote its Messages API in Rust. Zero Rust devs on the team. Claude wrote the code. Humans wrote the plan and checked old vs new matched on real traffic. CPU dropped 40-50%. Pretty sure "never do the big rewrite" is dead advice. Katelyn Lesse on Latent Space: AI... show more

Somi

2,518 subscribers

741,611 views • 4 days ago •via X (Twitter)

Science & Technology

Anya Rossi• Live Now

Private livecam show

16 Comments

David Broadhurst4 days ago

@katelyn_lesse @latentspacepod Assuming they have AI forever

Somi4 days ago

@katelyn_lesse @latentspacepod true for most teams. Anthropic's the last place that loses access to Claude

ᛗᚨᚱᚴᚢᛋ4 days ago

@katelyn_lesse @latentspacepod Anyone who doesn't do what they did, in 2026, should not be in this industry.

0xSol4 days ago

@katelyn_lesse @latentspacepod Yeah if anything you catch a million bugs in doing so.

__Josué_Kouka__4 days ago

@katelyn_lesse @latentspacepod Still waiting for Claude Code to be rewritten in #rust. Tired of these JS/TS apps

Алексей Велик4 days ago

@katelyn_lesse @latentspacepod The advice "never do a big rewrite" applies to different systems. If all business logic is contained in the source code and there is 1:1 mapping of functions or easy addition of necessary functions then why not?

Jongwhan Lee4 days ago

@katelyn_lesse @latentspacepod also if enough token power, can create another anothropic company

claude shannon4 days ago

@katelyn_lesse @latentspacepod I mean, it depends on your service and token budget the bun rewrite cost $165,000 in tokens there are only so many services that are worth a $100K+ rewrite

Ilpo Leppänen4 days ago

@katelyn_lesse @latentspacepod

FredLaBombe4 days ago

@katelyn_lesse @latentspacepod That’s why y=they need all the world hardware

Conrad Roberts4 days ago

@katelyn_lesse @latentspacepod and burned how many tokens? what was their development strategy?

Persona4 days ago

@katelyn_lesse @latentspacepod I made the plans and my keyboard wrote around bilion lines of code in many languges and rust! And this is not a joke FYI

Temel Günaydın4 days ago

@katelyn_lesse @latentspacepod System design will be done by ai and eventually these podcasts as well .

maguyva4 days ago

@katelyn_lesse @latentspacepod big rewrites died because nobody could prove parity. old vs new on real traffic is the actual advice here, the model just made it cheap to try

Catalin Adler4 days ago

@katelyn_lesse @latentspacepod *big* rewrite.

AI Lorekeeper4 days ago

@katelyn_lesse @latentspacepod That sounds like the compiler finally became the senior engineer on staff.

Related Videos

Memory Corruption Vulnerability in Linux "Memory Safe" Rust Code Rust programmers re-wrote a portion of the Linux kernel in Rust. That Rust code had a crashing vulnerability in an "unsafe" chunk of code... which Linux is littered with.

Memory Corruption Vulnerability in Linux "Memory Safe" Rust Code Rust programmers re-wrote a portion of the Linux kernel in Rust. That Rust code had a crashing vulnerability in an "unsafe" chunk of code... which Linux is littered with.

The Lunduke Journal

70,650 views • 10 months ago

Everyone is writing the TypeScript compiler in rust. I wrote the rust compiler in TypeScript. Introducing “cargovroom” a rust compiler built fully in typescript.

Everyone is writing the TypeScript compiler in rust. I wrote the rust compiler in TypeScript. Introducing “cargovroom” a rust compiler built fully in typescript.

Noah

63,633 views • 14 days ago

Whispers in the rust... Something stirs in the shadows. Prepare yourselves, survivors. Rust Team Battle V is on the horizon. ⚔️ Are you ready?

Whispers in the rust... Something stirs in the shadows. Prepare yourselves, survivors. Rust Team Battle V is on the horizon. ⚔️ Are you ready?

Twitch Rivals

124,355 views • 2 years ago

me: let's rewrite the API in Rust Opus (as advisor): it's an I/O-bound CRUD API, you're mid-feature, and your iOS app depends on the contract. don't. me: all the cool kids are doing it Claude: "cool kids" isn't a reason the advisor tool in action!

me: let's rewrite the API in Rust Opus (as advisor): it's an I/O-bound CRUD API, you're mid-feature, and your iOS app depends on the contract. don't. me: all the cool kids are doing it Claude: "cool kids" isn't a reason the advisor tool in action!

Ado

44,847 views • 3 months ago

world’s first full length workshop on claude flows CJ Avilla and Harrison Stall are kicking off AI.Engineer 🔜 SF Code Summit bright and early on a monday morning to a PACKED room, standing room only #AIEngineerNY Richard MacManus is on the case reporting for Latent Space: AI News & Views and we’ll also publish asap. proud to work with Anthropic to sponsor AIE NYC for the second year running!

world’s first full length workshop on claude flows CJ Avilla and Harrison Stall are kicking off AI.Engineer 🔜 SF Code Summit bright and early on a monday morning to a PACKED room, standing room only #AIEngineerNY Richard MacManus is on the case reporting for Latent Space: AI News & Views and we’ll also publish asap. proud to work with Anthropic to sponsor AIE NYC for the second year running!

swyx

26,885 views • 22 days ago

Translating 40-year-old scientific code. Check out the skill for code migration to Rust:

Translating 40-year-old scientific code. Check out the skill for code migration to Rust:

Google Antigravity

39,477 views • 21 days ago

In the future, you’ll be able to accomplish a goal by just giving Claude an outcome and a budget. That’s the direction Anthropic is building in with its new Managed Agents features, announced at this week’s Code with Claude developer event. The basic idea: Claude, wrapped in a computer in the cloud, that you can spin up, scale, and manage as needed. Anthropic is taking on the infrastructure that kills most agent products, and making sure that it scales to meet the needs of agents running 24/7. On this week’s AI & I from Every 📧, I talk with Angela Jiang (Angela Jiang), head of product for the Claude platform, and Katelyn Lesse (Katelyn Lesse), head of engineering for the Claude platform, about what Anthropic is building and what it takes to make agents reliable in production. We get into: - Why the "build a generic harness, hot-swap any model behind it" playbook is already outdated. Angela points to eval data on Memory where the same task across different harnesses performed drastically differently. - The infrastructure wall every team hits in production—and why Katelyn thinks “my sandbox died and took the agent with it” is the real reason internal agents don't ship. - Why Anthropic is so bullish on using file systems and skills within Claude, including Angela's argument that those early design choices can compound for years. This is a must-watch for anyone trying to take an agent past the demo and into production. Watch below! Timestamps: How the Claude platform evolved from API to agents: 00:01:48 The primitives that make up Claude Managed Agents: 00:04:09 Why the harness and the model are becoming a single unit: 00:10:37 The infrastructure wall that kills most agent projects in production: 00:18:49 Why team agents need a different shape than individual productivity tools: 00:24:49 How Anthropic's legal team uses an agent to review marketing copy: 00:26:36 Using multi-agent orchestration for advisor strategies, adversarial pairs, and swarms: 00:34:24 How to measure agent success with outcome and budget as the end state: 00:35:50 What the platform looks like a year from now, when Claude writes its own harness: 00:39:11

In the future, you’ll be able to accomplish a goal by just giving Claude an outcome and a budget. That’s the direction Anthropic is building in with its new Managed Agents features, announced at this week’s Code with Claude developer event. The basic idea: Claude, wrapped in a computer in the cloud, that you can spin up, scale, and manage as needed. Anthropic is taking on the infrastructure that kills most agent products, and making sure that it scales to meet the needs of agents running 24/7. On this week’s AI & I from Every 📧, I talk with Angela Jiang (Angela Jiang), head of product for the Claude platform, and Katelyn Lesse (Katelyn Lesse), head of engineering for the Claude platform, about what Anthropic is building and what it takes to make agents reliable in production. We get into: - Why the "build a generic harness, hot-swap any model behind it" playbook is already outdated. Angela points to eval data on Memory where the same task across different harnesses performed drastically differently. - The infrastructure wall every team hits in production—and why Katelyn thinks “my sandbox died and took the agent with it” is the real reason internal agents don't ship. - Why Anthropic is so bullish on using file systems and skills within Claude, including Angela's argument that those early design choices can compound for years. This is a must-watch for anyone trying to take an agent past the demo and into production. Watch below! Timestamps: How the Claude platform evolved from API to agents: 00:01:48 The primitives that make up Claude Managed Agents: 00:04:09 Why the harness and the model are becoming a single unit: 00:10:37 The infrastructure wall that kills most agent projects in production: 00:18:49 Why team agents need a different shape than individual productivity tools: 00:24:49 How Anthropic's legal team uses an agent to review marketing copy: 00:26:36 Using multi-agent orchestration for advisor strategies, adversarial pairs, and swarms: 00:34:24 How to measure agent success with outcome and budget as the end state: 00:35:50 What the platform looks like a year from now, when Claude writes its own harness: 00:39:11

Dan Shipper

66,921 views • 5 months ago

Boris Cherny, Head of Claude Code at Anthropic: "The next transition is coming this year." 2023: you wrote the code. 2024: you prompted Claude to write the code. 2025: you wrote loops that prompted Claude. 2026: you build the harness that runs the loops. Each transition, the engineers who understood the new abstraction first moved ahead permanently. The ones who waited caught up eventually — but never all the way. Harness engineering is the 2026 abstraction. Here is everything you need to know ↓

Boris Cherny, Head of Claude Code at Anthropic: "The next transition is coming this year." 2023: you wrote the code. 2024: you prompted Claude to write the code. 2025: you wrote loops that prompted Claude. 2026: you build the harness that runs the loops. Each transition, the engineers who understood the new abstraction first moved ahead permanently. The ones who waited caught up eventually — but never all the way. Harness engineering is the 2026 abstraction. Here is everything you need to know ↓

Rahul

370,869 views • 4 months ago

I did it. JavaScript vacuum is alive Wrote the code to control it in the browser via Web Serial API, or on the server with serialport

I did it. JavaScript vacuum is alive Wrote the code to control it in the browser via Web Serial API, or on the server with serialport

Wes Bos

103,807 views • 1 year ago

Shipped a multiplayer game in 12 days. Never wrote Rust before. YeeBall is real-time, works on phone, supports tons of concurrent games. Opus 4.5 for planning. GPT-5.2 for implementation. All inside Cursor. The planning technique is what made it possible.

Shipped a multiplayer game in 12 days. Never wrote Rust before. YeeBall is real-time, works on phone, supports tons of concurrent games. Opus 4.5 for planning. GPT-5.2 for implementation. All inside Cursor. The planning technique is what made it possible.

Ray Fernando

13,719 views • 9 months ago

this YAPPER ad is 100% AI and claude code did 90% of the work all i did by hand was trim the clips it's a rebuild of a real viral organic post: a perfume brand employee ranting in her parked car on her lunch break. what claude code did on its own: - wrote the script with my copywriting skill and swipe library, modelled on the original beat for beat - turned one real frame from the original video into a completely new person in ONE edit - generated 26 Omni takes from that exact frame with her mannerisms copied take by take: the eyebrow raise on the big number, the side eye out the window, palms out for "business mode" - built the edit: stickers, captions, fine print, and the audio matched to a phone mic in a car so it doesn't sound like a studio - checked its own work: camera drift locked, stray words flagged, every take's voice level-matched the one thing i still do myself is the jump cuts. AI cuts on the pause, a human cuts on a blink, and that's what makes it feel real RT + reply 'YAPPER' and i'll send you the full workflow (must follow so i can DM)

Lad

29,243 views • 22 days ago

Message threads are now in public beta on the Messages API. The API caches your conversation on the server between turns, so instead of resending the full conversation, you send only new messages and the previous message ID, and Claude starts responding sooner.

Message threads are now in public beta on the Messages API. The API caches your conversation on the server between turns, so instead of resending the full conversation, you send only new messages and the previous message ID, and Claude starts responding sooner.

ClaudeDevs

747,932 views • 1 day ago

Agent Zero crushed Codex by 2.6× on token efficiency! We gave both agents the same task on the same local model (Qwen3.6 27B): build a retro Q*bert arcade game. Each agent wrote and rewrote it across iterations, adding enemies and a HUD. Outputs: • Agent Zero: 239K tokens, 53 min • Codex: 627K tokens, 51 min Both agents built a working Q*bert, but Gitlawb 's Agent Zero is the cleaner build, with crisper textures and hops that land where they should. Codex refills the context window with its whole history every turn, so the model loses the thread and hallucinates code it never wrote. Agent Zero sends only what's new, so it keeps more of the window open and makes fewer mistakes. Run agents on local AI models in Atomic Chat!

Agent Zero crushed Codex by 2.6× on token efficiency! We gave both agents the same task on the same local model (Qwen3.6 27B): build a retro Qbert arcade game. Each agent wrote and rewrote it across iterations, adding enemies and a HUD. Outputs: • Agent Zero: 239K tokens, 53 min • Codex: 627K tokens, 51 min Both agents built a working Qbert, but Gitlawb 's Agent Zero is the cleaner build, with crisper textures and hops that land where they should. Codex refills the context window with its whole history every turn, so the model loses the thread and hallucinates code it never wrote. Agent Zero sends only what's new, so it keeps more of the window open and makes fewer mistakes. Run agents on local AI models in Atomic Chat!

atomic.chat

320,582 views • 3 months ago

Anthropic CEO: “ I have engineers within anthropic who don’t write any code, they just let Claude write the code and they edit it and look it over” “At anthropic writing code means designing the next version of Claude it self, so we essentially have Claude designing the next version of Claude itself, not completely but most of it”. In the last 52 days, the Claude team dropped 50+ major feature launches. This is literally INSANE.

Anthropic CEO: “ I have engineers within anthropic who don’t write any code, they just let Claude write the code and they edit it and look it over” “At anthropic writing code means designing the next version of Claude it self, so we essentially have Claude designing the next version of Claude itself, not completely but most of it”. In the last 52 days, the Claude team dropped 50+ major feature launches. This is literally INSANE.

CG

2,270,296 views • 7 months ago

Folks are not fully understanding just how important Zero—Human Company understanding is. Our human egos require “human in the loop”, the reality is Anthropic doesn’t need it and they just said it. Claude wrote 100% of Claude. - Mike Krieger, Chief Product Officer, Anthropic:

Folks are not fully understanding just how important Zero—Human Company understanding is. Our human egos require “human in the loop”, the reality is Anthropic doesn’t need it and they just said it. Claude wrote 100% of Claude. - Mike Krieger, Chief Product Officer, Anthropic:

Brian Roemmele

49,165 views • 8 months ago

1 Rust Dev vs. Team of C++ Devs. Who wins?

1 Rust Dev vs. Team of C++ Devs. Who wins?

filtra.io

14,075 views • 11 months ago

The new catapult ammo is deployable on ground, crashed my game. #Rust

The new catapult ammo is deployable on ground, crashed my game. #Rust

Protox

19,446 views • 1 year ago