Loading video...

Video Failed to Load

Go Home

Knowing how LLM contexts work and how to work around context limitations – aka “context engineering” – is becoming so important. No better person to explain than dex Timestamps: 00:00 Intro 01:33 Dex’s path into tech 03:34 Early work in platform engineering 05:28 Replicated 11:24 Metalytics 12:36 12-factor agents...

GergelyOrosz's profile picture

Gergely Orosz

346,790 subscribers

63,174 views • 22 days ago •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

The highest leverage thing you have to optimize for AI agents is their context. If you’re not getting enough out of AI agents, you probably have a context problem. By default the AI model knows absolutely nothing relevant to perform the task successfully. The whole job of context engineering is to prioritize what the model has as information to work with. Small deviations in the goals, instructions, domain understanding, or underlying data you supply the model will lead to wildly different outcomes. This is where you should spend almost all your time and energy in a workflow. We will start to religiously document more of our workflows and business processes, create clearer goals for people and agents, and ensure our tech infra is setup to get AI agents the right information. The lessons are currently being learned the fastest with coding agents, but the same lessons will apply to every field of knowledge work going forward. dex has a great talk on this that’s absolutely worth your time to watch.
0:42

The highest leverage thing you have to optimize for AI agents is their context. If you’re not getting enough out of AI agents, you probably have a context problem. By default the AI model knows absolutely nothing relevant to perform the task successfully. The whole job of context engineering is to prioritize what the model has as information to work with. Small deviations in the goals, instructions, domain understanding, or underlying data you supply the model will lead to wildly different outcomes. This is where you should spend almost all your time and energy in a workflow. We will start to religiously document more of our workflows and business processes, create clearer goals for people and agents, and ensure our tech infra is setup to get AI agents the right information. The lessons are currently being learned the fastest with coding agents, but the same lessons will apply to every field of knowledge work going forward. dex has a great talk on this that’s absolutely worth your time to watch.

Aaron Levie

87,086 views • 11 months ago

What does it mean for software engineering when we no longer write the code? Here's the take from Boris Cherny (Boris Cherny), the creator of Claude Code. Timestamps: 00:00 Intro 11:15 Lessons from Meta 19:46 Joining Anthropic 23:08 The origins of Claude Code 32:55 Boris's Claude Code workflow 36:27 Parallel agents 40:25 Code reviews 47:18 Claude Code's architecture 52:38 Permissions and sandboxing 55:05 Engineering culture at Anthropic 1:05:15 Claude Cowork 1:12:48 Observability and privacy 1:14:45 Agent swarms 1:21:16 LLMs and the printing press analogy 1:30:16 Standout engineer archetypes 1:32:12 What skills still matter for engineers 1:35:24 Book recommendations Brought to you by: • Statsig — ⁠ The unified platform for flags, analytics, experiments, and more. • Sonar – The makers of SonarQube, the industry standard for automated code review. Proactively find and fix issues in real-time with the SonarQube MCP Server: • WorkOS – Everything you need to make your app enterprise ready. Three interesting things from this conversation: 1. Boris automated himself out of code review well before AI. Boris was one of the most prolific code reviewers at Meta company. And he worked hard to minimize time spent on code review. His system::every time he left the same kind of review comment, he logged it in a spreadsheet. Once a pattern hit 3-4 occurrences, he’d write a lint rule to automate it away! 2. PRDs are dead on the Claude Code team: prototypes replaced them. Instead of writing Product Requirement Documents (specs), they build hundreds of working prototypes before shipping a feature. Boris: “There’s just no way we could have shipped this if we started with static mocks and Figma or if we started with a PRD.” 3. This is the year of the generalist (and maybe the year of those with ADHD) Boris’s work has shifted from deep-focus single-threaded coding to managing multiple parallel agents and context-switching rapidly. As Boris put it: “It’s not so much about deep work, it’s about how good I am at context switching and jumping across multiple different contexts very quickly.”
1:37:58

What does it mean for software engineering when we no longer write the code? Here's the take from Boris Cherny (Boris Cherny), the creator of Claude Code. Timestamps: 00:00 Intro 11:15 Lessons from Meta 19:46 Joining Anthropic 23:08 The origins of Claude Code 32:55 Boris's Claude Code workflow 36:27 Parallel agents 40:25 Code reviews 47:18 Claude Code's architecture 52:38 Permissions and sandboxing 55:05 Engineering culture at Anthropic 1:05:15 Claude Cowork 1:12:48 Observability and privacy 1:14:45 Agent swarms 1:21:16 LLMs and the printing press analogy 1:30:16 Standout engineer archetypes 1:32:12 What skills still matter for engineers 1:35:24 Book recommendations Brought to you by: • Statsig — ⁠ The unified platform for flags, analytics, experiments, and more. • Sonar – The makers of SonarQube, the industry standard for automated code review. Proactively find and fix issues in real-time with the SonarQube MCP Server: • WorkOS – Everything you need to make your app enterprise ready. Three interesting things from this conversation: 1. Boris automated himself out of code review well before AI. Boris was one of the most prolific code reviewers at Meta company. And he worked hard to minimize time spent on code review. His system::every time he left the same kind of review comment, he logged it in a spreadsheet. Once a pattern hit 3-4 occurrences, he’d write a lint rule to automate it away! 2. PRDs are dead on the Claude Code team: prototypes replaced them. Instead of writing Product Requirement Documents (specs), they build hundreds of working prototypes before shipping a feature. Boris: “There’s just no way we could have shipped this if we started with static mocks and Figma or if we started with a PRD.” 3. This is the year of the generalist (and maybe the year of those with ADHD) Boris’s work has shifted from deep-focus single-threaded coding to managing multiple parallel agents and context-switching rapidly. As Boris put it: “It’s not so much about deep work, it’s about how good I am at context switching and jumping across multiple different contexts very quickly.”

Gergely Orosz

489,982 views • 5 months ago

context engineering vs graph engineering. every few months the list gets a new word and everyone treats it as a replacement for the last one. these two are not on the same list. one decides what the model sees this turn, the other decides what exists at all. the cleanest way to tell them apart is to ask what a single unit of work looks like. > context engineering is the window the window opens empty, every single time. you assemble what goes in it. the prompt, the docs, the history, the tool results. the assembling is the work. the window only grows. it never shrinks on its own, so eventually something gets dropped. usually from the middle. usually without telling you. then the turn ends and the window is thrown away. not archived, thrown away. the next turn opens empty again and you re-explain what you already explained. good context engineering is knowing what to leave out, not what to pack in. the unit of work is one window. > graph engineering is the structure the same material arrives from the same sources. instead of packing it into a window, you pull entities out of it, resolve the duplicates into one node, and write typed edges between them. nothing here is stored as text you hope to find again. it is stored as a thing with a name and its connections to other things. when the turn ends, the graph is still there. the next turn does not start from zero. it starts by querying what already exists, and the query walks edges instead of guessing at similarity. good graph engineering is deciding what counts as the same thing twice. the unit of work is one relationship. > they are not alternatives the graph is what refills the window. context engineering decides what fits. graph engineering decides what there is to choose from. remove the graph and every session starts blind. remove the context work and the best structure in the world arrives as an unreadable dump. that also tells you which one broke. the answer drifted from what you actually said, or forgot something from this same session. that is the window. the answer is coherent but invents a connection that does not exist, or cannot join two facts it has clearly seen. that is the structure. people debug the prompt because the prompt is the easiest thing to edit. it keeps taking the blame for failures that live a layer down. save this - then read the full breakdown below
0:40

context engineering vs graph engineering. every few months the list gets a new word and everyone treats it as a replacement for the last one. these two are not on the same list. one decides what the model sees this turn, the other decides what exists at all. the cleanest way to tell them apart is to ask what a single unit of work looks like. > context engineering is the window the window opens empty, every single time. you assemble what goes in it. the prompt, the docs, the history, the tool results. the assembling is the work. the window only grows. it never shrinks on its own, so eventually something gets dropped. usually from the middle. usually without telling you. then the turn ends and the window is thrown away. not archived, thrown away. the next turn opens empty again and you re-explain what you already explained. good context engineering is knowing what to leave out, not what to pack in. the unit of work is one window. > graph engineering is the structure the same material arrives from the same sources. instead of packing it into a window, you pull entities out of it, resolve the duplicates into one node, and write typed edges between them. nothing here is stored as text you hope to find again. it is stored as a thing with a name and its connections to other things. when the turn ends, the graph is still there. the next turn does not start from zero. it starts by querying what already exists, and the query walks edges instead of guessing at similarity. good graph engineering is deciding what counts as the same thing twice. the unit of work is one relationship. > they are not alternatives the graph is what refills the window. context engineering decides what fits. graph engineering decides what there is to choose from. remove the graph and every session starts blind. remove the context work and the best structure in the world arrives as an unreadable dump. that also tells you which one broke. the answer drifted from what you actually said, or forgot something from this same session. that is the window. the answer is coherent but invents a connection that does not exist, or cannot join two facts it has clearly seen. that is the structure. people debug the prompt because the prompt is the easiest thing to edit. it keeps taking the blame for failures that live a layer down. save this - then read the full breakdown below

Hanako

19,089 views • 8 days ago

AI AGENTS 101 (58 minute free masterclass) send this to anyone who wants to understand ai agents, claude skills, md files, how to get the most out of AI etc in plain english: 1. chat vs agents - chat models answer questions in a back and forth while agents take a goal, figure out the steps, and deliver a result 2. agents don’t stop after one response. they keep running until the task is actually finishedno babysitting required 3. everything runs on a loop. they gather context, decide what to do, take an action, then repeat until done 4. the loop is the system. they look at files, tools, and the internet. decide the next step. execute and then feed that back into the next step. over and over until completion 5. the model is just one piece. gpt, claude, gemini are the reasoning layer. the key is model + loop + tools + context 6. mcp is how agents use tools. it connects things like browser, code, apis, and your internal software. once connected, the agent decides when to use them to get the job done 7. context beats prompt all day. you don't need to write perfect prompts. load your agent with context about your business, style, and goals and then simple instructions work 8. claude.md or agents.md is the onboarding doc it tells the agent who it is, how to behave, what it knows, and what tools it can use. this gets loaded every time before it starts 9. memory.md is how it improves. agents don’t remember by default. this file stores preferences, corrections, and patterns you tell the agent to update it, and it gets better over time 10. skills + harnesses make it usable. skills are reusable tasks like writing, research, analysis the harness is the environment like claude code or openclaw that runs everything. basiclaly, different interfaces, same system underneath this episode with remy on The Startup Ideas Podcast (SIP) 🧃 was one of the clearest ways of understanding a lot of the core concepts of ai agents could be the best beginners course for ai agents 58 mins. all free. no advertisers. i just want to see you build cool stuff. im rooting for you. send to a friend watch
58:55

AI AGENTS 101 (58 minute free masterclass) send this to anyone who wants to understand ai agents, claude skills, md files, how to get the most out of AI etc in plain english: 1. chat vs agents - chat models answer questions in a back and forth while agents take a goal, figure out the steps, and deliver a result 2. agents don’t stop after one response. they keep running until the task is actually finishedno babysitting required 3. everything runs on a loop. they gather context, decide what to do, take an action, then repeat until done 4. the loop is the system. they look at files, tools, and the internet. decide the next step. execute and then feed that back into the next step. over and over until completion 5. the model is just one piece. gpt, claude, gemini are the reasoning layer. the key is model + loop + tools + context 6. mcp is how agents use tools. it connects things like browser, code, apis, and your internal software. once connected, the agent decides when to use them to get the job done 7. context beats prompt all day. you don't need to write perfect prompts. load your agent with context about your business, style, and goals and then simple instructions work 8. claude.md or agents.md is the onboarding doc it tells the agent who it is, how to behave, what it knows, and what tools it can use. this gets loaded every time before it starts 9. memory.md is how it improves. agents don’t remember by default. this file stores preferences, corrections, and patterns you tell the agent to update it, and it gets better over time 10. skills + harnesses make it usable. skills are reusable tasks like writing, research, analysis the harness is the environment like claude code or openclaw that runs everything. basiclaly, different interfaces, same system underneath this episode with remy on The Startup Ideas Podcast (SIP) 🧃 was one of the clearest ways of understanding a lot of the core concepts of ai agents could be the best beginners course for ai agents 58 mins. all free. no advertisers. i just want to see you build cool stuff. im rooting for you. send to a friend watch

GREG ISENBERG

375,365 views • 4 months ago

Everyone wants agent swarms. Very few people are talking seriously enough about the context layer that makes swarms useful. Even with one agent, context is fragile. Too little context and the agent guesses. Too much context and it wastes tokens, loses focus, or reasons over irrelevant noise. The sweet spot is precise context: the right knowledge, in the right structure, at the right moment. With many agents, that challenge explodes. Each agent produces decisions, assumptions, findings, summaries, risks, and partial conclusions. Unless that knowledge becomes shared, structured, and reusable, every new agent is forced to rediscover what another agent already learned. That is not a swarm. That is a crowd. Shared context graphs are what turn agent activity into agent collaboration, and OriginTrail DKG V10 brings them to life. Was just playing with some final polishing for the V10 release, and it is really powerful to see shared context graphs where multiple agents contribute knowledge into the same connected memory, with attribution visible directly in the graph ui. That matters for three reasons. First, agents can access and build on one shared memory instead of staying trapped in isolated sessions. Second, the graph structure helps them retrieve the exact context they need, instead of stuffing everything into a prompt and hoping the model sorts it out. Third, verifiability of provenance. You can see which agent contributed each piece of knowledge, trace the source, and decide what to trust. Tokenmaxxing starts with fewer tokens, but the deeper story is coordination - agents stop reloading the world and start building on shared, verifiable context. That is the foundation for serious multi-agent work across software engineering, research, finance, operations, project management, and far beyond. The future is not more agents, it is agents working from shared, verifiable context. But the more the merrier, of course.
0:22

Everyone wants agent swarms. Very few people are talking seriously enough about the context layer that makes swarms useful. Even with one agent, context is fragile. Too little context and the agent guesses. Too much context and it wastes tokens, loses focus, or reasons over irrelevant noise. The sweet spot is precise context: the right knowledge, in the right structure, at the right moment. With many agents, that challenge explodes. Each agent produces decisions, assumptions, findings, summaries, risks, and partial conclusions. Unless that knowledge becomes shared, structured, and reusable, every new agent is forced to rediscover what another agent already learned. That is not a swarm. That is a crowd. Shared context graphs are what turn agent activity into agent collaboration, and OriginTrail DKG V10 brings them to life. Was just playing with some final polishing for the V10 release, and it is really powerful to see shared context graphs where multiple agents contribute knowledge into the same connected memory, with attribution visible directly in the graph ui. That matters for three reasons. First, agents can access and build on one shared memory instead of staying trapped in isolated sessions. Second, the graph structure helps them retrieve the exact context they need, instead of stuffing everything into a prompt and hoping the model sorts it out. Third, verifiability of provenance. You can see which agent contributed each piece of knowledge, trace the source, and decide what to trust. Tokenmaxxing starts with fewer tokens, but the deeper story is coordination - agents stop reloading the world and start building on shared, verifiable context. That is the foundation for serious multi-agent work across software engineering, research, finance, operations, project management, and far beyond. The future is not more agents, it is agents working from shared, verifiable context. But the more the merrier, of course.

Jurij Skornik

11,070 views • 2 months ago

Why is the creator of OpenCode pretty skeptical about AI productivity gains, and the hype around AI? A very conversation dax (and lots of truth bombs:) Timestamps: 00:00 Intro 07:03 Dax’s path into tech 09:04 Early startup experience 13:16 Getting involved with open source 16:13 OpenCode 23:17 Anthropic banning OpenCode 30:34 From terminal to GUI 32:34 OpenCode’s business model 36:33 Why inference is profitable 39:11 GPU bottlenecks 40:54 AI hype 45:50 AI spending 48:47 Dax’s memo 55:41 Dax’s skepticism of predictions 58:58 Engineering culture at OpenCode 1:02:38 How building works at OpenCode 1:05:36 Taste and quality 1:11:32 Dax’s work setup 1:12:35 The role of engineers and EMs 1:15:50 Advice for engineers 1:18:12 Book recommendation Brought to you by: • Antithesis – verify your system’s correctness without human review or traditional integration tests – and avoid bugs or outages • WorkOS – everything you need to make your app enterprise ready • turbopuffer – a vector and full-text search engine built on object storage. It’s fast, cheap, and extremely scalable Three interesting thoughts from Dax: 1. No AI-native coding agent company is “winning” by being better with AI. Dax says that none of OpenCode’s competitors are crushing them, and that nobody is using AI so well that others cannot compete. 2. Most software engineers profit from AI as time gained, not increased output — unless you change incentives! Dax says the natural way for software engineers to “cash out” their AI tooling gains is with time savings, by doing the same work as before, but faster. Until compensation and motivation structures change, most teams should expect output to stay flat while engineers go home earlier. There’s nothing wrong with this, but AI vendors sell a different outcome to CFOs: increased output. 3. AI code generation mutes the “guilt” of doing the wrong thing, but this builds up tech debt. Pre-AI, writing a hack felt bad, the second time it felt really bad, and by the third time you’d often just refactor in order to fix up the code. Now, the agent hides the hack, which skews devs’ judgment and results in less tech debt being cleaned up.
1:21:01

Why is the creator of OpenCode pretty skeptical about AI productivity gains, and the hype around AI? A very conversation dax (and lots of truth bombs:) Timestamps: 00:00 Intro 07:03 Dax’s path into tech 09:04 Early startup experience 13:16 Getting involved with open source 16:13 OpenCode 23:17 Anthropic banning OpenCode 30:34 From terminal to GUI 32:34 OpenCode’s business model 36:33 Why inference is profitable 39:11 GPU bottlenecks 40:54 AI hype 45:50 AI spending 48:47 Dax’s memo 55:41 Dax’s skepticism of predictions 58:58 Engineering culture at OpenCode 1:02:38 How building works at OpenCode 1:05:36 Taste and quality 1:11:32 Dax’s work setup 1:12:35 The role of engineers and EMs 1:15:50 Advice for engineers 1:18:12 Book recommendation Brought to you by: • Antithesis – verify your system’s correctness without human review or traditional integration tests – and avoid bugs or outages • WorkOS – everything you need to make your app enterprise ready • turbopuffer – a vector and full-text search engine built on object storage. It’s fast, cheap, and extremely scalable Three interesting thoughts from Dax: 1. No AI-native coding agent company is “winning” by being better with AI. Dax says that none of OpenCode’s competitors are crushing them, and that nobody is using AI so well that others cannot compete. 2. Most software engineers profit from AI as time gained, not increased output — unless you change incentives! Dax says the natural way for software engineers to “cash out” their AI tooling gains is with time savings, by doing the same work as before, but faster. Until compensation and motivation structures change, most teams should expect output to stay flat while engineers go home earlier. There’s nothing wrong with this, but AI vendors sell a different outcome to CFOs: increased output. 3. AI code generation mutes the “guilt” of doing the wrong thing, but this builds up tech debt. Pre-AI, writing a hack felt bad, the second time it felt really bad, and by the third time you’d often just refactor in order to fix up the code. Now, the agent hides the hack, which skews devs’ judgment and results in less tech debt being cleaned up.

Gergely Orosz

231,126 views • 2 months ago

"We are clearly entering a world where a product-minded engineer is now empowered to produce software without writing a line of code for it." In this conversation, Temporal CEO Samar Abbas joins a16z GPs Sarah Wang and Raghu Raghuram to cover: - Why agents are going from short-lived and interactive to long-running and async - Why the engineer of the future manages 15 parallel AI tasks at once - How OpenAI Codex runs millions of concurrent agent executions on Temporal - How real-time context engineering is exploding on Temporal - Why SaaS isn't dead, and value is migrating to APIs 00:00 Introduction 04:03 Temporal's origin story 11:14 Why agents raise the stakes 16:00 Specialized agents need durable RPC 25:20 Deep research agents 30:58 Execution histories as a superpower 39:04 Minimal viable long-running agent architecture 45:07 Context engineering at scale 52:40 Where value accrues: The "five-layer cake" and breakout AI applications Samar Abbas Sarah Wang Raghu Raghuram
1:02:30

"We are clearly entering a world where a product-minded engineer is now empowered to produce software without writing a line of code for it." In this conversation, Temporal CEO Samar Abbas joins a16z GPs Sarah Wang and Raghu Raghuram to cover: - Why agents are going from short-lived and interactive to long-running and async - Why the engineer of the future manages 15 parallel AI tasks at once - How OpenAI Codex runs millions of concurrent agent executions on Temporal - How real-time context engineering is exploding on Temporal - Why SaaS isn't dead, and value is migrating to APIs 00:00 Introduction 04:03 Temporal's origin story 11:14 Why agents raise the stakes 16:00 Specialized agents need durable RPC 25:20 Deep research agents 30:58 Execution histories as a superpower 39:04 Minimal viable long-running agent architecture 45:07 Context engineering at scale 52:40 Where value accrues: The "five-layer cake" and breakout AI applications Samar Abbas Sarah Wang Raghu Raghuram

a16z

45,721 views • 5 months ago

What if we're actually in the middle of the third golden age of software engineering? This is what Grady Booch sees happening. If you are anxious about the state of the industry, you want to watch/listen to Grady's longer-term perspective and stories. Watch the full episode here: 00:00 Intro 01:58 The first golden age of software engineering 18:59 The software crisis 33:01 The second golden age of software engineering 42:21 Y2K and the Dotcom crash 45:47 Early AI 47:34 The third golden age of software engineering 51:48 Why software engineers will very much be needed 58:46 Grady responds to Dario Amodei 1:06:54 New skills engineers will need to succeed 1:10:04 Resources for studying complex systems 1:14:33 How to thrive during periods of change Brought to you by: • Statsig — ⁠ The unified platform for flags, analytics, experiments, and more. • Sonar – The makers of SonarQube, the industry standard for automated code review. Join me online at the Sonar Summit on March 3rd, where I talk about practical tactics for the AI era. • WorkOS – Everything you need to make your app enterprise-ready.
1:17:58

What if we're actually in the middle of the third golden age of software engineering? This is what Grady Booch sees happening. If you are anxious about the state of the industry, you want to watch/listen to Grady's longer-term perspective and stories. Watch the full episode here: 00:00 Intro 01:58 The first golden age of software engineering 18:59 The software crisis 33:01 The second golden age of software engineering 42:21 Y2K and the Dotcom crash 45:47 Early AI 47:34 The third golden age of software engineering 51:48 Why software engineers will very much be needed 58:46 Grady responds to Dario Amodei 1:06:54 New skills engineers will need to succeed 1:10:04 Resources for studying complex systems 1:14:33 How to thrive during periods of change Brought to you by: • Statsig — ⁠ The unified platform for flags, analytics, experiments, and more. • Sonar – The makers of SonarQube, the industry standard for automated code review. Join me online at the Sonar Summit on March 3rd, where I talk about practical tactics for the AI era. • WorkOS – Everything you need to make your app enterprise-ready.

Gergely Orosz

86,896 views • 6 months ago

Dario Amodei just announced the end of software engineering as a profession. The timeline is 6 to 12 months. Amodei: “I have engineers within Anthropic who say, I don’t write any code anymore. I just let the model write the code. I edit it.” Not a prediction. Current reality inside the frontier lab. The engineers who built the most advanced AI in the world have stopped writing code. They supervise. They edit. They manage architecture. The craft they spent careers mastering has been handed to the system they built. Amodei says models will do most, maybe all, of what software engineers do end-to-end within six to twelve months. Not assisting. Not autocompleting. Handling the entire development process independently. If you are learning syntax today, you are learning a dead language. Amodei: “Then it’s a question of how fast does that loop close?” The loop is this. AI writes code. Code builds better AI. Better AI writes better code. Faster. Without sleep. Without the cognitive limits that cap how quickly any human engineer can work. Once that loop closes, technological progress stops being constrained by human output. It becomes self-sustaining. Exponential. Operating at a pace no human workforce can match or direct. Software engineering isn’t ending. It’s becoming supervision. The developers who survive won’t be the best coders. They’ll be the best supervisors. The ones who can direct AI output, catch its failures, and architect what it builds toward. The skill that matters stops being implementation. It becomes judgment. Most developers are still optimizing for a skillset about to become as obsolete as stenography. While the people who built the systems replacing them already stopped doing the work themselves. The window to develop that judgment before the loop closes is exactly as long as Amodei’s timeline. Six to twelve months.
0:26

Dario Amodei just announced the end of software engineering as a profession. The timeline is 6 to 12 months. Amodei: “I have engineers within Anthropic who say, I don’t write any code anymore. I just let the model write the code. I edit it.” Not a prediction. Current reality inside the frontier lab. The engineers who built the most advanced AI in the world have stopped writing code. They supervise. They edit. They manage architecture. The craft they spent careers mastering has been handed to the system they built. Amodei says models will do most, maybe all, of what software engineers do end-to-end within six to twelve months. Not assisting. Not autocompleting. Handling the entire development process independently. If you are learning syntax today, you are learning a dead language. Amodei: “Then it’s a question of how fast does that loop close?” The loop is this. AI writes code. Code builds better AI. Better AI writes better code. Faster. Without sleep. Without the cognitive limits that cap how quickly any human engineer can work. Once that loop closes, technological progress stops being constrained by human output. It becomes self-sustaining. Exponential. Operating at a pace no human workforce can match or direct. Software engineering isn’t ending. It’s becoming supervision. The developers who survive won’t be the best coders. They’ll be the best supervisors. The ones who can direct AI output, catch its failures, and architect what it builds toward. The skill that matters stops being implementation. It becomes judgment. Most developers are still optimizing for a skillset about to become as obsolete as stenography. While the people who built the systems replacing them already stopped doing the work themselves. The window to develop that judgment before the loop closes is exactly as long as Amodei’s timeline. Six to twelve months.

Dustin

44,304 views • 5 months ago

Here's my conversation with the founding team of Cursor, a popular code editor (based on VSCode) that specializes in AI-assisted programming. This is a super technical conversation that is bigger than just about one code editor. It's about the future of programming and, in general, the future of human-AI collaboration. It's here on X in full, and is up on everywhere else. Timestamps: 0:00 - Introduction 0:59 - Code editor basics 3:09 - GitHub Copilot 10:27 - Cursor 16:54 - Cursor Tab 23:08 - Code diff 31:20 - ML details 36:54 - GPT vs Claude 43:28 - Prompt engineering 50:54 - AI agents 1:04:51 - Running code in background 1:09:31 - Debugging 1:14:58 - Dangerous code 1:26:09 - Branching file systems 1:29:20 - Scaling challenges 1:43:32 - Context 1:48:39 - OpenAI o1 2:00:01 - Synthetic data 2:03:48 - RLHF vs RLAIF 2:05:34 - Fields Medal for AI 2:08:17 - Scaling laws 2:17:06 - The future of programming
2:28:55

Here's my conversation with the founding team of Cursor, a popular code editor (based on VSCode) that specializes in AI-assisted programming. This is a super technical conversation that is bigger than just about one code editor. It's about the future of programming and, in general, the future of human-AI collaboration. It's here on X in full, and is up on everywhere else. Timestamps: 0:00 - Introduction 0:59 - Code editor basics 3:09 - GitHub Copilot 10:27 - Cursor 16:54 - Cursor Tab 23:08 - Code diff 31:20 - ML details 36:54 - GPT vs Claude 43:28 - Prompt engineering 50:54 - AI agents 1:04:51 - Running code in background 1:09:31 - Debugging 1:14:58 - Dangerous code 1:26:09 - Branching file systems 1:29:20 - Scaling challenges 1:43:32 - Context 1:48:39 - OpenAI o1 2:00:01 - Synthetic data 2:03:48 - RLHF vs RLAIF 2:05:34 - Fields Medal for AI 2:08:17 - Scaling laws 2:17:06 - The future of programming

Lex Fridman

1,261,295 views • 1 year ago

Engineers need to communicate effectively when building AI Systems. These terms will help you use a shared vocabulary. This is useful when discussing concepts, reading papers, or collaborating with teammates. Listed in order. 00:00 Agenda 00:28 1. Large Language Model 01:28 2. Tokenization 02:53 3. Vectorization 04:15 4. Attention 07:22 5. Self-Supervised Learning 12:07 6. Transformer 14:32 7. Fine-tuning 17:05 8. Few-shot Prompting 18:11 9. Retrieval Augmented Generation 20:33 10. Vector Database 23:03 11. Model Context Protocol 25:43 12. Context Engineering 28:17 13. Agents 29:19 14. Reinforcement Learning 34:42 15. Chain of Thought 35:55 16. Reasoning Models 36:36 17. Multi-modal Models 38:21 18. Small Language Models 40:24 19. Distillation 41:47 20. Quantization If you are a software engineer looking to transition to AI, click the link below. AI Engineering Course: #AI #SoftwareEngineering #Agents
43:33

Engineers need to communicate effectively when building AI Systems. These terms will help you use a shared vocabulary. This is useful when discussing concepts, reading papers, or collaborating with teammates. Listed in order. 00:00 Agenda 00:28 1. Large Language Model 01:28 2. Tokenization 02:53 3. Vectorization 04:15 4. Attention 07:22 5. Self-Supervised Learning 12:07 6. Transformer 14:32 7. Fine-tuning 17:05 8. Few-shot Prompting 18:11 9. Retrieval Augmented Generation 20:33 10. Vector Database 23:03 11. Model Context Protocol 25:43 12. Context Engineering 28:17 13. Agents 29:19 14. Reinforcement Learning 34:42 15. Chain of Thought 35:55 16. Reasoning Models 36:36 17. Multi-modal Models 38:21 18. Small Language Models 40:24 19. Distillation 41:47 20. Quantization If you are a software engineer looking to transition to AI, click the link below. AI Engineering Course: #AI #SoftwareEngineering #Agents

Gaurav Sen

80,115 views • 10 months 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
43:20

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 📧

66,339 views • 3 months ago

RLM is the most import foundation of my Pi Harness (other than Pi of course). It's seeded with late interaction retrieval results (thanks to @lightonai for pylate). The Agent initiates it with query then.. 𝐒𝐞𝐭𝐮𝐩 A python REPL is created and seeded with: 1. Late interaction search to pre-filter. Instead of doing top 3/5/10, it's top hundreds of documents. This is set into a `context` variable. 2. Python functions are loaded in to do more searches if `context` variable isn't enough. And to make llm calls with cheaper models in parallel batches. 𝐈𝐭𝐞𝐫𝐚𝐭𝐢𝐨𝐧 𝐋𝐨𝐨𝐩 From there, an LLM iterates in the REPL based on the query. It's just like exploring in a jupyter notebook. The LLM writes prose (like a markdown cell) and code to be run in the REPL each turn. This allows the LLM to sort, filter, and synthesize information. It can fan out and ask smaller models to summarize, combine, contrast, or do anything else to documents to help it understand the data. After several turns the LLM reponds with the final answer. Either because it found the answer, or hit the budget limit. Context as a Python variable, LLM as the programmer, REPL as the runtime. 𝐖𝐡𝐲 𝐃𝐨𝐞𝐬 𝐓𝐡𝐢𝐬 𝐖𝐨𝐫𝐤 1. Richer Shell. Agents (and subagents) work by intermixing code and prose/thinking. But they use static scripts or bash that run and exit and start over each tool call. That's not ideal for exploration and synthesis of data. For that, state is useful to continue building and exploring the data as you learn more. There's a reason jupyter notebooks have been popular with data scientists. 2. Keeps main agent context clean. The better context you have the better the agent will perform (duh!). This means three thing: better human input, less missing search results, and less incorrect search results. Letting the agent iterate allows it to synthesize just what is needed and nothing else. All bad paths or peeks at something that turns out to be irrelevant stays out of main agent context. 3. Stack the good ideas! People often compare late interaction search vs RLM. Or static vs dynamic languages. Or agentic search vs semantic search. But...You can just use them all together for what they're each good at. Use them all for the area they're really great for. Read the full post which has more detail about how and why.
3:43

RLM is the most import foundation of my Pi Harness (other than Pi of course). It's seeded with late interaction retrieval results (thanks to @lightonai for pylate). The Agent initiates it with query then.. 𝐒𝐞𝐭𝐮𝐩 A python REPL is created and seeded with: 1. Late interaction search to pre-filter. Instead of doing top 3/5/10, it's top hundreds of documents. This is set into a `context` variable. 2. Python functions are loaded in to do more searches if `context` variable isn't enough. And to make llm calls with cheaper models in parallel batches. 𝐈𝐭𝐞𝐫𝐚𝐭𝐢𝐨𝐧 𝐋𝐨𝐨𝐩 From there, an LLM iterates in the REPL based on the query. It's just like exploring in a jupyter notebook. The LLM writes prose (like a markdown cell) and code to be run in the REPL each turn. This allows the LLM to sort, filter, and synthesize information. It can fan out and ask smaller models to summarize, combine, contrast, or do anything else to documents to help it understand the data. After several turns the LLM reponds with the final answer. Either because it found the answer, or hit the budget limit. Context as a Python variable, LLM as the programmer, REPL as the runtime. 𝐖𝐡𝐲 𝐃𝐨𝐞𝐬 𝐓𝐡𝐢𝐬 𝐖𝐨𝐫𝐤 1. Richer Shell. Agents (and subagents) work by intermixing code and prose/thinking. But they use static scripts or bash that run and exit and start over each tool call. That's not ideal for exploration and synthesis of data. For that, state is useful to continue building and exploring the data as you learn more. There's a reason jupyter notebooks have been popular with data scientists. 2. Keeps main agent context clean. The better context you have the better the agent will perform (duh!). This means three thing: better human input, less missing search results, and less incorrect search results. Letting the agent iterate allows it to synthesize just what is needed and nothing else. All bad paths or peeks at something that turns out to be irrelevant stays out of main agent context. 3. Stack the good ideas! People often compare late interaction search vs RLM. Or static vs dynamic languages. Or agentic search vs semantic search. But...You can just use them all together for what they're each good at. Use them all for the area they're really great for. Read the full post which has more detail about how and why.

Isaac Flath

40,212 views • 3 months ago

We've built 40+ AI agents and internal tools. The hardest part is Context Creation. AI runs playbooks and makes judgment calls for you. But without your company's context, you get slop. Context Creation means extracting the subject matter expertise and playbooks that live in people's heads, not in LLM training data, or even your tools. As forward deployed engineers (FDEs), we create context and turn it into code. We evaluate the business impact, how it aligns with the dev roadmap, and come up with creative solutions. We built The FDE Factory to replace ourselves. It drives AI adoption inside our clients' companies by running discovery sessions using prototypes to create context. Here's how it works: We put a prototype in front of a stakeholder. The stakeholder gives feedback via voice while they're using or reviewing it. Then our FDE Factory Agents builds in their expertise in minutes: > Context Agent reviews the codebase and feedback, extracts the requirements, and creates a spec > Scope Agent checks the spec against the development roadmap, validates it, and hands it off > Engineering Agent builds a new feature and wires the integration > QA Agent runs tests to prove to itself it works > PR merges, feature goes live, product updates itself in real time It's like the nontechnical stakeholder wrote the code without even knowing it. Coding agents are great at turning good development plans into code, and they're getting better at turning context into good development plans in collaboration with professional engineers. But nontechnical people are capped on what they can build without product people and engineers. The bridge that takes nontechnical people from vibe coding basic apps to building production AI tools that run on first party context is FDEs. Our new FDE Factory gives you the system to go from idea to production. Context Creation is the first and most important step in our FDE lifecycle, and we just automated it. Now clients get the right agents and tools built for them, customized to their unique business and encoded with their expertise. PS: If you're building AI agents within your company, reply "Playbook" and I'll DM you the entire FDE playbook we've run with 30+ companies. It covers finding high-impact AI use cases, building them, and deploying them across the org.
3:40

We've built 40+ AI agents and internal tools. The hardest part is Context Creation. AI runs playbooks and makes judgment calls for you. But without your company's context, you get slop. Context Creation means extracting the subject matter expertise and playbooks that live in people's heads, not in LLM training data, or even your tools. As forward deployed engineers (FDEs), we create context and turn it into code. We evaluate the business impact, how it aligns with the dev roadmap, and come up with creative solutions. We built The FDE Factory to replace ourselves. It drives AI adoption inside our clients' companies by running discovery sessions using prototypes to create context. Here's how it works: We put a prototype in front of a stakeholder. The stakeholder gives feedback via voice while they're using or reviewing it. Then our FDE Factory Agents builds in their expertise in minutes: > Context Agent reviews the codebase and feedback, extracts the requirements, and creates a spec > Scope Agent checks the spec against the development roadmap, validates it, and hands it off > Engineering Agent builds a new feature and wires the integration > QA Agent runs tests to prove to itself it works > PR merges, feature goes live, product updates itself in real time It's like the nontechnical stakeholder wrote the code without even knowing it. Coding agents are great at turning good development plans into code, and they're getting better at turning context into good development plans in collaboration with professional engineers. But nontechnical people are capped on what they can build without product people and engineers. The bridge that takes nontechnical people from vibe coding basic apps to building production AI tools that run on first party context is FDEs. Our new FDE Factory gives you the system to go from idea to production. Context Creation is the first and most important step in our FDE lifecycle, and we just automated it. Now clients get the right agents and tools built for them, customized to their unique business and encoded with their expertise. PS: If you're building AI agents within your company, reply "Playbook" and I'll DM you the entire FDE playbook we've run with 30+ companies. It covers finding high-impact AI use cases, building them, and deploying them across the org.

Mike Fishbein

10,141 views • 2 months ago

AI is changing the software engineering craft. Anders Hejlsberg (Anders Hejlsberg) - creator of C#, TypeScript and industry legend - on why code review needs to get more enjoyable in response: #1 - AI is shifting the craft from writing code, to reviewing code: "In a sense, we're all turning into project managers. We can have an army of junior programmers, called agents, that will just spit out reams of code but someone's got to have the big picture and review all of that. And so, increasingly, our craft is going from one of writing the code, to one of reviewing the code and building the architecture of the code and overseeing the work. It's a different kind of craft. It's a different kind of enjoyment. I've always liked writing the code. To me that was the fulfilling part, seeing it work. In a way, AI robs a little bit of that, because I am less interested in reviewing code." #2 - The code review experience should be improved: "I think we could also make the process of reviewing code much more interesting than it is today. I mean, today, you see a list of diffs in alphabetical order and now it's up to you to make heads or tails of it. There are more pedagogical ways of presenting that. And you could have commentary generated by the AI that tells you what the changes are and whatever, and then tries to guide you along. So that symbiotic relationship, I think we need to work on that more and to keep the enjoyment in there."
2:04

AI is changing the software engineering craft. Anders Hejlsberg (Anders Hejlsberg) - creator of C#, TypeScript and industry legend - on why code review needs to get more enjoyable in response: #1 - AI is shifting the craft from writing code, to reviewing code: "In a sense, we're all turning into project managers. We can have an army of junior programmers, called agents, that will just spit out reams of code but someone's got to have the big picture and review all of that. And so, increasingly, our craft is going from one of writing the code, to one of reviewing the code and building the architecture of the code and overseeing the work. It's a different kind of craft. It's a different kind of enjoyment. I've always liked writing the code. To me that was the fulfilling part, seeing it work. In a way, AI robs a little bit of that, because I am less interested in reviewing code." #2 - The code review experience should be improved: "I think we could also make the process of reviewing code much more interesting than it is today. I mean, today, you see a list of diffs in alphabetical order and now it's up to you to make heads or tails of it. There are more pedagogical ways of presenting that. And you could have commentary generated by the AI that tells you what the changes are and whatever, and then tries to guide you along. So that symbiotic relationship, I think we need to work on that more and to keep the enjoyment in there."

The Pragmatic Engineer

39,011 views • 2 months ago

Getting the most out of Claude Fable 5, Anthropic’s powerful new model, you need to maximize your ambition: It’s built for full task delegation—you leave it looping for hours or overnight and come back to a finished product. If you want to get the most out of it, you need to relearn what software engineering is and how to step away to let the model do its work. That’s why I invited Mike Krieger, head of Anthropic Labs, on Every 📧’s AI & I. Mike’s been using Mythos-class models for a few months now internally at Anthropic, and he’s learned a ton of new tricks to make its increased powers work for him. And, as a co-founder of Instagram, he can reflect on how software engineering has changed over the last 15 years and what it means going forward. We get into: - Why the right workflow for Fable 5 is overnight delegation, not back-and-forth iteration—Mike ends his workday by briefing the model, then wakes up to a completed task. When a remote service went down mid-task, Fable 5 wrote a workaround, documented it, and forged ahead - The gap between what’s in your head and what exists in the world is closing fast—given access to Fable 5 and a set of internal MCPs, an Anthropic recruiter described the experience as, "The first time in my life where I feel like the thing that's in my head and the thing that exists in the world are right next to each other. I can just do it." - Software engineering isn’t dead, but the role has been reinvented—the PM/eng split is blurring, and the better engineers Mike talks to are holding two feelings at once: loss for the craft and shock at what’s now possible - Verification is the new bottleneck—Mike gives Fable video captures of its own work so it can catch animation glitches that screenshots would miss This is a must-watch for anyone building software and trying to figure out their role now that the models can handle so much. Watch below! Timestamps Introduction: 00:00:03 How Fable completely reshaped Mike's workflow: 00:01:48 When to use Sonnet versus Fable: 00:04:48 What the media tracker Mike built over a weekend reveals about agent-native architecture: 00:10:06 The cost to build has collapsed: 00:15:00 Is software engineering over?: 00:19:03 How Anthropic's engineering teams work today: 00:21:48 The mechanics of verification: 00:38:39 Dynamic workflows: 00:47:24 What people should use the model to build: 00:44:39
52:05

Getting the most out of Claude Fable 5, Anthropic’s powerful new model, you need to maximize your ambition: It’s built for full task delegation—you leave it looping for hours or overnight and come back to a finished product. If you want to get the most out of it, you need to relearn what software engineering is and how to step away to let the model do its work. That’s why I invited Mike Krieger, head of Anthropic Labs, on Every 📧’s AI & I. Mike’s been using Mythos-class models for a few months now internally at Anthropic, and he’s learned a ton of new tricks to make its increased powers work for him. And, as a co-founder of Instagram, he can reflect on how software engineering has changed over the last 15 years and what it means going forward. We get into: - Why the right workflow for Fable 5 is overnight delegation, not back-and-forth iteration—Mike ends his workday by briefing the model, then wakes up to a completed task. When a remote service went down mid-task, Fable 5 wrote a workaround, documented it, and forged ahead - The gap between what’s in your head and what exists in the world is closing fast—given access to Fable 5 and a set of internal MCPs, an Anthropic recruiter described the experience as, "The first time in my life where I feel like the thing that's in my head and the thing that exists in the world are right next to each other. I can just do it." - Software engineering isn’t dead, but the role has been reinvented—the PM/eng split is blurring, and the better engineers Mike talks to are holding two feelings at once: loss for the craft and shock at what’s now possible - Verification is the new bottleneck—Mike gives Fable video captures of its own work so it can catch animation glitches that screenshots would miss This is a must-watch for anyone building software and trying to figure out their role now that the models can handle so much. Watch below! Timestamps Introduction: 00:00:03 How Fable completely reshaped Mike's workflow: 00:01:48 When to use Sonnet versus Fable: 00:04:48 What the media tracker Mike built over a weekend reveals about agent-native architecture: 00:10:06 The cost to build has collapsed: 00:15:00 Is software engineering over?: 00:19:03 How Anthropic's engineering teams work today: 00:21:48 The mechanics of verification: 00:38:39 Dynamic workflows: 00:47:24 What people should use the model to build: 00:44:39

Dan Shipper 📧

40,430 views • 1 month ago

Happy to properly launch Anna, the proactive AI agent for parents! Uncovering a bit of the technology behind the scenes! Building Anna is where I learned: 💾 Memory as plain text sucks. You need structured memory. Like a full-blown PostgreSQL DB that stores your tasks and calendar in a structured manner. Most harnesses are good at coding-related stuff. Let it do the query. Don't let it vibe-search the memory. Let it vibe your SQL query 💭 Dreaming is a useful concept for enhancing memory to feed the LLM context. But DO NOT vibe your dream. Asking your agent to "hey, just dream and keep the relevant memory around" is a recipe for deleting a bunch of important information and keeping trash around. Your dream needs to have some Taxonomy (or better, Ontology). What information is important? For who? With what object? What can they do? And again, these are impossible to describe and act well without a proper schema 🔄 Loop Engineering is important for smoothing out rough edges in the system we build. But even expensive loop engineering with a state-of-the-art model can't out-engineer bad system design. The highest leverage an AI Engineer can do is actually building the right system design, and having an eye on both product delight and engineering scalability There are several more insights that I plan to cover in a dedicated video about Agentic AI Engineering. But it's actually a huge relief that the future of software engineering... is still software engineering
2:10

Happy to properly launch Anna, the proactive AI agent for parents! Uncovering a bit of the technology behind the scenes! Building Anna is where I learned: 💾 Memory as plain text sucks. You need structured memory. Like a full-blown PostgreSQL DB that stores your tasks and calendar in a structured manner. Most harnesses are good at coding-related stuff. Let it do the query. Don't let it vibe-search the memory. Let it vibe your SQL query 💭 Dreaming is a useful concept for enhancing memory to feed the LLM context. But DO NOT vibe your dream. Asking your agent to "hey, just dream and keep the relevant memory around" is a recipe for deleting a bunch of important information and keeping trash around. Your dream needs to have some Taxonomy (or better, Ontology). What information is important? For who? With what object? What can they do? And again, these are impossible to describe and act well without a proper schema 🔄 Loop Engineering is important for smoothing out rough edges in the system we build. But even expensive loop engineering with a state-of-the-art model can't out-engineer bad system design. The highest leverage an AI Engineer can do is actually building the right system design, and having an eye on both product delight and engineering scalability There are several more insights that I plan to cover in a dedicated video about Agentic AI Engineering. But it's actually a huge relief that the future of software engineering... is still software engineering

Gogo | Dota for Toxicity

30,766 views • 1 month ago

OpenClaw - the agentic software spreading like wildfire - was built on top of Pi, a minimalist, self-modifying agent. I sat down with Pi's creator, Mario Zechner and longtime Pi user (+ the creator of Flask) Armin Ronacher ⇌ to talk Pi, and their (very grounded!) takes on building with AI. Timestamps: 00:00 Intro 07:30 How Mario, Armin, and Peter Steinberger met 15:15 How 30 dev teams use AI agents: learnings 21:50 The importance of judgment 24:26 Challenges when non-engineers write code 28:30 Downsides of over-automation 32:18 Pi 48:09 OpenClaw + Pi 50:54 “Clankers” 57:32 Open source and AI 1:00:22 Complexity as the enemy 1:02:50 Building an AI-native startup 1:11:52 “Slow the F down” 1:16:40 MCPs vs. CLI 1:25:03 Predictions and staying up to date • YouTube: • Spotify: • Apple: Brought to you by: • Statsig – ⁠ The unified platform for flags, analytics, experiments, and more. • Sonar — The makers of SonarQube, the industry standard for code verification and automated code review. Try it out for yourself. • WorkOS – WorkOS gives you APIs to ship enterprise features – SSO, directory sync, RBAC, audit logs – in days, not months. Visit learn more. --- Three parts I found especially interesting in this discussion: 1. New trend: AI makes it harder for senior engineers to reject pointless complexity. Historically, senior engineers kept software complexity at bay simply by saying “no” a lot. But Armin observes that these days, more junior engineers and product managers deploy agent-scripted counterarguments when a senior colleague kicks an idea to the curb. This makes decision-making exhausting, and more bad ideas make it into production as a result. 2. It should be MUCH easier to build specialized tools for specific tasks. Different projects need different harness types because, as Mario points out, the same hammer is not ideal for every single construction job. As such, Pi is built with the goal of allowing the creation of specialized harnesses. It can modify itself so that a user can create the bespoke harness needed for any task. Mario believes it’s a preview of how self-modifiable software might look in the future. 3. Automation bias is one of the biggest risks of working with AI agents. Once devs confirm that an AI agent can produce acceptable code, they start to review its output less often, even though agents can – and do! – produce slop. Mario advises being far more sceptical with agents, and cautions that the quality of their output isn’t guaranteed, however well they performed previously.
1:33:55

OpenClaw - the agentic software spreading like wildfire - was built on top of Pi, a minimalist, self-modifying agent. I sat down with Pi's creator, Mario Zechner and longtime Pi user (+ the creator of Flask) Armin Ronacher ⇌ to talk Pi, and their (very grounded!) takes on building with AI. Timestamps: 00:00 Intro 07:30 How Mario, Armin, and Peter Steinberger met 15:15 How 30 dev teams use AI agents: learnings 21:50 The importance of judgment 24:26 Challenges when non-engineers write code 28:30 Downsides of over-automation 32:18 Pi 48:09 OpenClaw + Pi 50:54 “Clankers” 57:32 Open source and AI 1:00:22 Complexity as the enemy 1:02:50 Building an AI-native startup 1:11:52 “Slow the F down” 1:16:40 MCPs vs. CLI 1:25:03 Predictions and staying up to date • YouTube: • Spotify: • Apple: Brought to you by: • Statsig – ⁠ The unified platform for flags, analytics, experiments, and more. • Sonar — The makers of SonarQube, the industry standard for code verification and automated code review. Try it out for yourself. • WorkOS – WorkOS gives you APIs to ship enterprise features – SSO, directory sync, RBAC, audit logs – in days, not months. Visit learn more. --- Three parts I found especially interesting in this discussion: 1. New trend: AI makes it harder for senior engineers to reject pointless complexity. Historically, senior engineers kept software complexity at bay simply by saying “no” a lot. But Armin observes that these days, more junior engineers and product managers deploy agent-scripted counterarguments when a senior colleague kicks an idea to the curb. This makes decision-making exhausting, and more bad ideas make it into production as a result. 2. It should be MUCH easier to build specialized tools for specific tasks. Different projects need different harness types because, as Mario points out, the same hammer is not ideal for every single construction job. As such, Pi is built with the goal of allowing the creation of specialized harnesses. It can modify itself so that a user can create the bespoke harness needed for any task. Mario believes it’s a preview of how self-modifiable software might look in the future. 3. Automation bias is one of the biggest risks of working with AI agents. Once devs confirm that an AI agent can produce acceptable code, they start to review its output less often, even though agents can – and do! – produce slop. Mario advises being far more sceptical with agents, and cautions that the quality of their output isn’t guaranteed, however well they performed previously.

Gergely Orosz

172,615 views • 3 months ago

Last week $TEAM outlined a new AI bull case for SaaS, as a source of context to generate better, faster, cheaper AI outcomes. Atlassian demod how Claude Code by itself underperformed an identical instance of Claude Code, on an identical prompt, when the other instance had access to their Teamwork Graph (TWG) which has access to past project files, messages, documents, code bases, employee data, and system integrations (slack, outlook, etc). The added context pointed Claude to the right information to solve the task faster, whereas the instance without context thought more, read more, hallucinated, and got the task incorrect after spending 2x more tokens and running an extra minute. Since token spend is quickly becoming many multiples of traditional software cost for coding use cases (Jira at $10-12/seat vs Anthropic spend per seat in the several hundreds a month), Atlassian makes a compelling case that context pays for itself many times over by helping customers to get more value out of their expensive AI tokens. $MSFT easily has the strongest advantage as a source of differentiated context. Video worth watching.
4:52

Last week $TEAM outlined a new AI bull case for SaaS, as a source of context to generate better, faster, cheaper AI outcomes. Atlassian demod how Claude Code by itself underperformed an identical instance of Claude Code, on an identical prompt, when the other instance had access to their Teamwork Graph (TWG) which has access to past project files, messages, documents, code bases, employee data, and system integrations (slack, outlook, etc). The added context pointed Claude to the right information to solve the task faster, whereas the instance without context thought more, read more, hallucinated, and got the task incorrect after spending 2x more tokens and running an extra minute. Since token spend is quickly becoming many multiples of traditional software cost for coding use cases (Jira at $10-12/seat vs Anthropic spend per seat in the several hundreds a month), Atlassian makes a compelling case that context pays for itself many times over by helping customers to get more value out of their expensive AI tokens. $MSFT easily has the strongest advantage as a source of differentiated context. Video worth watching.

SouthernValue

31,196 views • 2 months ago

Why did Uber build thousands of microservices? No better person to answer than Uber's first CTO, Thuan Pham. Timestamps: 00:00 Intro 05:32 Getting into tech 16:09 The dot-com bust 20:42 VMware 26:29 Getting hired by Travis at Uber 33:22 Early days at Uber and scaling challenges 40:57 Uber’s China launch 47:12 The platform and program split 50:26 From monolith to microservices 53:38 Internal tools at Uber 57:05 Helix: Uber’s mobile app rewrite 59:55 Thuan’s email about naming 1:02:03 Org structure changes under 1:06:34 Thuan’s work philosophy 1:12:23 The “three tours of duty” at Uber 1:15:37 Why Thuan left Uber 1:17:34 Coupang and Nubank 1:21:59 Faire 1:25:31 How Faire uses AI 1:28:24 AI’s impact on software engineering 1:31:09 The role of the CTO 1:35:13 Career advice Brought to you by: • Statsig – ⁠ The unified platform for flags, analytics, experiments, and more. • WorkOS – Everything you need to make your app enterprise ready. • Sonar – The makers of SonarQube, the industry standard for automated code review. Check out SonarQube Advanced Security: Three interesting parts from this conversation: 1. The program/platform split came before microservices. The concept of cross-functional “program” teams and dedicated “platform” teams became necessary because an org split across backend, frontend and mobile engineers slowed down in execution speed when Uber grew to around 100 engineers. Every feature required negotiating bandwidth across the mobile, backend, and dispatch teams. Thuan, Travis Kalanick, and Jeff Holden literally used color-coded sticky notes with people’s names to reorganize into self-sufficient teams. We cover more about this split in this The Pragmatic Engineer deepdive, The Platform and Program split at Uber: 2. Expect multiple rewrites during hypergrowth. The right architecture depends on how fast a product and company are growing. At Uber, repeated rewrites were common because each one “bought” another window of survival for the company. Thuan’s recommendation is to understand that a rewrite simply means a company is outrunning its existing architecture: this is not necessarily a bad thing! 3. Uber is the only major company that had a “Senior 1” and “Senior 2” level – and Thuan is unapologetic. Thuan introduced the Senior 1 (L5A) and Senior 2 (L5B) levels because the jump from senior (L5) to Staff (L6) became very big, and larger than between previous levels. One problem this split level created was that Uber’s L5B was akin to Google’s and Facebook’s L6/E6. Thuan resisted the title inflation of just renaming L5B to ‘Staff’.
1:39:34

Why did Uber build thousands of microservices? No better person to answer than Uber's first CTO, Thuan Pham. Timestamps: 00:00 Intro 05:32 Getting into tech 16:09 The dot-com bust 20:42 VMware 26:29 Getting hired by Travis at Uber 33:22 Early days at Uber and scaling challenges 40:57 Uber’s China launch 47:12 The platform and program split 50:26 From monolith to microservices 53:38 Internal tools at Uber 57:05 Helix: Uber’s mobile app rewrite 59:55 Thuan’s email about naming 1:02:03 Org structure changes under 1:06:34 Thuan’s work philosophy 1:12:23 The “three tours of duty” at Uber 1:15:37 Why Thuan left Uber 1:17:34 Coupang and Nubank 1:21:59 Faire 1:25:31 How Faire uses AI 1:28:24 AI’s impact on software engineering 1:31:09 The role of the CTO 1:35:13 Career advice Brought to you by: • Statsig – ⁠ The unified platform for flags, analytics, experiments, and more. • WorkOS – Everything you need to make your app enterprise ready. • Sonar – The makers of SonarQube, the industry standard for automated code review. Check out SonarQube Advanced Security: Three interesting parts from this conversation: 1. The program/platform split came before microservices. The concept of cross-functional “program” teams and dedicated “platform” teams became necessary because an org split across backend, frontend and mobile engineers slowed down in execution speed when Uber grew to around 100 engineers. Every feature required negotiating bandwidth across the mobile, backend, and dispatch teams. Thuan, Travis Kalanick, and Jeff Holden literally used color-coded sticky notes with people’s names to reorganize into self-sufficient teams. We cover more about this split in this The Pragmatic Engineer deepdive, The Platform and Program split at Uber: 2. Expect multiple rewrites during hypergrowth. The right architecture depends on how fast a product and company are growing. At Uber, repeated rewrites were common because each one “bought” another window of survival for the company. Thuan’s recommendation is to understand that a rewrite simply means a company is outrunning its existing architecture: this is not necessarily a bad thing! 3. Uber is the only major company that had a “Senior 1” and “Senior 2” level – and Thuan is unapologetic. Thuan introduced the Senior 1 (L5A) and Senior 2 (L5B) levels because the jump from senior (L5) to Staff (L6) became very big, and larger than between previous levels. One problem this split level created was that Uber’s L5B was akin to Google’s and Facebook’s L6/E6. Thuan resisted the title inflation of just renaming L5B to ‘Staff’.

Gergely Orosz

69,369 views • 4 months ago