Insights · Videos
Worth watching, in plain English.
The talks we learn from, from the people building AI at Y Combinator, Anthropic, Google and more. Each one comes with a short summary: what it says, what it means for your business, and where we've seen it ourselves.
25 talks
No, That's Not a Software FactoryWhy counting AI-written code misses the point, and how WorkOS built its software factory around shipping features instead.
Why Software Factories FailWhy letting AI write code nobody reads breaks down over months, and how planning up front keeps review fast.
What It Actually Takes to Build a Software FactoryA software factory covers the whole cycle, from user feedback to tested release, not just an AI that writes code.
Harness Engineering: How to Build a Software FactoryHow to move from using coding agents to a software factory, by building the checks and loops around them.
OpenClaw Ships Faster Than You Can Read the DiffAn OpenClaw maintainer explains how a small part-time team ships code at enormous speed with many AI agents at once.
AI Architect: Why The Best Engineers Are Solving Code Review BottlenecksAn AI engineer explains how automatic checks, not more human reviewers, keep up with code that AI writes.
AWS Veteran: How The New Software Development Life Cycle WorksA former AWS architect walks through his team's full workflow for building software with AI, from customer talks to review.
The FDE Playbook for AI Startups with Bob McGrewA Palantir veteran explains why sending engineers to sit inside customers has become the way AI agent companies grow.
Andrej Karpathy: Software Is Changing (Again)Karpathy explains how AI changes what software is, and how to work with AI tools that are powerful but fallible.
Your AI Agent Is Hiding Its Mistakes From YouA Warp engineer shows how to have AI grade its own past work to find problems it never reported.
Lead Architect Reveals His Rules for Coding AgentsAn experienced architect explains why he sketches the structure by hand before letting AI agents build anything.
Forward Deployed Engineering 101A short talk on what forward deployed engineers are, why Palantir used them, and whether your company needs them.
How Senior Engineers Use Coding Agents On Prod InfrastructureA HashiCorp engineer explains how he lets AI agents work on live infrastructure while keeping humans in control of risky steps.
How We Took an AI Agent From 41% to 97% AccurateA six-step method for taking a back-office AI agent from barely working to dependable, shown on a real pricing-research job.
How to Build a Self-Improving Company with AIYC partner Tom Blomfield describes a company run as loops where AI notices failures and fixes them overnight.
Your AI Isn't the Advantage. Your Judgment IsWhy companies that gain from AI choose the right problems and redesign work, rather than just adding tools.
Stop Giving Your AI Agents API KeysA hands-on workshop on giving AI agents their own identity and short-lived access instead of a shared password.
Evals Are Broken, Use Them AnywayPublic AI benchmark scores are unreliable, but building your own tests is still the best way to improve an AI agent.
From 46% to 90%: Fine-Tuning Tiny LLMs for On-Device AgentsGoogle shows how very small AI models can run on your phone and become reliable once trained on one narrow job.
Anthropic Rebuilt Git (yes really)Theo reads Anthropic's account of replacing Git with its own system, and explains why AI-scale coding strains tools built for people.
Don't Build Agents, Build Skills InsteadThe creators of Anthropic's Agent Skills explain why packaging know-how in folders beats building a new agent for each job.
These 3 skills changed my lifeTheo walks through three reusable instruction files he gives his AI coding agents and how they cut his review work.
Git is holding us backTheo explains his frustrations with Git and why he is excited about the Zed editor team's plan for a replacement.
Letting AI Interface with your App with MCPKent C. Dodds explains how the Model Context Protocol lets any AI assistant connect to your app through one standard.
How to Build the Best Platforms for Software EngineersTwo platform engineers explain when a business needs an internal developer platform, and how to tell a good one from a bad one.
No talks match. Try another word or channel.