Fragmented - AI Developer Podcast

Andrej Karpathy says the goal is to maximize how long an agent runs without your intervention. But there's a false summit most teams hit first: individual speed goes up while system speed stalls, your laptop roars under four parallel Gradle builds, and review queues back up. Kaushik and Iury trace the full arc — from local multitasking to cloud-hosted async work to fully autonomous agents that fire on repo events and put PRs in your inbox. Show Notes Andrej Karpathy on agents and token throughput - NoPriors podcast — maximize agent runtime, not token burn Cursor Agent Mode - Multiagent interface - introduced the multi-agent board as a new paradigm for local parallel agents Google Antigravity - Agent Manager interface Claude Code Agent Teams - spawn sub-agents from a main orchestrator, with tmux pane integration Git worktrees - /reddit Remote Background Agents in the cloud Google Jules - hosted GitHub-connected agent, proposes a plan, edits code, runs tests, opens a PR Cursor Cloud Agents - remote agents that clone your repo in the cloud and work in parallel OpenAI Codex - cloud software engineering agent for parallel tasks Claude Code on the web - cloud-hosted Claude Code sessions decoupled from your local machine Building trust Episode 307 - Harness Engineering - the earlier episode on shaping agent environments — and why this ceiling exists Get in touch We'd love to hear from you. Email is the best way to reach us or you can check our contact page for other ways. We want to hear all the feedback: what's working, what's not, topics you'd like to hear more on. Contact us Newsletter Youtube Website Co-hosts: Kaushik Gopal Iury Souza [!fyi] We transitioned from Android development to AI starting withEp. #300. Listen to that episode for the full story behind our new direction.

A lively 20-minute walk through how image diffusion models generate pictures from noise. They cover why raw pixels fail, how VAEs create compact latent spaces, and how interpolation on latents blends visuals. Learn why adding and then removing noise is the core trick, with a sculptor analogy linking Michelangelo to Stable Diffusion.

They dig into harness engineering: the five pillars for shaping an agent’s environment. Topics include making repos agent-friendly, closing feedback loops with tests and logs, and using persistent memory to avoid repeated mistakes. They cover entropy control to prevent codebase drift and blast radius controls and permissioning to limit risky agent actions.

They tackle how AGENTS.md emerged as a common language for AI coding tools and the messy reality of keeping repo, personal, and tool rules synchronized. They compare different sync workflows like RuleSync, dotfiles, and symlink scripts. They also review recent research on auto-generated versus human-written AGENTS.md and when explicit repo-level instructions actually help.

Clear definitions of what an AI agent is and how its reactive loop works. A deep dive into modes versus subagents and how mode switching alters system behavior. An RTS-style scout analogy shows how parallel subagents gather intel without polluting context. Practical guidance on when to spawn isolated, scoped workers and when to avoid them.

They explain when to reach for agent skills versus simpler commands and why skills reduce repetition across teams. You’ll hear how skills are organized as folders with markdown, scripts, and assets. They unpack progressive disclosure and context forks to keep agents lean. They also cover installing skills, tool selection, and important security warnings about untrusted skill code.

A clear walkthrough of tokenization and how text becomes machine-friendly tokens. A tour of embeddings and analogies that show how meaning is encoded as vectors. A plain explanation of pretraining as the way models learn semantic maps. A practical look at inference, sampling methods, context windows, and cost tradeoffs.

They unpack what Model Context Protocols are and how they let models plug into external tools and data. They compare MCPs with CLIs and explain server and client roles. They cover real uses like GitHub PR review, Playwright browser control, and instant library docs. They also warn about token bloat, confusion from many tools, and share practical tips for efficient tool use.

Dive into the fascinating world of AI coding paradigms with insights on super autocomplete tools like GitHub Copilot and the evolution to chat-oriented programming. Discover how agents and innovative prompts enhance coding autonomy, including planning and deeper reasoning techniques. Explore the future of orchestrating multiple agents and the challenges that come with it. Highlights also include practical tips for working with AI, emphasizing the importance of clear communication with your virtual assistants. This discussion is a must for developers looking to stay ahead!

Fragmented is changing. New direction, new cohost. Kaushik explains the pivotfrom Android to AI development and introduces Iury Souza.From vibe coding to software engineering — one episode at a time.Full shownotes at fragmentedpodcast.com.Contact usNewsletterWebsiteContact usYoutubeCo-hosts:Kaushik GopalIury Souza