307 - Harness Engineering - the hard part of AI coding

310 - Mitchell Hashimoto on Ghostty & His Agentic Coding Workflow

Mitchell Hashimoto co-founded HashiCorp, built some of the most impressive DevOps tools like Vagrant and Terraform, sold the company to IBM — and then built a terminal. Ghostty is now where a huge chunk of agentic coding actually happens. Mitchell was an AI skeptic. We walk through his six-step adoption framework and the workflows he uses day to day — warm-start research, Hail Mary prompts across twenty GitHub issues, and knowing when to let the agent slam dunk it. Full shownotes at fragmentedpodcast.com [https://fragmentedpodcast.com/episodes/310]. SHOW NOTES * HashiCorp [https://en.wikipedia.org/wiki/HashiCorp] * Vagrant [https://developer.hashicorp.com/vagrant] * Terraform [https://developer.hashicorp.com/terraform] * IBM acquires Hashicorp [https://newsroom.ibm.com/2025-02-27-ibm-completes-acquisition-of-hashicorp,-creates-comprehensive,-end-to-end-hybrid-cloud-platform] GHOSTTY * Ghostty [https://ghostty.org] - Mitchell's fast, native terminal built for platform integration across Mac and Linux * Terminal shell [https://en.wikipedia.org/wiki/Shell_(computing)] * SSH - secure shell [https://en.wikipedia.org/wiki/Secure_Shell] * PTY - pseudoterminals [https://en.wikipedia.org/wiki/Pseudoterminal] * Terminal Multiplexers [https://en.wikipedia.org/wiki/Terminal_multiplexer] * tmux [https://en.wikipedia.org/wiki/Tmux] - most popular open source one * XTGETTCAP [https://invisible-island.net/xterm/ctlseqs/ctlseqs.html#h3-Device-Control-functions] by xterm [https://en.wikipedia.org/wiki/Xterm] * libghostty [https://github.com/ghostty-org/ghostty] - the cross-platform terminal emulation library that powers Ghostty's core * xterm-js [https://xtermjs.org/] - powers terminal for apps like VSCode and the cloud * Jedi Term [https://github.com/JetBrains/jediterm] - Intellij's embedded terminal * Ghostty is now a non-profit [https://mitchellh.com/writing/ghostty-non-profit] * cmux [https://github.com/manaflow-ai/cmux] - native macOS terminal multiplexer built on libghostty — a fork Mitchell champions * Free Software Definition [https://en.wikipedia.org/wiki/The_Free_Software_Definition] - the 4 essential freedoms 1. The freedom to run the program as you wish, for any purpose. 2. The freedom to study how the program works, and change it to make it do what you wish. 3. The freedom to redistribute copies so you can help others. 4. The freedom to distribute copies of your modified versions to others. * Mitchell's tweet on unsolicited PRs and transfer of ownership [https://x.com/mitchellh/status/1986933060643045646] THE AI ADOPTION JOURNEY * My AI Adoption Journey [https://mitchellh.com/writing/my-ai-adoption-journey] - Mitchell's blog post outlining his five-step framework * Step 1: Drop the Chatbot * Episode 301 - AI Coding ladder - Different stages of AI adoption * Step 2: Reproduce Your Own Work * Step 3: End-of-Day Agents * OpenAI Deep Research [https://openai.com/index/introducing-deep-research/] - kick off research tasks for a "warm start" the next morning * Spine AI research [https://www.getspine.ai/] - deep research tool for longer, hour-long analysis tasks * Step 4: Outsource the Slam Dunks * Claude status hooks - warcraft peons [https://x.com/kaushikgopal/status/2023919843016348027?s=20] * Conductor [https://www.conductor.build/] * Step 5: Engineer the Harness * Episode 307 - Harness Engineering - Fragmented's deep dive on harness engineering, heavily inspired by Mitchell's post * Step 6: Always have an Agent running * Peter Steinberger [https://en.wikipedia.org/wiki/Peter_Steinberger_(programmer)] * Codex plugin for Claude Code [https://github.com/openai/codex-plugin-cc] GET IN TOUCH We'd love to hear from you. Email [contact@fragmentedpodcast.com] 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 [https://fragmentedpodcast.com/contact] * Newsletter [https://buttondown.com/fragmentedcast] * Youtube [https://www.youtube.com/@fragmentedpodcast] * Website [https://fragmentedpodcast.com] CO-HOSTS: * Kaushik Gopal [https://kau.sh/] * Iury Souza [https://iurysouza.dev/] > [!fyi] We transitioned from Android development to AI starting with > Ep. #300. Listen to that episode for the full story behind > > our new direction.

309 - Background Agents

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 [https://www.youtube.com/watch?v=kwSVtQ7dziU] - NoPriors podcast — maximize agent runtime, not token burn * Cursor Agent Mode - Multiagent interface [https://cursor.com/changelog/2-0#multi-agents] - introduced the multi-agent board as a new paradigm for local parallel agents * Google Antigravity - Agent Manager interface [https://antigravity.google/docs/agent-manager] * Claude Code Agent Teams [https://code.claude.com/docs/en/agent-teams] - spawn sub-agents from a main orchestrator, with tmux pane integration * Git worktrees [https://www.reddit.com/r/ProgrammerTIL/comments/mtjg0c/git_til_about_git_worktrees/] - /reddit REMOTE BACKGROUND AGENTS IN THE CLOUD * Google Jules [https://jules.google.com/] - hosted GitHub-connected agent, proposes a plan, edits code, runs tests, opens a PR * Cursor Cloud Agents [https://cursor.com/blog/cloud-agents] - remote agents that clone your repo in the cloud and work in parallel * OpenAI Codex [https://openai.com/blog/introducing-codex/] - cloud software engineering agent for parallel tasks * Claude Code on the web [https://claude.ai/code] - 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 [contact@fragmentedpodcast.com] 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 [https://fragmentedpodcast.com/contact] * Newsletter [https://buttondown.com/fragmentedcast] * Youtube [https://www.youtube.com/@fragmentedpodcast] * Website [https://fragmentedpodcast.com] CO-HOSTS: * Kaushik Gopal [https://kau.sh/] * Iury Souza [https://iurysouza.dev/] > [!fyi] We transitioned from Android development to AI starting with > Ep. #300. Listen to that episode for the full story behind > > our new direction.

308 - How Image Diffusion Models Work - the 20 minute explainer

You already know how LLMs work from our popular 20-minute explainer. Now we take it to images. What does Michelangelo have to do with stable diffusion? More than you'd think. Walk away knowing how image generation actually works — and what it has in common with the text models you already understand. Full shownotes at fragmentedpodcast.com [https://fragmentedpodcast.com/episodes/308]. SHOW NOTES * Episode 303 - How LLMs work in 20 minutes - text generation * VAE - Variational Autoencoder [https://www.ibm.com/think/topics/variational-autoencoder] * RGB Color model [https://en.wikipedia.org/wiki/RGB_color_model] - wikipedia * Word2Vec technique [https://en.wikipedia.org/wiki/Word2vec] - wikipedia * Efficient Estimation of Word Representation [https://arxiv.org/abs/1301.3781] - original Word2Vec paper by Mikolov et al. * High-Resolution Image Synthesis with Latent Diffusion Models [https://arxiv.org/abs/2112.10752] - Rombach et al. (2022) — the paper behind Stable Diffusion * Image Training data * LAION-5B [https://arxiv.org/abs/2210.08402] - 5 billion image-text pairs scraped from the web, used to train many image generation models * WebLI [https://www.aibase.com/news/15390] - Google's internal image-text dataset * Michelangelo [https://www.goodreads.com/quotes/1191114-the-sculpture-is-already-complete-within-the-marble-block-before] GET IN TOUCH We'd love to hear from you. Email [contact@fragmentedpodcast.com] 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 [https://fragmentedpodcast.com/contact] * Newsletter [https://buttondown.com/fragmentedcast] * Youtube [https://www.youtube.com/@fragmentedpodcast] * Website [https://fragmentedpodcast.com] CO-HOSTS: * Kaushik Gopal [https://kau.sh/] * Iury Souza [https://iurysouza.dev/] > [!fyi] We transitioned from Android development to AI starting with > Ep. #300. Listen to that episode for the full story behind > > our new direction.

307 - Harness Engineering - the hard part of AI coding

The hard part of AI coding isn't generating code — it's controlling quality, safety, and drift. Kaushik and Iury break down harness engineering: the five pillars for shaping an agent's environment and what it looks like when teams build custom harnesses from scratch. Full shownotes at fragmentedpodcast.com [https://fragmentedpodcast.com/episodes/307]. SHOW NOTES WHY IT MATTERS * Harness Engineering [https://openai.com/index/harness-engineering/] - OpenAI's post on building their Codex codebase (~1M lines of code, 1,500 PRs merged, zero manually written) SHAPING THE HARNESS * The Feed's Lost and Found [https://iurysouza.dev/newsletter/the-feeds-lost-and-found-01/] - Iury's newsletter consolidating harness engineering themes 1. Agent legibility 2. Closed feedback loops 3. Persistent memory 4. Entropy control 5. Blast radius controls BUILDING THE HARNESS * Minions: Stripe's one-shot, end-to-end coding agents [https://stripe.dev/blog/minions-stripes-one-shot-end-to-end-coding-agents] - Stripe forked Goose to build custom agents for their codebase * Goose [https://github.com/block/goose] - open-source coding agent from Block * Superpowers [https://github.com/obra/superpowers] by Jesse Vincent - skills that enforce a proper software engineering process * Open Code [https://opencode.ai/] - open-source coding agent you can fork and customize OTHER RESOURCES * Agent Harness Glossary [https://latentpatterns.com/glossary/agent-harness] - Latent Patterns * Towards self-driving codebases [https://cursor.com/blog/self-driving-codebases] - Cursor * Agentic Workflows [https://githubnext.com/projects/agentic-workflows] - GitHub Next * Future of Software Development [https://www.thoughtworks.com/content/dam/thoughtworks/documents/report/tw_future%20_of_software_development_retreat_%20key_takeaways.pdf] - ThoughtWorks GET IN TOUCH We'd love to hear from you. Email [contact@fragmentedpodcast.com] 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 [https://fragmentedpodcast.com/contact] * Newsletter [https://buttondown.com/fragmentedcast] * Youtube [https://www.youtube.com/@fragmentedpodcast] * Website [https://fragmentedpodcast.com] CO-HOSTS: * Kaushik Gopal [https://kau.sh/] * Iury Souza [https://iurysouza.dev/] > [!fyi] We transitioned from Android development to AI starting with > Ep. #300. Listen to that episode for the full story behind > > our new direction.

306 - Keeping your agent instructions in sync and effective

AGENTS.md is becoming the common language for AI coding tools, but keeping repo rules, personal rules, and tool-specific files in sync is still messy. In this episode, Kaushik and Iury break down the sync problem, compare their own setups, and unpack what the latest AGENTS.md research actually says. Full shownotes at fragmentedpodcast.com [https://fragmentedpodcast.com/episodes/306]. SHOW NOTES THE SYNC PROBLEM * AGENTS.md [https://agents.md] - Official spec * Custom instructions with AGENTS.md [https://developers.openai.com/codex/guides/agents-md/] - Open AI * Keep your AGENTS.md in sync [https://kau.sh/blog/agents-md/] - Kaushik's post * Rulesync [https://github.com/jpcaparas/rulesync] - What Iury uses * Tweet by Ryan Carson and Claude frustrations [https://x.com/ryancarson/status/2024144258702143782] OTHER LINKS * Evaluating AGENTS.md: Are Repository-Level Context Files Helpful for Coding Agents? [https://arxiv.org/abs/2602.11988] * Harness engineering - Check the section about using AGENTS.md as a table of contents [https://openai.com/index/harness-engineering/] * OpenCode [https://opencode.ai/] GET IN TOUCH We'd love to hear from you. Email [contact@fragmentedpodcast.com] 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 [https://fragmentedpodcast.com/contact] * Newsletter [https://buttondown.com/fragmentedcast] * Youtube [https://www.youtube.com/@fragmentedpodcast] * Website [https://fragmentedpodcast.com] CO-HOSTS: * Kaushik Gopal [https://kau.sh/] * Iury Souza [https://iurysouza.dev/] > [!fyi] We transitioned from Android development to AI starting with > Ep. #300. Listen to that episode for the full story behind > > our new direction.