Show Notes
Brian Grinstead is a distinguished engineer at Mozilla, where he’s worked on Firefox and the web platform since 2013 (he joined to help launch Firefox DevTools). Recently he and his team pointed an agentic bug-finding pipeline at Firefox—a codebase with tens of thousands of files and tens of millions of lines of code—and shipped a record month of security fixes. The viral chart everyone saw gave the credit to Anthropic’s new Mythos model. Brian’s take is that the harness and pipeline did just as much of the work, and he walks through exactly how it runs and how anyone can build a starter version.
What you’ll learn: How to build a basic bug-finding harness by running Claude Code or Codex with one prompt and the -p flag, no SDK required Why pointing an agent at a whole codebase fails, and how an LLM judge can score and rank files before you spend any compute How a verifier subagent kills false positives by catching the agent when it cheats The goal-loop pattern: give an agent a tightly scoped problem, a clear pass/fail signal, and let it retry far past the point a human would quit Why teams that already invested in fuzzing, CI, and dev tooling are so far ahead How to weigh model versus harness, and why Brian splits the credit close to 50-50 How a non-engineer can reuse the same score, verify, and fix the loop for design quality, conversion rate, or tech debt Why AI-generated patches still can’t ship on their own, and where humans stay in the loop — Brought to you by: WorkOS—Make your app enterprise-ready today Metaview—The agentic recruiting platform for winning teams — In this episode, we cover: (00:00) Introduction to Brian Grinstead (02:43) The viral chart: Firefox Security Bug Fixes by Month (05:32) How the custom harness works (10:22) Goal loops and guardrails (14:45) How they built it (16:55) Real bugs, including a 15-year-old one (23:00) Open-sourcing it (26:26) Why humans still review every fix (32:30) Live demo and prioritizing files (40:18) Mobilizing the team and recap (42:33) Lightning round — Tools referenced: • Claude Code: https://claude.ai/code • Claude Agent SDK: https://code.claude.com/docs/en/agent-sdk/overview • Codex: https://openai.com/index/openai-codex/ • OpenAI Agent SDK: https://developers.openai.com/api/docs/guides/agents • VS Code: https://code.visualstudio.com/ • Docker: https://www.docker.com/ • Firefox: https://www.mozilla.org/firefox/ • Address Sanitizer: https://github.com/google/sanitizers • RLBox: https://rlbox.dev/ — Other references: • Mozilla Bug Bounty Program: https://www.mozilla.org/security/bug-bounty/ • Mozilla GitHub: https://github.com/mozilla — Where to find Brian Grinstead: LinkedIn: https://www.linkedin.com/in/bgrins/ GitHub: https://github.com/bgrins — Where to find Claire Vo: ChatPRD: https://www.chatprd.ai/ Website: https://clairevo.com/ LinkedIn: https://www.linkedin.com/in/clairevo/ X: https://x.com/clairevo — Production and marketing by https://penname.co/. For inquiries about sponsoring the podcast, email [email protected].
More from How I AI
Anthropic's new personal agent, Claude Duet, lives on your phone, inherits all your Claude plugins and connectors, and is pitched at small business owners and busy professionals. It's the most polished agent launch I've seen: thoughtful onboarding, task lists, first-class artifacts, and Slack-style channels. But polish isn't enough when you're competing with every other personal agent for attention. I put it to work on my inbox and Slack, my sales pipeline, a Google Docs sales deck, a launch video, and finding Warriors tickets for my kids, and I'll walk you through what it nails, where it stalls, and what Anthropic needs to fix before it can become anyone's number one agent.
Brooke Lacey has been making videos since before Facebook had video, and she built a TikTok audience of more than 290,000 by learning to cut short-form videos herself. With over 25 years in tech, she now builds her own apps, including Zosha, a free mobile app for learning AI. In this episode, Brooke walks me through the Claude Code, FFmpeg, and Whisper workflow that edits her TikToks in about 15 seconds, how she mines her comments for new scripts, and how she runs a weekly short-form operation with no editor.
I put Meta’s new Muse Wonder (and Wonder Full) frontier models through the How I AI bench, which covers nearly 20 tasks across prototyping, agentic workflows, coding, writing, and creative work. I scored everything with an LLM-as-a-judge pass and my own voice notes, then pulled together the first-look results you see here.
Alex Cornell is a design director at Meta, where he leads design on Muse, Meta’s consumer AI agent. He came to the role with a long background in software product design, including time as a founder, and has spent the past year building one of the most thoughtful agent interfaces in the consumer AI space. Muse ships with a custom avatar, a three-layer status display, a personalized ideas surface, and a generative feed. Alex has been sharing the reasoning behind each of those design decisions publicly on Threads.
Kath Korevec is a member of the Product staff at OpenAI working on Codex, and she spent over a year building and using ChatGPT Sites internally before its public launch. She’s been on the front lines of shipping Plugin Insights, MCP plugin hosting, and the connector ecosystem, which now includes around 60 integrations.