Hey Everyone 👋,
John Lindquist here with the 42nd issue of AI Dev Essentials!
Last week was about Fable 5 landing. This week is about the aftermath of Governments stepping in to take away our favorite new toy.
Anthropic had to disable Fable 5 and Mythos 5 after a U.S. government export-control directive. OpenAI gave Codex controlled Chrome DevTools Protocol access. Unreal Engine 5.8 shipped experimental MCP support. Claude Design started syncing with Claude Code. Cursor teased Origin, a waitlist-stage git forge aimed at agent-era workflows.
That Fable shutdown hit harder than I expected. For three days it felt like we all had self-driving cars, then suddenly we were back on manual, remembering how easy it was every time we changed gears. It was a strong reminder to me how unpredictable the AI landscape is, but I never really expected it to move backwards.
Get your ticket for the next Codex Power User Workshop. Standard tickets are $450, and yearly egghead members can register for $350.
⚡ Codex Power User Workshop — 📅 Fri June 26, 9am PT Regular (no membership required): $450 — Get Regular Ticket Pro Yearly Member (egghead.io yearly subscription): $350 (~~$450~~ regular price), you save $100 — Get Pro Ticket
🎤 Speaking Events
Check out my recent/upcoming speaking engagements:
O'Reilly: This Week in AI — I joined YK Sugi on O'Reilly Radar to unpack Fable 5, the token-cost problem around agentic coding loops, and why throwing agents at a legacy codebase without the right environment is such an easy way to burn budget. Read/watch the episode recap →
AI ❤️ Monorepos Conf — On June 23, I'm giving The Agentic Power User's Playbook, a practical talk about running swarms of agents in parallel: tiled terminals, fast-model routing, and the muscle memory that makes multi-agent work feel less chaotic. Register free →
AI Engineer World's Fair — At AI Engineer World's Fair, I'm teaching The Agentic Power User's Playbook: Tips and Tricks for Swarm-Style Agentic Development. This is the more hands-on version: how to make agents cooperate, verify each other, and stay useful across real projects. See the schedule →
🚀 Major Announcements
Policy Anthropic Disables Fable 5 and Mythos 5 After U.S. Export Directive
Anthropic said on June 12, 2026 that it received a U.S. government export-control directive requiring the company to suspend access to Fable 5 and Mythos 5 by foreign nationals, including foreign-national Anthropic employees. Anthropic says it could not reliably separate that access in real time, so it disabled both models for all customers while leaving other Claude models unaffected.
What Changed:
- Fable 5 and Mythos 5 were disabled globally for customers
- Access to other Claude models was not affected
- Anthropic says the order applies to foreign nationals inside and outside the United States
- Anthropic says the government's concern involved a narrow potential jailbreak, not a disclosed universal jailbreak
- The incident followed the Fable 5 launch and the earlier backlash over invisible safeguard behavior
- The Verge reports that experts see this as an unsettled and possibly unprecedented use of export controls for access to a hosted AI model
(Anthropic — Fable and Mythos Access Statement, The Verge — Export Controls Coverage)
💬 This one hurt. I loved the progress I was making with Fable. I had it build my own custom "Brain" system (like an OpenClaw/Hermes, but the way I wanted it). I was making incredible progress on the next version of Script Kit which has been an extremely difficult project for other models since it leverages the GPUI framework from Zed. It felt like 3 pure days of progress. I've been building my custom version of "Fusion" (a popular "council" concept that OpenRouter mainstreamed). It's something I'd done in the past, but updated it to the latest models: Opus 4.8, GPT 5.5, GLM 5.2, and Gemini 3.5. It's working really well and achieving near-Fable levels (based on vibes, I don't have the budget for evals), but definitely will keep tweaking it. I'll be teaching the setup in my next Codex Power User workshop for sure.
Tool Codex Developer Mode Gets Chrome DevTools Protocol Access
OpenAI added Developer mode for Browser use in Chrome and the Codex in-app browser, giving Codex controlled access to Chrome DevTools Protocol for deeper browser debugging. Codex can inspect console output, network traffic, runtime errors, DOM state, applied styles, and JavaScript performance when the feature is enabled.
Developer Notes:
- Works with Browser use in Chrome and the Codex in-app browser
- Full CDP access is off by default and can be enabled from Codex app settings
- Codex asks for explicit approval before using full CDP access to inspect a website
- Organizations can restrict full CDP access through managed configuration
- The feature is designed for debugging live browser issues, not just clicking through visible UI
- Pairs naturally with Codex browser tasks, app previews, and web app verification loops
(OpenAI Codex — In-App Browser, OpenAI Codex — Changelog, OpenAI Codex — App Settings)
💬 I love this workflow. I mainly use Agent Browser, but glad to see this becoming the standard. Whenever I work in a non-web project, one of my first prompts is: "What dev tools can we build to give us Chrome's CDP-like capabilities."
Web Chrome's WebMCP Origin Trial Gives Websites Agent Tools
Chrome opened the WebMCP origin trial, letting web apps expose structured tools to agents through imperative JavaScript APIs or declarative form annotations. The goal is to help agents complete tasks from explicit tool definitions instead of inferring intent from pixels and labels.
Developer Notes:
- Origin trial is available in Chrome 149
- Web apps can expose actions as structured tools
- Supports JavaScript APIs and declarative HTML form annotations
- Local testing is available through
chrome://flags/#enable-webmcp-testing - Current implementation requires an open browser tab or webview
(Chrome for Developers — WebMCP Origin Trial, Chrome for Developers — WebMCP Docs)
💬 I have too many ideas for this. Definitely going to record lessons about it as I explore it.
Game Dev Unreal Engine 5.8 Ships Experimental MCP Server Support
Epic released Unreal Engine 5.8 with an experimental Model Context Protocol plugin that lets MCP-compatible agents like Claude Code, Cursor, Codex, VS Code, Gemini, and MCP Inspector drive Unreal Editor workflows over a local connection.
What Developers Get:
- Experimental MCP plugin for Unreal Engine 5.8
- Direct agent connection to project context and Unreal workflows
- The docs call this experimental, incomplete in places, and subject to API and data-format changes
- Part of a larger 5.8 release focused on performance, content creation, animation, and virtual production
- Epic describes this as a way for models to become active collaborators inside the editor
💬 Can't wait to see what the next generation of game devs can make when they can build with natural language.
🛠️ Developer Tooling Updates
Tool Claude Design Adds Claude Code Sync and Design-System Workflows
Anthropic's Claude Design support docs now describe Claude Code workflows, including /design-sync for syncing a design system from Claude Code and /design for starting design work directly from Claude Code. Claude Design remains in beta, with support for design systems, direct edits, comments, and handoff into implementation workflows.
Developer Notes:
/design-syncsyncs a design system from Claude Code/designstarts Claude Design work from Claude Code- Claude Design can package handoff bundles for Claude Code implementation
- Support docs call out large-codebase considerations and known beta limitations
- Enterprise availability is beta and default-off, and the docs still call out workflow edges around large codebases
(Claude Support — Get Started with Claude Design, Anthropic — Claude Design)
💬 I have a big design project coming up next week. Definitely going to be using this and I'll report back.
Model Update Z.ai Releases GLM-5.2 for Long-Horizon Coding Tasks
Z.ai released GLM-5.2, a flagship model positioned around long-horizon tasks, project-scale engineering context, and 1M-token context windows. Z.ai says GLM-5.2 is built for stable execution across full development workflows, from requirements to deployable products.
What Stands Out:
- 1M-token context aimed at project-scale engineering work
- Focus on long-horizon coding and agentic engineering tasks
- Two thinking-effort levels in the developer docs
- Coding-plan guides for Claude Code, Cursor, Cline, Trae, Kilo Code, and more
- Z.ai's own benchmark claims are strong, but I would still treat this as vendor positioning until independent evals settle
(Z.ai — GLM-5.2, Z.ai Docs — GLM-5.2, Z.ai — Subscribe)
💬 Z.ai is cooking. I've only used GLM 5.2 a tiny bit, but I can definitely say the model gap between open and closed models is closing.
Open Source OpenAI Joins the Rust Foundation as a Platinum Member
The Rust Foundation announced that OpenAI joined as a Platinum Member and made a $600,000 total contribution supporting the Rust Project and broader Rust ecosystem. The Foundation says the funding will support maintainers, technical priorities, Project Goals, the Rust Foundation Maintainers Fund, and the Rust Innovation Lab.
Why It Matters:
- OpenAI is publicly backing Rust's maintainer ecosystem and technical priorities
- Predrag Gruevski, maintainer of
cargo-semver-checks, will join the Rust Foundation Board as OpenAI's Platinum Member representative - This is direct support for maintainer infrastructure, not just a sponsorship logo
(Rust Foundation — OpenAI Platinum Member, Rust Foundation — On OpenAI's Support for Rust)
💬 Love this. My main personal projects are all now built in Rust. Thanks OpenAI.
Platform Cursor Teases Origin, a Git Forge for Agent-Era Workflows
Cursor has opened a waitlist for Origin, which it describes as "a git forge for the agentic era." The public page is still light on product detail, so I would treat this as a strategic signal rather than a fully specified GitHub alternative.
Why It Matters:
- Cursor is signaling interest beyond the editor and into the code-hosting layer
- The public page frames Origin around agent-era infrastructure, but does not yet spell out review, collaboration, or migration details
- Keep this as a watch item unless Cursor publishes deeper docs or launch notes
💬 Git and GitHub are so engrained into my processes, it will be difficult to try something else. I'm really curious how tied to the Cursor ecosystem it will be and how much they'll be able to open-source.
🤖 AI Ecosystem Updates
Ecosystem Midjourney Announces Midjourney Medical
Midjourney now lists "Medical" among its projects, and The Verge reports that David Holz showed a proposed ultrasound-based full-body scanner developed with Butterfly Network modules. The reported initial focus is body-composition mapping rather than diagnostic imaging.
What To Watch:
- Midjourney appears to be exploring medical imaging hardware beyond its image-generation business
- The reported scanner concept uses ultrasound modules around a water-filled platform
- The company is not positioning the first version as a diagnostic medical device
- Privacy, regulation, FDA classification, and data access will matter more than the demo
(Midjourney, The Verge — Midjourney Medical)
💬 Um, what? How am I supposed to react to this? I'm no medical expert and have no clue what to say. I do love MidJourney's image design taste though...
⚡ Quick Updates
Tool Codex Mobile Adds Profile Usage Stats
- Released: June 9, 2026
- OpenAI added a Codex profile screen to ChatGPT for iOS with usage stats and token activity charts.
✨ Workshop Spotlight
Codex Power User Workshop
Next Friday, seats are still available
- Context Packaging: Bundle files, logs, screenshots, and traces so Codex understands the real problem
- Terminal Workflows: Run parallel research, planner, builder, and validator sessions without losing the thread
- Agent Profiles: Compare broad defaults vs. narrow prompts, focused skills, and limited tool access
- Tools & Sandboxing: Separate read-only from mutation tools and map agents to repeatable tool profiles
- Skills, Hooks & Memory: Build reusable expertise, enforce guardrails, and capture durable decisions
- Codex SDK: Embed Codex into CI, dashboards, and internal tools when terminal sessions aren't enough
Friday, Jun 26 @ 9:00 AM to 2:00 PM Pacific
Regular (no membership required): $450 — Get Regular Ticket
Pro Yearly Member (egghead.io yearly subscription): $350 (~~$450~~ regular price), you save $100 — Get Pro Ticket
Did you enjoy this issue? Share it with a friend.
© 2026 John Lindquist • egghead.io