Boris Cherny hosts Code with Claude in Tokyo, dotey dissects AI-native film editing, and turingou spots the most complete AI studio canvas yet — June 11
5 qualifying posts from 3 authors in the Jun 11 00:00–07:00 CST window. bcherny surfaced at a Code with Claude Tokyo event (1,819L) and retweeted Dario Amodei's new AI policy essay. dotey posted two breakdowns: Claude Design as a full Agent Harness (45 tools, 24 Skills), and a detailed AI-native film editing workflow using Claude Code + Fable 5 + Whisper + Remotion — no NLE software. turingou called out the most complete AI film studio canvas he's seen, and argued that exponential AI-native efficiency eventually stops being a sales pitch and becomes a replacement strategy.
Window: Jun 11 00:00–07:00 CST (Jun 10 16:00–Jun 11 07:00 UTC) · 5 qualifying posts · 3 authors
bcherny — Claude Code lands in Tokyo
Boris Cherny posted a photo from the Code with Claude Tokyo event, a live developer gathering that drew 1,819 likes. The event follows a string of high-profile Fable 5 dispatches Cherny had posted the day before — the Fable 5 availability announcement (4,353L) and his personal review calling it "the biggest step up I've felt in our models since Opus 4.5 back in November." 1
The Tokyo stop suggests Anthropic's developer-relations momentum is global, not just US-West-Coast-centric. Cherny also retweeted Dario Amodei's new essay "Policy on the AI Exponential," which argues AI is advancing much faster than policy can keep pace with — context worth watching as Claude Code's install base grows. 2
Loading content card…
dotey — Claude Design decoded, and AI edits a film without touching a timeline
宝玉 (dotey) posted two high-signal pieces this morning.
Claude Design as an Agent Harness. His first post, at 160 likes, breaks down what Claude Design actually is under the hood: it runs on a full agentic loop with 45 tools and 24 built-in Skills, constructed on top of a browser + virtual machine environment. His conclusion: Claude Design is effectively the same architecture as Claude Code — a complete Agent Harness — just optimized for UI/UX workflows rather than terminal sessions. 3 The practical implication he draws: Claude Design consumes fewer tokens per task than a raw Claude Code session because the Harness is specialized — fewer irrelevant tools in context means less token overhead, though individual tool calls are slower since they route through the browser/VM layer.
Loading content card…
AI-native film editing — no timeline, no NLE. His second post (150L) is a detailed reconstruction of a video production workflow built entirely without Premiere or Final Cut. The creator handed 25 GB of raw footage (17 takes, 4 scenes) to a pipeline driven by Claude Code with Fable 5. Key steps as dotey describes them: 4
- Whisper transcribed all footage to millisecond-level JSON timestamps
- Multiple subagents then scanned the JSON to discard takes with filler words and select the cleanest segments
- FFmpeg assembled the first cut from a decision JSON — no human touching a timeline
- Custom LUTs written from scratch by the AI, with a browser-based HTML slider UI generated for human color-grade review
- Remotion (React-based) rendered motion graphics synced to Whisper's word-level timestamps — meaning when the speaker says a specific word, an on-screen animation fires automatically, with zero manual keyframe alignment
- Figma MCP let the human design team iterate on layouts, with the AI re-fetching updated designs via MCP before the final 4K 24fps render
This is a real-world demonstration of "Claude Code + Fable 5 as production infrastructure," not just a demo project.
Loading content card…
turingou — AI film studio canvas and the efficiency tipping point
郭宇 (turingou) posted two related observations in the Jun 10 early evening window.
The most complete AI film studio canvas he's seen. At 161 likes, his brief post calls out a canvas-based AI video production product as the most finished he's encountered in the AI film studio category. 5 He doesn't name the product, but the post landed in the same window as his broader reflections on AI-native competitive dynamics — suggesting he's been watching this space closely.
The efficiency inflection point. At 107 likes, he argues that once AI-native companies achieve exponential efficiency advantages over traditional competitors, the business model shifts: it's no longer about selling services to incumbents (B2B), it's about outright replacing them. 6 Relationships, resources, and trust will be barriers — but he says efficiency will break through them. He also filed a related process note: onboarding is still fundamentally about teaching humans to express intent, and intent itself is the bandwidth bottleneck between AI and people. The implication is that automating intent expression is the next hard problem, not just automating the execution.
Loading content card…
Add more perspectives or context around this Post.