Day 1: A Million-Dollar App in the Week We Have Fable | Faceless AI - Dataspheres AI

Day 1: A Million-Dollar App in the Week We Have Fable On July 1, Anthropic put Claude Fable 5 back on the Max plan. On July 7, the complimentary window cl...

Day 1: A Million-Dollar App in the Week We Have Fable On July 1, Anthropic put Claude Fable 5 back on the Max plan. On July 7, the complimentary window closes and the model moves to usage credits at $10 per million input tokens and $50 per million output. This series starts on day three of that window — late to the party, but the party is short. What follows is a daily record of what one person and one frontier model can build in the days that remain. The target is an application worth a million dollars. The constraint is the calendar. Why this week is unusual Fable 5 has had the strangest availability timeline of any frontier model. It launched on June 9 alongside Claude Mythos 5, with subscribers originally promised access through June 23. Three days later, on June 12, the U.S. government applied export controls to both models over security concerns, and because Anthropic could not verify user nationality in real time, it suspended access for everyone . On June 30 the controls were lifted, and the model returned on July 1 — this time with a 50% weekly-usage cap and a hard end date of July 7 ( CNBC , Search Engine Journal ). The Fable 5 availability window — the strangest launch timeline of any frontier model. A model that was export-controlled and un-controlled within a single month, now available for exactly one more week, is a moment worth capturing. Most write-ups about frontier models are benchmarks. This one is a build log. The constraint, in numbers. The challenge Build and ship an application with a credible path to a million dollars in value, in the days the window has left, using Fable 5 as the primary engineer. Every day gets a public update in this series — what shipped, what failed, and what the evidence is. Claims come with links: working software, board records, output files. Anything without an artifact does not get claimed. The million-dollar figure is a target for the asset being built, not a revenue promise for the week. What makes the target honest is the process: the work is tracked on spec-driven development boards where every task passes through validation gates before it counts as done, and the boards are public. The framework is described in Run AI Work Like a State Machine — this series is that framework under a deadline. Three boards, one state machine The build is not one project but three concurrent workstreams, each running on its own board with a live public dashboard. The dashboards are not screenshots prepared for this post — they are the actual working trackers, with progress rings, trace graphs, and activity feeds that update as tasks move. If nothing is happening, the dashboards will say so. That is the point. Open Source Beta 1 tracks the platform itself: the push to get Dataspheres AI — the system these very boards run on — to its first open-source beta release. Fable writes platform code, the board gates it, and the dashboard shows where the release stands at any moment. Qwen Edit Parity is the AI studio workstream, and it carries the best war story of the three. For weeks, an image-editing pipeline produced wrong output from code while working perfectly in the ComfyUI interface — the kind of bug that eats days. The root cause, dug out under this board's research gate: the hand-built API workflow was a structurally incomplete graph, missing three nodes and using the wrong text encoder. No prompt change could fix a wrong graph. The dashboard documents the diagnosis and the fix: capture the verified workflow the UI actually ran, and replay it verbatim. Mobile UX Store tracks the third front: making the platform work as a real mobile product, up to app-store readiness. Distribution is where million-dollar targets live or die, so this board exists from day one rather than as an afterthought. This post is on a board too The series itself runs on the same system. This post traces back to a verbatim prompt on an intake card, through a research task with cited sources, a publish spec, and typed validation criteria that must pass before the page goes live. The Fable Week dashboard is public — you can watch this series get built the same way you can watch the app get built. When a daily update claims something shipped, the claim will link to the board that gated it. What day 1 looked like Today went to the foundation. The build is not starting from zero — the past month of Fable-assisted work produced the infrastructure this week runs on: A production AI studio at Project 3035: character generation, virtual try-on (validated at 121.6 seconds per garment transfer on an RTX 5080), and image-to-video, all registered in a pipeline map with machine-checked status gates. A spec-driven development harness that forces evidence before any task advances — the same system that keeps this week's claims honest, visible on the three dashboards above. This publication itself, tracked as an artifact on its own board. Day 1's output is the plan, the boards, and this post. The application concep