How to use Unknown to rewrite a platform with AI
The author reports a two-month rewrite with fewer bugs, strong performance and easier feature development, while describing difficult reviews and substantial overtime.
news.ycombinator.comPrompt to try
Editorial adaptation, not the source author’s exact prompt. Replace bracketed inputs and review tool permissions first.
Plan a rewrite of [existing platform] into [target API] and [target frontend]. Propose core architecture changes and [quality invariants]. Implement with meaningful tests and deterministic checks, report review risks, and track architectural improvements in [issue tracker].
Reach people exploring AI agents
See placements →What happened
The author reports a two-month rewrite with fewer bugs, strong performance and easier feature development, while describing difficult reviews and substantial overtime.
Reported by enraged_camel on Hacker NewsHow the agent did it
- Replace the old platform with separate API and frontend codebases.
- Re-architect core functionality and remove accumulated technical debt.
- Enforce coverage, test quality and deterministic checks.
Where a person stepped in
The team invested heavily in code review, trained colleagues on AI development, and held weekly reviews of fixed bugs.
A person approved each step before the agent continued.
The author describes a complete workplace platform rewrite that replaced a Phoenix LiveView codebase with separate Phoenix API and Vue with TypeScript codebases. No code was shared with the abandoned application. The team used the opportunity to redesign core functionality and remove technical debt accumulated over six years.
They report finishing in two months, with a little over 700,000 lines of code, and attribute 95% of planning, orchestration, and writing to AI. Quality controls included a requirement for full test coverage on both client and server, AI review rules requiring non-vacuous tests, and deterministic checks enforcing invariants, formatting, and other constraints. Weekly meetings examined fixed bugs and considered checks to prevent recurrence.
A dedicated Fable agent managed an Architecture epic, tracking initiatives such as accessibility, shared helpers, and common patterns. The author reports fewer bugs, strong performance, and easier feature development despite the combined codebases being roughly 35% larger. They also describe difficult reviews, team training, and long hours under an external deadline. These are the author's reported results, rather than independently verified measurements.
Who it's for: Teams undertaking a substantial rewrite with capacity for rigorous review and enforceable quality requirements.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Hacker News · enraged_camel
Frequently asked questions
What stack replaced the old platform?
The team replaced Phoenix LiveView with two new codebases: a Phoenix API and a Vue frontend using TypeScript.
How did the team address code quality?
They required full client and server test coverage, non-vacuous tests, deterministic invariant checks, and weekly review of fixed bugs to identify additional preventive checks.
More details
Remaining steps
- Review fixed bugs weekly for additional preventive checks.
About this account
The author attributes 95% of planning, orchestration and code writing to AI, with a Fable agent tracking architecture initiatives.
Platform: Hacker News. Attribution: https://news.ycombinator.com/item?id=49679798. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: enraged_camel on Hacker News
Source published 13 Sept 2026
Source checked 14 Oct 2026
Tools and topics
Fable · Phoenix · Vue · TypeScript · Deterministic checks
Platform rewrite · Code quality
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.