Build software from a spec with Claude Code and review sub-agents
A commenter has Claude Code build from a detailed spec using code-review sub-agents, context management and automated tests while they supervise. ·
Done withClaude Code
- What they did
- A commenter described using Claude Code in auto mode to build software from a detailed spec, with code-review subagents, context-management rules, and extensive automated tests running as it builds. The human role was supervisory: leave it to work, then come back the next morning, decide what to change or refactor, update the spec, and repeat.
- How it went
- They said this could be especially powerful for solo developers and could create strong productivity gains. But they gave no concrete project, metrics, or finished result, and noted it is only worth trying if you can afford the token usage.
- Worth knowing
- The main catch they mention is cost: token use may be high enough that affordability is part of whether this workflow is practical.
Try it yourself with Claude Code
Build [what you want, e.g. a habit tracker web app] from the spec in [path to spec file]. Work in small steps, write automated tests as you go, and use code-review sub-agents to check each major piece before moving on. Keep your context tidy by summarizing progress in a notes file. Stop when all tests pass and give me a short list of decisions you want me to review. Ask before deleting files or installing anything unusual.
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
Source: Reddit · Apr 1, 2026
Related use cases
Turn tickets into merged pull requests
Pulls tasks from GitHub Issues or Linear, then runs Claude Code or Codex in isolated Kubernetes pods to carry each one through to a PR.
Build and balance a live Telegram RPG
A non-coding PM had Claude Code write a Telegram RPG bot, deploy it over SSH, and run tens of thousands of test fights to rebalance classes.
Automate content distribution across Reddit and LinkedIn
A Claude Code agent runs six cloud automations that turn new articles into Reddit posts and replies and LinkedIn posts with infographics; Reddit replies need Slack approval.
Submitted by from search gap: productivity
Build a C compiler with a team of parallel agents
Anthropic's engineering write-up on tasking Opus 4.6 agent teams with building a C compiler.
Scaffold a new service end to end with agents
Cursor agents create Terraform, k8s manifests and CI, validate them and open a draft PR; one cleanup run wrongly deleted Terraform files.
Run multi-file refactors and sysadmin tasks on network machines
A user uses Codex for multi-file refactors and precisely defined tasks on their network PCs and servers, then walks away.
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.