How to use Claude to run parallel web app development
A freelancer scopes parallel Claude Code sessions around selected rules, plan approval, code and UI review, tests and documentation updates.
news.ycombinator.com
Prompt to try
Editorial adaptation, not the source author’s exact prompt. Replace bracketed inputs and review tool permissions first.
For [single-system task], read [relevant rules] and [system documentation], then inspect [similar feature]. Plan the work for one session and wait for plan approval. Implement the approved plan, create tests for [behaviors], then update the feature docs and record decisions and lessons.
Reach people exploring AI agents
See placements →What happened
The author reports producing more code without typing it, with code quality almost matching their own and devlogs helping refine the workflow.
Reported by hakanderyal on Hacker NewsHow the agent did it
- Selected a bounded task and loaded relevant rules and system documents.
- Started plan mode and asked the agent to inspect similar existing features.
- Examined and refined the plan before approving implementation.
Where a person stepped in
The author approves plans, watches implementation, intervenes when needed, reads code and manually tests the UI.
A person approved each step before the agent continued.
A solo freelancer describes using Claude Code for C#/.NET backends and React web applications while no longer typing code themselves. Their workflow runs three to eight visible sessions in parallel, depending on task demands and available attention. They credit existing examples and selectively loaded rules with keeping the output close to the code they would write.
For each session, the author chooses a task confined to one system, loads relevant rule files and feature documentation, and starts plan mode. Similar existing features become examples for the agent to inspect. Larger efforts get an initial plan, followed by separate sessions that plan individual phases in depth.
The author examines each plan, discusses changes and approves it before implementation. They watch the main tasks and intervene when necessary. Afterwards, they inspect code in an editor and test through the UI while the agent writes guided tests. The agent then updates feature and system documentation.
A newer devlog command records session metadata, decisions and lessons, and adds changelog entries to relevant documents. The author says this helped refine the workflow. Who it's for: experienced developers who can divide work into bounded tasks and actively review several concurrent sessions.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Hacker News · hakanderyal
Frequently asked questions
What does the author review before implementation?
The author examines the agent’s plan, goes back and forth on it, and approves it before watching implementation.
How is completed work checked?
The author reads the code in an editor and tests manually through the UI while the agent creates tests under specific guidance.
More details
Remaining steps
- Watched implementation and intervened on main tasks when needed.
- Reviewed code and manually tested the UI while the agent wrote guided tests.
- Asked the agent to update documentation and record session devlogs.
About this account
Claude Code plans and implements scoped tasks across parallel sessions, writes guided tests and maintains documentation.
Platform: Hacker News. Attribution: https://news.ycombinator.com/item?id=46701913. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: hakanderyal on Hacker News
Source published 21 Jan 2026
Source checked 14 Oct 2026
Tools and topics
Claude Code · C#/.NET · React · Markdown · CLI tools
Parallel sessions · Development workflow · Documentation
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.