Project ideas from Hacker News discussions.

How I write software with LLMs

📝 Discussion Summary (Click to expand)

1. Splitting the agent into specialized “sub‑agents” (architect, developer, reviewer, etc.)
Many users argue that breaking a single LLM into role‑specific agents helps manage context, enforce permissions, and reduce hallucinations.

“If you want to separate capabilities, definitely.” – chriswarbo
“The orchestrator runs the whole thing… architect → developer → reviewer.” – marcus_holmes

2. Cost‑vs‑quality trade‑off and token budgeting
A recurring point is that multi‑agent pipelines can be expensive, but they can also save tokens by delegating simple tasks to cheaper models.

“One tier, one model is cheaper, but the quality comes with the reviewers.” – stavros
“You spend your token & context budget in full in 3 phases.” – hakanderyal

3. Human understanding and the “vibe‑coding” critique
Several comments caution that relying on LLMs without reading or reviewing the code turns developers into “no‑code” users and erodes architectural insight.

“If you fail to even read the code produced, then I might as well treat it like a no‑code system.” – ashwinsundar
“I can understand the high‑levels of how no‑code works, but as soon as it breaks, it might as well be a black box.” – ashwinsundar

4. Workflow & tooling integration (CLI vs IDE, markdown artifacts, harnesses)
Users discuss how to embed agents into existing toolchains, the value of markdown‑based plan files, and the pros/cons of terminal‑based vs. IDE‑based agents.

“I’m using a hierarchy of artifacts: requirements doc → design docs → code+tests.” – aix1
“All artifacts are version controlled.” – aix1
“I just want to talk to a model all day, but that’s not the same as writing code.” – lbreakjai

These four themes capture the main strands of opinion in the discussion.


🚀 Project Ideas

Generating project ideas…

AgentFlow: Lightweight Multi‑Agent Orchestration Platform

Summary

  • Provides a simple, declarative way to define and run multi‑agent pipelines (architect, planner, developer, reviewer) with role‑based context windows and permission boundaries.
  • Automates cost‑aware model selection, token budgeting, and seamless GitHub integration for pull‑request generation.

Details

Key Value
Target Audience Developers and teams experimenting with LLM‑powered coding workflows.
Core Feature YAML/JSON pipeline definition, role‑based context isolation, cost‑aware model routing, GitHub PR automation.
Tech Stack Node.js + Express, TypeScript, OpenAI/Anthropic API wrappers, GitHub Actions, Docker.
Difficulty Medium
Monetization Revenue‑ready: $29/month per team, free tier with 10k tokens/month.

Notes

  • HN commenters love the idea of “splitting into subagents” and “separate agents with separate context windows” (e.g., “split on domain of side effects, not on task complexity”).
  • Provides a practical utility for teams that want to bolt an orchestrator onto an existing repo without messy integration, addressing “Can you bolt superpowers onto an existing project?”.

CodeTrace: AI‑Generated Code Traceability & Review

Summary

  • Automatically generates version‑controlled markdown artifacts (requirements, design, test plans, implementation notes) for every LLM‑generated change.
  • Integrates with CI to run tests, static analysis, and produce review comments, ensuring traceability and maintainability.

Details

Key Value
Target Audience Teams using LLMs for code generation who need audit trails and reviewability.
Core Feature Artifact generation, diff summarization, automated test & lint runs, GitHub PR comment bot.
Tech Stack Python, FastAPI, GitHub API, PyLint, pytest, OpenAI API.
Difficulty Medium
Monetization Hobby

Notes

  • Addresses frustration that “LLM‑generated code is a maintenance nightmare” and the need for “reviewable code” (see “reviewable code, that is code specifically delivered in a manner that makes code review more straightforward”).
  • Enables developers to “anchor” decisions in timestamped files, a practice praised by “I want the LLM to write the plan to a file as well”.

ReqFlow: AI‑Powered Dynamic Requirements Capture

Summary

  • Conversational interface that turns natural‑language feature requests into structured flowcharts, IPO diagrams, and markdown requirement specs.
  • Supports iterative refinement, version control, and export to project‑management tools (Jira, Trello, Notion).

Details

Key Value
Target Audience Product managers, designers, and developers who struggle with evolving requirements.
Core Feature LLM‑driven diagram generation, versioned requirement docs, integration with PM tools.
Tech Stack React, Node.js, D3.js for diagrams, OpenAI API, Notion/Trello APIs.
Difficulty Medium
Monetization Revenue‑ready: $15/month per user, free tier with 5 projects.

Notes

  • Directly tackles the pain point “one big problem: ability to come up with requirements” and the desire for “flow charts, IPO charts” (see “I would like to see a resurgence of flow charts”).
  • Provides a practical utility for teams that need to keep requirements traceable as they change.

CodeGuard: LLM‑Driven Code Quality Auditor

Summary

  • Runs LLM‑based code reviews, static analysis, and security scans on every commit or PR, producing actionable reports and diff comments.
  • Offers IDE plugins (VS Code, JetBrains) for real‑time feedback and a web dashboard for historical quality metrics.

Details

Key Value
Target Audience Developers and QA teams looking to automate review of LLM‑generated code.
Core Feature LLM review bot, static analysis integration, security scanning, IDE plugin.
Tech Stack Go for backend, TypeScript for plugins, OpenAI/Anthropic APIs, ESLint, Bandit, OWASP ZAP.
Difficulty High
Monetization Revenue‑ready: $49/month per repo, enterprise licensing.

Notes

  • Responds to the recurring complaint that “LLM code is messy and hard to review” and the need for “automated code coverage and theorem proving” (see “we need to automate things like theorem provers and code coverage”).
  • Offers a concrete tool that can be discussed in terms of cost‑benefit and integration complexity, a common HN discussion theme.

Read Later