Project ideas from Hacker News discussions.

The state of AI in 2026: On the road to ROI

📝 Discussion Summary (Click to expand)

1. Skepticism about AI ROI and hype
Many commenters doubt that AI is delivering measurable returns yet, comparing it to past tech bubbles and noting the long timelines for payoff.
- “The ROÍ will be happening any day now, we promise. Please hire more McKinsey consultants who say they’re AI experts.” – mjhay
- “Do I understand correctly that means there is no AI ROI to show off yet? If correct that’s pretty bad, no?” – dgellow
- “For the internet in many cases it took a decade or longer too, no?” – Kassam_L (referencing the productivity paradox)

2. AI’s impact on software development
Opinions split on whether LLMs aid coding or create new technical debt; they speed up tasks like test generation and refactoring but can encourage quick‑fix rewrites that leave messy codebases.
- “LLMs are pretty good at helping with that though…. You can also ask them to write a bunch of tests and then rip out the old implementation and fix it with a new one.” – jeffreyrogers
- “Yeah, unreasonable deadlines… spent trying to untangle a mass of overly verbose spaghetti code… No, thank you. I’ll leave the underpaid disinfestation job to someone else.” – sph
- “These days ppl often… simply ask an LLM to replace something rather than debugging it… So rather than debug a steaming pile of code you might be able to treat it as a greenfield opportunity.” – gumby

3. Consulting/enterprise IT dynamics
A recurring theme is the reliance on high‑priced consultants (e.g., McKinsey) to “fix” IT messes, often seen as a costly band‑aid for deeper organizational issues.
- “You have to recognize that a vast majority of enterprises fall in the category of 'Dumb IT'… they look up to the likes of McKinsey for wisdom.” – bwfan123
- “Being a consultant to fix all those messes could be a good seam of code face to get stuck into…” – PaulRobinson
- “It's going to be a horrible job though, the equivalent of fixing Y2K bugs in Cobol code except with code bases that even more vast and unintelligible.” – rwmj


🚀 Project Ideas

LegacyCodeDocuBot

Summary

  • Automatically extracts business logic from undocumented legacy Python scripts and generates up-to-date documentation, diagrams, and unit‑test skeletons.
  • Core value: saves engineers hours of manual reverse‑engineering and reduces risk when modifying or replacing old code.

Details

Key Value
Target Audience Enterprise developers/maintainers wrestling with undocumented Python scripts and legacy codebases
Core Feature Ingest a folder of scripts → produce markdown docs, Mermaid flowcharts, and pytest stubs; suggest refactoring opportunities
Tech Stack Python (AST, pyflakes), LangChain/LlamaIndex for LLM integration, FastAPI for optional web UI, Docker
Difficulty Medium
Monetization Revenue-ready: SaaS subscription per codebase scanned (tiered by lines of code)

Notes

  • "Business logic obviously not documented anywhere, just mountains and mountains of slop readmes that might be accurate, but also might not be." – addresses the pain of missing docs.
  • "Unreasonable deadlines spent trying to untangle a mass of overly verbose spaghetti code" – the tool cuts down that untangling time.
  • HN commenters would love a concrete way to turn opaque scripts into maintainable assets without hiring expensive consultants.

AI Impact Meter

Summary

  • Measures the productivity impact and ROI of AI usage across engineering teams by integrating with Git, CI/CD, issue trackers, and chat logs.
  • Core value: translates AI activity into concrete metrics (time saved, defect reduction, cost avoidance) for executives and finance.

Details

Key Value
Target Audience Engineering managers, CTOs, and finance leaders who need to justify AI spend
Core Feature Collects activity data, computes PR cycle‑time changes, test automation gains, estimated cost savings; shows dashboards and trend reports
Tech Stack Backend: Go or Node.js; Data store: ClickHouse/Prometheus; Frontend: React; Optional LLM summarizer for insights
Difficulty High
Monetization Revenue-ready: tiered SaaS pricing based on number of engineers or repositories monitored

Notes

  • "It makes organizations more productive at producing code … but translating that into something that affects PnL is different." – this tool directly tackles that translation problem.
  • "I think companies will get better at measuring the impact of AI … and that the companies that are better at this will have an advantage." – provides the measurement capability HN users call for.
  • Enables data‑driven discussions about AI ROI, reducing reliance on vague consultant claims.

InternalAppForge

Summary

  • A framework that lets organizations replace SaaS tools with internally built, LLM‑assisted custom applications, providing scaffolding, security review, and incremental migration paths.
  • Core value: reduces vendor lock‑in and licensing costs while empowering teams to build fit‑for‑purpose tools quickly.

Details

Key Value
Target Audience Mid‑size enterprises tired of SaaS fees and seeking internal alternatives but lacking full‑stack expertise
Core Feature Generate a full CRUD app (backend API, frontend UI, auth, deployment configs) from a natural‑language spec using LLMs; includes governance checks for data security and audit logs
Tech Stack Backend: Python/FastAPI or Node/NestJS; Frontend: React/Terraform or Pulumi for infra; LLM integration via OpenAI or open‑source models
Difficulty High
Monetization Revenue-ready: enterprise license or per‑app fee; optional paid support & customization tier

Notes

  • "MatrixMan: We can do it ourselves. Simple custom tools are so much more effective than a conglomeration of general purpose ones. I've been ripping out SaaS offerings left and right, it's great." – aligns with the desire to replace SaaS with bespoke tools.
  • "Plenty of companies run most of their internal operations on Excel and Access for exactly this reason." – offers a safer, scalable alternative to those ad‑hoc solutions.
  • HN audience would appreciate a pragmatic path to internal tooling that leverages LLMs without sacrificing security or governance.

Read Later