IdeaWell

Project ideas from Hacker News discussions.

AI migrated legacy COBOL programs to Java, bugs included

Original Article

Hacker News Discussion

📝 Discussion Summary (Click to expand)

Theme 1 – Bug‑for‑bug migration & Hyrum’s Law

"Bug‑for‑bug migration is a real thing in large code‑bases. You want to replicate all behavior of the code regardless whether the behavior is a feature or a bug." — wwind123
"Sometimes if the bug exists for enough time it is intended behavior." — nicman23

Theme 2 – Deterministic tooling vs. non‑deterministic LLMs

"The biggest problem is not that bugs are migrated with COBOL, but that lots of new bugs are going to be introduced. AI is not deterministic, it will be making tons of mistakes." — toplinesoftsys
"AI is not deterministic, the only realistic low‑error solution is not a more complex use of non‑deterministic AI, but deterministic transpilation." — dragonwriter

Theme 3 – Scale and legacy ecosystem of COBOL

"The largest test case was 4kloc. There are hundreds of billions of lines of COBOL in production. The IRS alone has approx 160 COBOL programs, averaging 230 kloc each." — pacaro

🚀 Project Ideas

Deterministic COBOL-to-Java Migration Engine

Summary

  • Generates Java code that precisely preserves original COBOL behavior, including known bugs, using an AST‑based deterministic translator.
  • Core value: Enables safe, low‑risk mainframe migrations without introducing new defects.

Details

Key Value
Target Audience Mainframe maintenance teams, fintech firms, insurance companies still on COBOL
Core Feature Bug‑for‑bug transpilation with an integrated test harness that validates output against production I/O
Tech Stack Rust for deterministic parsing, TypeScript for UI, PostgreSQL for storing test cases, Docker for reproducible builds
Difficulty High
Monetization Revenue-ready: Subscription SaaS $199/mo per mainframe instance

Notes

  • HN commenters constantly stress “bug‑for‑bug” migration; this tool directly solves that by automatically capturing and replaying legacy behavior.
  • Could spark discussion about deterministic AI pipelines and provide a practical way to implement the “incremental step‑by‑step” approach mentioned.

Legacy Test Data Vault for Mainframe Applications

Summary

  • A cloud service that catalogs, enriches, and version‑controls real production I/O datasets (files, DB2 records, CICS messages) used for verifying migrations.
  • Core value: Guarantees that migrated code behaves identically because tests are run against authentic data patterns.

Details

Key Value
Target Audience Database administrators, migration consultants, audit teams in regulated industries
Core Feature Ingests mainframe file/DB extracts, auto‑generates synthetic yet statistically faithful test suites, stores snapshots for regression checks
Tech Stack Python backend, AWS S3 + Glue, GraphQL API, React dashboard
Difficulty Medium
Monetization Revenue-ready: Pay‑per‑GB stored $0.02/GB + $49/mo base plan

Notes

  • Several HN remarks highlight “testing the migration” and “bugs will appear if test data isn’t representative”; this service provides the missing data layer.
  • Opens conversation about data‑centric approaches to legacy modernization.

Visual COBOL Flow Analyzer (VS Code Extension)

Summary

  • Parses COBOL source to generate interactive flow graphs that expose PERFORM ranges, GOTO jumps, and implicit variable dependencies.
  • Core value: Helps developers “understand why a duplicate INSERT is needed” and spot hidden control‑flow bugs before migration.

Details

Key Value
Target Audience Legacy COBOL maintainers, new hires, consulting firms performing migrations
Core Feature Real‑time syntax‑highlighted graph view, hover tooltips with variable scopes, exportable DOT files for external analysis
Tech Stack TypeScript, Tree‑sitter grammar for COBOL, Cytoscape.js for rendering, Node.js runtime
Difficulty Low
Monetization Hobby

Notes

  • Directly addresses “sometimes if the bug exists for enough time it is intended behavior” and “now no one understands the code”.
  • Likely to generate HN discussion about tools that make opaque legacy code approachable.

Read Later