Claude vs Codex: Fix a Real Bug

Claude vs Codex: Fix a Real Bug

A live, side-by-side engineering session on a real production-scale codebase

AI coding tools are everywhere. Very few engineers know how they behave on a large, real production system, or how to use them at their level of seniority.

In this session, we put Claude and Codex through the same production bug on Apache Fineract, an open-source core banking platform used by financial institutions worldwide. A large, multi-module Java & Spring Boot system with real loan, repayment and accounting logic.

THE BUG

A customer makes a loan repayment. The network is slow, the request is retried, and the repayment is posted twice. Wrong balance, broken accounting, and no obvious error in the logs.

THE SHOWDOWN: FOUR ROUNDS, FOUR ROLES

Both tools tackle the same bug, round by round, the way real engineering teams work:

Round 1 | Software Engineer: understand an unfamiliar codebase and reproduce the bug
Round 2 | Senior Software Engineer: find the root cause with evidence and apply a minimal, tested fix
Round 3 | Lead Engineer: enforce team standards, automate code review and add guardrails
Round 4 | Principal Architect: make AI safe at scale with permissions, integrations and quality measurement
Every round is scored live on evidence quality, correctness, test quality, change size and review findings. You'll also see exactly where AI gets it wrong, and how an engineer catches it.

WHAT YOU WILL LEARN

- How Claude and Codex differ in how they work, and when to use each
- A production-grade AI bug-fixing workflow: reproduce, test, fix, verify, review
- How AI changes the work at every level, from Software Engineer to Principal Architect
- The skills that keep you valuable as AI changes software engineering
- Prompts and practices you can apply at work immediately

WHO SHOULD ATTEND
Engineers with 0–16 years of experience: Software Engineers, Senior Software Engineers, Lead Engineers, Principal Engineers, Architects and Engineering Managers.
The demo is in Java. The workflow applies to Java, .NET, Python and other stacks.
FORMAT
2 hours | Live demo-led | Interactive Q&A
Limited seats to keep the session interactive.
Optional preparation: explore Apache Fineract at github.com/apache/fineract

━━━━━━━━━━━━━━━━━━

VENKATESH DB
Enterprise AI that ships.
━━━━━━━━━━━━━━━━━━