AI coding tools shine in clean tutorials and side projects. But what happens when you bring them to a 500,000-line codebase with years of history, cross-service dependencies, and a strong team style?
This is a practical talk — no demos prepared in advance. We'll open real codebases on the web together, throw the same problems at AI coding tools, and you'll decide for yourself whether the result holds up.
We'll work through 10 practical cases drawn from real engineering work on large codebases:
1. Adding a feature end-to-end across a monorepo (DB → API → frontend), without manual orchestration.
2. Removing a deprecated library from hundreds of call sites in one pass.
3. Porting a service between languages (e.g., Python → Go) while preserving behavior.
4. Generating test coverage for legacy modules that have never been tested.
5. Refactoring duplicated code spread across many files and services.
6. Migrating major framework versions (React, Next.js, Strapi) with breaking changes.
7. Onboarding a new engineer — letting them query the codebase instead of reading docs for a week.
8. Running parallel YOLO-mode development across 3+ independent tasks at once.
9. Auditing a large codebase against security and compliance guardrails.
10. Implementing a feature directly from a PDF spec — an RFC, a book chapter, or an internal design doc.
Evgeny Potapov, engineering manager with over 20 years of experience and co-founder of ApexData, will share what consistently works and what still breaks on large codebases, with concrete prompts you can take home and try.
For developers, tech leads, and engineering managers who already use AI coding tools but want to push them further on real code.
AI coding tools shine in clean tutorials and side projects. But what happens when you bring them to a 500,000-line codebase with years of history, cross-service dependencies, and a strong team style?
This is a practical talk — no demos prepared in advance. We'll open real codebases on the web together, throw the same problems at AI coding tools, and you'll decide for yourself whether the result holds up.
We'll work through 10 practical cases drawn from real engineering work on large codebases:
1. Adding a feature end-to-end across a monorepo (DB → API → frontend), without manual orchestration.
2. Removing a deprecated library from hundreds of call sites in one pass.
3. Porting a service between languages (e.g., Python → Go) while preserving behavior.
4. Generating test coverage for legacy modules that have never been tested.
5. Refactoring duplicated code spread across many files and services.
6. Migrating major framework versions (React, Next.js, Strapi) with breaking changes.
7. Onboarding a new engineer — letting them query the codebase instead of reading docs for a week.
8. Running parallel YOLO-mode development across 3+ independent tasks at once.
9. Auditing a large codebase against security and compliance guardrails.
10. Implementing a feature directly from a PDF spec — an RFC, a book chapter, or an internal design doc.
Evgeny Potapov, engineering manager with over 20 years of experience and co-founder of ApexData, will share what consistently works and what still breaks on large codebases, with concrete prompts you can take home and try.
For developers, tech leads, and engineering managers who already use AI coding tools but want to push them further on real code.
AI coding tools shine in clean tutorials and side projects. But what happens when you bring them to a 500,000-line codebase with years of history, cross-service dependencies, and a strong team style?
This is a practical talk — no demos prepared in advance. We'll open real codebases on the web together, throw the same problems at AI coding tools, and you'll decide for yourself whether the result holds up.
We'll work through 10 practical cases drawn from real engineering work on large codebases:
1. Adding a feature end-to-end across a monorepo (DB → API → frontend), without manual orchestration.
2. Removing a deprecated library from hundreds of call sites in one pass.
3. Porting a service between languages (e.g., Python → Go) while preserving behavior.
4. Generating test coverage for legacy modules that have never been tested.
5. Refactoring duplicated code spread across many files and services.
6. Migrating major framework versions (React, Next.js, Strapi) with breaking changes.
7. Onboarding a new engineer — letting them query the codebase instead of reading docs for a week.
8. Running parallel YOLO-mode development across 3+ independent tasks at once.
9. Auditing a large codebase against security and compliance guardrails.
10. Implementing a feature directly from a PDF spec — an RFC, a book chapter, or an internal design doc.
Evgeny Potapov, engineering manager with over 20 years of experience and co-founder of ApexData, will share what consistently works and what still breaks on large codebases, with concrete prompts you can take home and try.
For developers, tech leads, and engineering managers who already use AI coding tools but want to push them further on real code.
