Claude Code vs. Codex vs. Antigravity 2.0:Battle of Coding Agent

Hack Panel

About the session

What happens when developers hand over part of the coding task to AI and make the tools prove themselves live? 

In Battle of the Coding Agents, four panelists take on the same ML challenge using three different AI coding agents: Claude Code, Codex, and Google Antigravity 2.0.

The goal is simple: build a working LoRA fine-tune of DistilBERT on IMDB. 

For developers, ML engineers, architects, and AI builders, the session shows where these tools speed up development, where they still struggle, and where human judgment continues to matter.

Session Takeaways

See three AI coding agents solve the same ML task live. 

Watch how each tool handles setup, code generation, debugging, and last-mile fixes. 

Compare different agent styles, from pair-programming support to more autonomous execution. 

Understand where AI coding agents can accelerate real development workflows. 

See what still needs human review, intervention, and technical judgment. 

Walk away with a clearer view of which coding agent fits your workflow.