Google launches Gemini 4 Argon in Antigravity
An author used a 2015 Amazon Kindle Paperwhite with a jailbroken firmware and a Home Assistant dashboard to compare two AI coding agents. The agents were Google’s Gemini 4 Argon, run through the Antigravity desktop app on a $20 Google AI Pro plan, and Anthropic’s Claude Code, run on an Opus 5.5 model under a $100 Claude Max plan. Both were given the same one‑line brief and a single hour to work…
Key points
- Gemini 4 Argon used 18% of a $20 Google AI Pro plan in a 28‑minute run.
- Claude Code generated 63 tests versus Argon's 4, but made fewer visual changes to the Kindle dashboard.
- Both agents ran in about half an hour on a Debian LXC with Node 20, but only Claude Code kept the layout unchanged.
Gemini 4 Argon used 18 % of the weekly allowance in a 28‑minute run, while Claude Code consumed $5.33 of API‑equivalent usage. Argon produced four tests; Claude Code produced 63, including a check for an ancient WebSocket draft. Visually, Argon widened forecast columns and removed a sensor card, whereas Claude Code left the layout unchanged but added style rules. Both agents fixed hidden bugs and added a Git attributes file, but only Claude Code logged each missing calendar separately and added a test for it.
The experiment shows that even with identical prompts, the agents differ in test coverage, visual impact, and cost. Developers should verify changes on real hardware and consider the trade‑off between thorough testing and minimal visual disruption.
Model pages: Gemini 4 Argon → · Claude Code → · Opus 5.5 →
The story so far
4 episodes →- Google launches Gemini 4 Argon in Antigravitythis story
I gave Gemini 4 Argon in Antigravity and Claude Code the same messy side project, and only one of them left it better than it found it
xda-developers.com · 6 November 2026
Loading the full article…
This text was published by xda-developers.com and written by Joe Rice-Jones. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Agents & Tools
All →- Claude Code boosts blog output to 140 posts per month · 1 src
- Unite.AI ranks 10 AI image generation APIs for developers · 1 src
- Build AI agent memory with OpenAI Agents SDK and Mem0 · 1 src
- Author builds local travel planner with Hermes Agent and Ollama · 2 src
- Python agent gains memory with JSON file and OpenRouter API · 1 src
Comments
via GitHub Discussions