I have access to Claude Code at work and used both Claude Code and Antigravity at home. I have shipped several projects/features with both tools.
First of all, AI Pro is simply the best value for the family, you are getting four accounts not just one! My entire family can do Vibe Coding on a B+ product, maybe A- if the project is mostly front end, which most home projects are. This is on top of higher limit with other Google AI Products, Gemini in Docs/Sheets/Slides, etc. In 2026, I consider vibe coding (and agentic document creation) to be an universal skill, all four of us being able to do this for less than $20 per month, is massive. We already paid $99 a year for 2TB Google Drive, so $99 more for 3 more TB Google Drive and all this AI features? I just can’t justify using anything else. Even the Chinese models cost way more the way we are using it, and that is for Vibe Coding alone. This is too good to be true and I fully expect Google to nerf it someday when they get market share advantage over Open AI and Anthropic.
I do think the weekly quota on Gemini Pro Low and Sonnet are a little too low, I get maybe 2-3 hours a week out of it, planning only. But Flash is a monster, it is very close to Gemini Pro in execution after you create the AGENTS.md, DESIGN.md and implementation plan with bigger models. I hope they will make it more agent friendly, as I do find Claude Code, especially with Gary Tan G Stack simply tries harder, and goes through more iterations before handing it to me to review. I actually don’t think Opus vs Sonnet/Flash matters as much the willingness to iterate. Whether the agent performs the task with 95% accuracy or 85% accuracy doesn’t matter, both need to be tested before saying “I am done.”.
I do find Antigravity really strong at searching the Internet for most up to date information on new API’s, images, config files, that type of thing. This helps a lot with the overall project. I love pasting inspirational screenshots into the IDE, and ask Antigravity to use Nano Banana 2 to generate mock ups for the site design and update DESIGN.md. 30 minutes later, my site looks exactly like the mock up!
Most importantly, Google has more user data, money and compute via TPU than rest of the AI coding companies combined. Their biggest problem is having three coding products due to politics: Antigravity, Gemini CLI, Jules . . . According to Bloomberg they are taking this very seriously and Antigravity is chosen to be the flagship going forward. Pichai talked about it this past week and mentioned how Gemini for Mac was built with it. Once they kill off the other two, consolidate the product, if Claude Code is 100 out of 100, the gold standard, Antigravity will easily be 95+. Right now I would rate it 85 overall, 80 if you are doing heavy backend work, 90 if you are doing heavy front end work, and you get to use it for the whole family, that is massive.
None of the bugs, quota issues are difficult to resolve, I expect big things next month.