Add more A/B data points (n>1) to docs/AB-RESULTS.md

[yuting0624]

Current results are n=1 per arm — direction is strong but confidence is loose.

Task: repeat the large-task A/B (solo@high / solo@max / hybrid), measure with scripts/measure-session.py <session-id>, and add a row/table to docs/AB-RESULTS.md.

Acceptance: at least one additional run per arm, with session ids and the COST-WEIGHTED numbers.

Great way to learn the workflow end to end.

[vaishnavidesai09]

Hi i would like to work issue please assign me

[yuting0624]

Hi @vaishnavidesai09 — thanks for your interest, assigning you now! 🙌 Happy to have you as the first contributor here.

A couple of heads-ups before you dive in:

• Setup required: you'll need the Antigravity CLI (agy) installed & authenticated, plus Claude Code. Running these incurs your own API / cloud costs (Claude + Gemini side), so just be aware of that.
• Pick a sizable task: the savings only show clearly on a reasonably large build (multi-file, lots of bulk work). Tiny tasks won't produce meaningful A/B numbers.
• Scope is flexible: you don't need all three arms — a single Claude solo run vs a hybrid run is enough to add a useful data point. Measure with scripts/measure-session.py <session-id> and add the session ids + COST-WEIGHTED numbers to docs/AB-RESULTS.md.

Take your time, and ask here if you hit anything. Looking forward to your PR! 🚀

[vaishnavidesai09]

Hi @yuting0624 — appreciate the heads-up on costs upfront. I wasn't able to cover the API expenses needed to run the sessions, so feel free to reassign this to someone else. Thanks for the opportunity!

[yuting0624]

Hi @vaishnavidesai09 — totally understand, no worries at all! Thanks a lot for being upfront about it. 🙏 I'll unassign this so it's open for others. You're very welcome to pick up any of the non-cost issues anytime — would love to have you contribute. Thanks again, and good luck! 🚀