GEMINI LABJP
PRO35 — Gemini 3.5 Pro is reported to target July 17, though the date, context window, and pricing all remain unconfirmed by GoogleREBUILD — The model is said to be a ground-up rebuild after the 2.5 Pro architecture was scrapped over structural failures in recursive tool-calling and SVG generationFLASH35 — Gemini 3.5 Flash is the newest broadly available release, tuned for faster, lower-cost coding, agents, multimodal, and long-document workSAATHI — ATL Saathi, built on Gemini 3.5 Flash, begins piloting in 100 Indian schools as a 24/7 planning and training assistant for Tinkering Lab educatorsSEA — In Southeast Asia, nearly 70% of prompts arrive in native languages — Vietnam 89%, Thailand 87%, Indonesia 84% — and 3 in 4 requests come from mobileSHEETS — From July 15, AI Expanded Access licenses raise the usage limits for Fill with Gemini and the AI function in SheetsPRO35 — Gemini 3.5 Pro is reported to target July 17, though the date, context window, and pricing all remain unconfirmed by GoogleREBUILD — The model is said to be a ground-up rebuild after the 2.5 Pro architecture was scrapped over structural failures in recursive tool-calling and SVG generationFLASH35 — Gemini 3.5 Flash is the newest broadly available release, tuned for faster, lower-cost coding, agents, multimodal, and long-document workSAATHI — ATL Saathi, built on Gemini 3.5 Flash, begins piloting in 100 Indian schools as a 24/7 planning and training assistant for Tinkering Lab educatorsSEA — In Southeast Asia, nearly 70% of prompts arrive in native languages — Vietnam 89%, Thailand 87%, Indonesia 84% — and 3 in 4 requests come from mobileSHEETS — From July 15, AI Expanded Access licenses raise the usage limits for Fill with Gemini and the AI function in Sheets
TAG

Pairwise Comparison

1 articles
Back to all tags
Related:
Gemini API1Vision1ASO1App Store1LLM-as-judge1Python1Indie Dev1
Gemini API/2026-07-16Advanced

I Asked Gemini to Grade My App Store Screenshots. Everything Scored 78–85.

Ask Gemini Vision to grade App Store screenshots out of 100 and good candidates and deliberately broken ones both land at 78–85. Here is how I measured the judge's discrimination power, dropped absolute scoring, and rebuilt it as debiased pairwise comparison.