March 10, 2026
Genius or just “good”?
Maybe the G in AGI stands for Gemini
Fans crown Gemini the “G” in AGI — skeptics say it’s just good, not godlike
TLDR: Google launched Gemini 3.1 Flash-Lite, a fast, cheap AI known for image skills, while retiring Gemini 3 Pro quickly. The crowd is split: some praise Gemini’s visual magic and “general” usefulness, others stick with ChatGPT, and many blast Google’s model churn as risky for people building on these tools.
Google just dropped the mini-but-mighty Gemini 3.1 Flash-Lite, and writer Robin Sloan is hyped: fast, affordable, and sharp with images. He even hints Google’s chasing something more general—as in AGI, “artificial general intelligence” that’s useful beyond code—while rivals train code-writing bots. The comments? A spicy split.
Philosophy mode: skyberrys asks, “what do we really mean by General?” Cue existential thread about whether “smart” means “does everything” or “talks to everyone.” Industry tea: vadepaysa’s friend says Gemini’s visual skills are “best in class” and already powering real apps. But the duel isn’t settled—synergy20 says ChatGPT still answers better, so they pay for both. Crown on ice.
The chaos challenge: jrm4 wants to feed an attendance sheet of 50 students’ wild handwriting to see if any AI can decipher pure scribble. Meanwhile, 0gs side-eyes Sloan’s enthusiasm with a wink—“his first novel was about Google,” implying hometown bias. Drama, meet popcorn.
Then Sloan drops the rant: Google’s quick retirement of Gemini 3 Pro feels like model mayflies, leaving developers anxious about sudden yank-and-swap updates and weird regressions. You could self-host to avoid that… except nothing matches Gemini’s visual chops—yet. Translation: the AI race is speeding up, and the finish line keeps moving.
Key Points
- •Google released the Gemini 3.1 Flash‑Lite model, described as fast and capable, with documentation available via the Gemini API site.
- •The author states a preference for Gemini models due to speed, cost, and visual task performance, integrating them programmatically into other systems.
- •The post asserts that Anthropic and OpenAI are focusing on coding agents, while Google is refining a more general‑purpose intelligence with Gemini.
- •Google is deprecating Gemini 3 Pro and urging migration to Gemini 3.1 Pro (preview) before March 9, 2026, per the referenced announcement.
- •The author notes self‑hosting can avoid sudden deprecations, but currently sees no self‑hosted model matching Gemini’s visual acuity, expecting improvements in the near future.