DeepMind's Gemini 3.5 Pro Has a Secret Talent — and a Major Weakness
Leaked sentiment claims Gemini 3.5 Pro excels at one-shots but fails at agentic coding.
Deep Dive
A prominent AI Twitter poster speculates that “3.5 Pro” will be “very fun for creative one shots” but “abysmal for agentic coding.” Despite Google’s unmatched data, infrastructure, and cash flow, the poster finds it “crazy that they’re still struggling to catch up” after multiple times looking set to take a clear lead.
Key Points
- Gemini 3.5 Pro reportedly excels at creative one-shot tasks but struggles with agentic coding.
- Google DeepMind is said to be prioritizing world models over coding and RSI improvements.
- Despite massive data and infrastructure advantages, Google has repeatedly lost its lead to Anthropic and OpenAI.
Why It Matters
Even with vast resources, Google may not automatically win the AI race—execution and product focus are decisive.