I took a closer look at what sparked the noise: not a new model release, pricing change, or API update, but Demis Hassabis's public manifesto and the visibly nervous reaction surrounding it. And here I had to pause: the market reads such texts not as philosophy, but as a signal of where a company's real focus lies.
If I'm involved in AI integration or planning to build AI automation on a specific vendor's stack, what matters isn't elegant phrasing—it's the rhythm of model delivery, API stability, and roadmap predictability. So the community's reaction is more telling than the manifesto itself.
But there's an important factual caveat. I don't see confirmed public evidence of mass Gemini 3 delays or any official code red inside DeepMind. On the contrary, open data shows Google maintained a dense release pace in 2026: Gemini 3.5 Flash, Gemini 3.1 Pro, Gemini 3 Deep Think, Gemma 4, and several specialized launches.
So the debate isn't about a proven failure—it's about perception. A part of the tech community senses a gap between the leader's image, public rhetoric, and the speed at which a competitive market, especially Chinese players, enforces tempo in coding, reasoning, and price.
I don't ignore such signals. When a strong lab attracts chatter like "enough manifestos, show steady product progress," it usually means trust in the vendor is being measured not by benchmarks, but by execution discipline.
Business and Automation Impact
For businesses, the takeaway is simple: I wouldn't bet on a single AI automation provider just because of brand. If there's too much noise around the roadmap, I immediately design a multi-vendor AI architecture with the ability to quickly switch models within the same pipelines.
Those who already have model abstraction layers win. The ones who hard-coded processes, prompts, and quality evaluation to a single API lose, as they become hostage to someone else's pace.
These are exactly the forks we tackle with clients at Nahornyi AI Lab: where you need a single flagship model, and where it's smarter to go for AI solution development with a backup circuit, routing, and cost control. If your platform or internal processes already hit a model-selection bottleneck, let's review the architecture together and build an implementation without blind faith in anyone's manifestos.