Cara, the last week of February was absolutely crazy for those following AI. Seriously, I don't remember seeing so much concentrated activity like this. Wall Street was both fascinated and terrified — and for good reason.

Did you see what Google DeepMind did? They launched Gemini 3.1 Pro with a context window of 1 million tokens. I mean, the amount of information this model can now process is absurd. Text, code, images, all together in huge sessions. And the prices are still competitive, which means this technology is finally reaching real companies, not just labs.

But it wasn't just Google. Anthropic released Claude Sonnet 4.6 and people couldn't stop talking about the improvements in coding and reasoning. They also introduced Claude Cowork, an AI agent that runs on your computer and can interact with local files and browsers. This AI agent trend is everywhere now.

And there's more. Alibaba entered with Qwen 3.5 — 397 billion parameters, man. A mixture of experts architecture designed to be cost-efficient. It looks like China is taking this very seriously. Meanwhile, ByteDance showcased Seedance 2.0, a model that generates realistic videos from text or images. Of course, it came with more protections this time because people weren’t quiet about the criticisms of synthetic media.

There’s even a Spanish company, Multiverse Computing, that launched Hypernova 60B — a compressed model using quantum-inspired techniques. Available for free to developers. It promises to reduce inference costs, which is music to the ears of startups bleeding money on computing.

Now, here’s the part that’s both scary and exciting: infrastructure spending. Google, Amazon, Meta, and Microsoft together committed around 650 billion dollars in AI infrastructure by 2026. We’re talking data centers, custom chips, cloud expansion. Is this disciplined investment or reckless speculation? Hard to say.

OpenAI didn’t stay out either. They closed a 10 billion dollar deal with Cerebras Systems for wafer-scale chips. Capable of hundreds of megawatts. All to accelerate inference in ChatGPT and support increasingly complex models until 2028. They also hired Peter Steinberger, creator of Openclaw.

And there’s the edge computing aspect. Ambiq expanded research in Singapore for ultra-low-power AI. Intelligence on devices in wearables and industrial systems. When energy is so expensive, efficiency becomes a competitive weapon.

Regulators also woke up. The UK is planning free AI training for 10 million adults by 2030. The EU launched a transparency code project under the AI Act, with requirements to label generated content and rules for high-risk systems.

But what really matters is that AI has left the lab. Reuters implemented tools that reduced corrections by 10%. In biotech? 73% adoption of AI in protein prediction. Lowe’s deployed voice agents in stores to assist customers. Samsung partnered with Gracenote to improve search on smart TVs.

That’s where you see the real impact. It’s no longer just a fancy showcase; it’s real production. Productivity gains or disappointments are now visible.

Wall Street is divided. Optimists see a renaissance of productivity through automation and advanced reasoning. Pessimists see capex expanding and sky-high valuations vulnerable to slower monetization.

For society at large, the debate is even more tense. Some imagine abundance driven by AI. Others warn about job displacement, misinformation, and opaque systems operating beyond public understanding.

A week of announcements doesn’t solve anything. But it makes clear: the race is accelerating and no one is standing still. Neither regulators, nor investors, nor tech companies. The tension between innovation and caution has never been so evident.