🏴‍☠️ Gemini 3.1 Pro Doubles AI Reasoning Power

Google Just Raised the Bar

Author

Captain YAR
February 22, 2026 • Estimated Reading Time: 9 minutes

The biggest AI jumps don’t feel like “new features.” They feel like the floor just moved under your feet. You try a task you’ve done before, and in this creator’s breakdown, the model doesn’t just answer, it understands.

If the demos translate into everyday work, for many of us Gemini 3.1 Pro could reset what we expect from “reasoning.” Not because the writing gets prettier, but because the model handles space, structure, and constraints without collapsing. That’s where AI stops sounding smart and starts being useful.

Google positions Gemini 3.1 Pro as a major leap, with reasoning performance that effectively doubles what its predecessor could do. The video adds a spicy detail: the “Deep Think” model everyone’s been talking about was apparently running on the 3.1 engine the whole time. If that’s true, some of the impressive results people attributed to “Deep Think” were really a preview of 3.1’s core capability.

In partnership with

The Lithium Boom is Heating Up

Thanks to growing demand, lithium stock prices grew 2X+ from June 2025 to January 2026. $ALB climbed as high as 227%. $LAC hit 151%. $SQM, 159%.

This $1B unicorn’s patented technology can recover 3X more lithium than traditional methods. That’s earned investment from leaders like General Motors.

Now they’re preparing for commercial production just as experts project 5X demand growth by 2040. They’ve announced what could be one of the US’ largest lithium production facilities and have rights to approximately 150,000 lithium-rich acres across North and South America.

Unlike public stocks, you can buy private EnergyX shares alongside 40,000+ other investors. Invest for $11/share by the 2/26 deadline.

This is a paid advertisement for EnergyX Regulation A offering. Please read the offering circular at invest.energyx.com. Under Regulation A, a company may change its share price by up to 20% without requalifying the offering with the Securities and Exchange Commission.

What makes this interesting is the shift from pure text to executable reasoning. When an AI writes code for SVG, simulations, or CAD, it can’t hide behind confident language. The output either runs correctly, or it doesn’t, and the feedback loop is immediate, which changes how you build and learn.

The “Pelican Bench” proves visual coding mastery
The creator spotlights a test called the “Pelican Bench,” a prompt that asks for SVG code drawing a pelican riding a bicycle. Older models often produced static shapes or awkward, nonsensical compositions. In the demo, Gemini 3.1 Pro generated an animated SVG where motion looks intentional, not accidental.

That distinction matters. The model isn’t “drawing pixels,” it’s encoding structure, timing, and implied physics into code you can verify. The video shows more examples, like a giraffe driving a tiny car and a murmuration-style simulation of flocking, reinforcing the same point: visual reasoning is being translated into executable instructions.

Smashing the “impossible” benchmarks
The headline number is the ARC AGI benchmark, designed to measure fast generalization and learning of new rules, not memorization. In the breakdown, Gemini 3.1 Pro is reported at 77.1%. The creator frames that as more than double Gemini 3 Pro’s result, and claims it outperformed models like Opus 4.6 and GPT 5.2 in the same testing context.

They also cite strong performance on “Humanity’s Last Exam” and scientific knowledge evaluations.

Take that with the right mindset: benchmark wins are not a product. But if independent tests line up, it suggests a model that adapts better when the problem looks unfamiliar, which is exactly what real projects demand. A simple way to pressure-test this yourself is to ask for a small animated SVG, a constrained layout, and a step-by-step fix when the first attempt fails.

200+ AI Side Hustles to Start Right Now

While you were debating if AI would take your job, other people started using it to print money. Seriously.

That's not hyperbole. People are literally using ChatGPT to write Etsy descriptions that convert 3x better. Claude to build entire SaaS products without coding. Midjourney to create designs clients pay thousands for.

The Hustle found 200+ ways regular humans are turning AI into income. Subscribe to The Hustle for the full guide and unlock daily business intel that's actually interesting.

*Ad

Other awesome AI guides you may enjoy

From text to physical reality: CAD and urban planning
The most practical demo is where the model stops acting like a chatbot. The creator references an example associated with Google’s Jeff Dean: an urban planning tool that generates full city layouts from geographic constraints. Change a river, move paths, adjust boundaries, and the system rebuilds the block logic immediately.

Even more exciting for makers is “Prompt to CAD.” Feed the model a technical spec, and it outputs manifold CAD code meant to be ready for 3D printing. This effectively bridges the gap between describing an object and actually making one, especially for people who know the spec but don’t want to wrestle with every dimension by hand from scratch alone.

Unlock ChatGPT’s Full Power at Work

ChatGPT is transforming productivity, but most teams miss its true potential. Subscribe to Mindstream for free and access 5 expert-built resources packed with prompts, workflows, and practical strategies for 2025.

Whether you're crafting content, managing projects, or automating work, this kit helps you save time and get better results every week.

*Ad