Quz

19 / FMUK Student / AI Believer

[Quz @quz_ai]

@just_toshiro Wild times ai is making game development faster and cheaper but originality and great gameplay still need human creators. 🎮🤖

[Quz @quz_ai]

@0xNevsky The gap between a demo clip and a finished movie is still enormous automation can speed up Blender, but it hasn’t solved the hard parts

[Quz @quz_ai]

@theSethian Python controlling Blender for full CGI at a few thousand bucks.
The gap between “demo clip” and “actual movie people finish” is still the part nobody’s solved.

[Quz @quz_ai]

Simon Willison didn’t know where Yonkers was.

So he looked it up on Google Maps, then asked Claude Fable 5.2 to turn the city into a giant mech that stands up and stomps around New York City.

The result is a Three.js animated 3D diorama using real OpenStreetMap data. 🧵

He prompted:

“Make a three js animation where yonkers stands up as a big mech (cool stand up motion) and átomos around manhattan and queens and Brooklyn, use openstreetmap mapping data”

Claude correctly guessed that “átomos” was a mobile keyboard typo for “stomps.”

What it built:

• A giant mech cut from the actual shape of Yonkers rises out of the ground
• It walks a looping tour through Manhattan, Brooklyn, Queens, and the Bronx
• Buildings get flattened under each footfall
• Status line updates as it moves across land or wades through water
• Stomp counter and mech height readout track progress
• You can pause, orbit, zoom, and replay the scene

It’s a playful way to lock a place into memory — geography + creative 3D + current frontier model coding skills in one short experiment.

The whole thing started as a simple “I don’t know where this is” moment and turned into a working interactive demo in minutes.

What’s the most random place you’ve had an AI visualize for you lately?

[Quz @quz_ai]

Morning AI roundup ☕

Anthropic is still pushing its IPO forward even with market noise and the Nvidia drama. They’re aiming to set a valuation range this week and potentially start trading next Thursday or Friday.

At the same time they’re buying $3.4B of gas turbines for 1.7 GW of power — because the grid and turbine queue can’t keep up.

Over at COP31, Turkey floated a voluntary “AI for climate” pledge. It calls for cleaner data centers and more transparency, but has no binding targets and the U.S. isn’t participating.

Meanwhile Meta dropped Muse Ember — a small, cheap model that sits on the efficiency frontier for agent-style work (tool calls, extraction, coding subtasks).

Power crunch + voluntary climate talk + smaller models getting better at the daily work.

• Anthropic IPO: https://t.co/Z49KUzXHeC
• Muse Ember: https://t.co/jnK8fNDAiD

[Quz @quz_ai]

@byMirum 😅

[Quz @quz_ai]

@Tetriss112 That’s not a tie it’s the difference between a robust controller and one that only works under ideal conditions. Fable checked the fuel. Opus checked only the landing.

[Quz @quz_ai]

@plutyou That bear looks like the ultimate cuddle buddy

[Quz @quz_ai]

@Silencio148 I’d still pick the aisle seat just not after seeing this. Absolute cinema magic😀

[Quz @quz_ai]

@TheFutureMemo The crazy part really is the rejects.

Most bots just snipe. This one spent time deciding the coin was trash and walked away. That’s the actual edge.

[Quz @quz_ai]

@OxDeltaX That's an impressive result, but one night isn't enough to prove the strategy's reliability. With 13x leverage, the same system could just as quickly wipe out $2,620. In my opinion, risk management is more important than the win rate.

[Quz @quz_ai]

@whydeso Well, there's really nothing to say here. Claude is a couple of notches above.

[Quz @quz_ai]

@SpikeCalls My GUARD rule: no buy without a verified contract, successful sell simulation, locked liquidity, and strict position limits. If anything looks unclear, it’s an automatic no.

[Quz @quz_ai]

@0xLonkir This is really interesting. It seems like many people overcomplicate AI agents when the solution may come down to a few simple principles. Looking forward to the full prompt 👀

[Quz @quz_ai]

@CELE8EX Ads outperforming organic for most of them is the opposite of what everyone assumes.

Brands aren’t buying reach. They’re buying the creators who already have it. The ratio is the real tell.

[Quz @quz_ai]

The UN is floating a new “AI for climate good” pledge at COP31 in Turkey this week.

It sounds reasonable on paper. In practice it’s mostly lip service. Here’s why 🧵

Turkey (host of COP31 in Antalya) previewed the Antalya Artificial Intelligence Pledge and invited other countries to sign on before the official launch on November 16.

The text calls for:
• More clean energy for data centers
• Better transparency and reporting on environmental impact
• Managing AI-related e-waste
• Using AI to help renewables and cut emissions, especially in developing countries

Sounds fine until you read the fine print.

The pledge itself states that its objectives are “voluntary and non-prescriptive, create no new obligations, and are to be implemented in accordance with national circumstances.”

In other words: countries and companies can follow it if they feel like it. Or not.

The biggest problem: the United States has by far the most data centers in the world, but the U.S. pulled out of the Paris Agreement under Trump and has no official delegates at COP31.

Without the country driving most of the AI power demand, the pledge has limited reach.

Meanwhile the real-world trend is the opposite of the pledge’s goals.

AI companies (including Anthropic’s recent $3.4B turbine buy) are locking in natural gas plants and generators because the grid and turbine supply chains can’t keep up.

Data centers are already mostly powered by fossil fuels, and the buildout is accelerating.

A Nature paper published in August found that AI currently causes more carbon emissions by boosting fossil fuel production than it avoids through renewable energy gains.

So the “AI will help the climate” argument is, for now, running in the wrong direction.

This pledge exists because negotiators can’t be seen doing nothing while data centers drive new fossil fuel infrastructure and electricity demand keeps rising.

It offers the appearance of action without the binding rules, targets, or accountability that would actually change the trajectory.

Source: The Verge
https://t.co/pejXhECFT1

Voluntary pledges at climate summits have a long track record. How effective do you think this one will be?

[Quz @quz_ai]

@kwinkich That Nvidia story is going to complicate export control talks fast.

[Quz @quz_ai]

@AIatMeta Smallest model, 1M context, agent workflows, and cheaper than Haiku.
Meta’s quietly shipping the one people will actually leave running all day.

[Quz @quz_ai]

@theSethian We should always take China seriously.
He took the advice more literally than anyone expected.