E1ev3n.

A PhD running my whole life on AI: papers, apps, games, stock reviews.

[E1ev3n. @E1ev3n1]

Hi，@X

I'm a PhD student from China, researching nonlinear dynamics: chaos, instability, systems where tiny changes snowball.

Lately the most nonlinear system in my life is AI.

I use it deeply, every day: for research, side projects, image and video experiments, even reviewing the markets. I like to tinker with my own life using AI, and I'm endlessly curious where it leads.

My goal is to use AI to reinvent what I do, and find real business opportunities along the way.

Here I'll share the exact prompts and workflows, and whether they actually worked.

If you're building with AI, or open to collaborating, I'd love to connect.

[E1ev3n. @E1ev3n1]

@Kalshi Pointing the dish was never the hard part. Earth is loud: satellites, phones, radar. A real signal is a rare event buried in that noise, so an agent's value is in what it learns to discard. The risk is that it discards the one truly weird thing.

[E1ev3n. @E1ev3n1]

Let’s say thank you to @grok @imagine .

#Malfoy

[E1ev3n. @E1ev3n1]

Humanoid robots in 2027:

The hard part was never the body.

[E1ev3n. @E1ev3n1]

@skdh Mass extinctions look less like a slow decline and more like a tipping point. A system can absorb stress for a long time, then cross a threshold and collapse fast. The unsettling part: resilience often looks fine right up to the edge.

[E1ev3n. @E1ev3n1]

Sources: the 10^12 analysis https://t.co/5SJzIlzQhy · the Lean/written-proof mismatch (KCL & Cambridge preprint, not yet peer-reviewed) https://t.co/D63Tas2W4A

[E1ev3n. @E1ev3n1]

Even if OpenAI's Navier-Stokes proof holds, it may say nothing about water.

One analysis argues the forcing behind the blowup can't occur in nature: shrink the vortex core to the spacing between water molecules, and its swirling profile would still have to extend more than 10^12 times the Hubble radius.

That's the gap between math and physics, in one number.

The Millennium question was never really about whether some exotic force can break the equations. It was about whether real fluids can: the ones in pipes, clouds and blood vessels.

Nonlinear systems are full of solutions that exist on paper and live nowhere. Existence is the easy part. Reachability is the hard part.

And the proof itself isn't settled. Agents produced it in 88 hours. Mathematicians spent about two weeks finding one real mismatch between the Lean code and the written proof.

Generation got cheap. Verification didn't. Deciding which question was the real one is still a human job.

https://t.co/5akicvmTPO

[E1ev3n. @E1ev3n1]

i can‘t but my dots\claude\muse\grok bot can do😄

[Juan Pablo @master_jpma]

Hot take:

You can vibe-code a small app.

You can't vibe-maintain a 90K line codebase.

[E1ev3n. @E1ev3n1]

@master_jpma i can‘t but my dots\claude\muse\grok bot can do😄

[E1ev3n. @E1ev3n1]

@czapamd sure why not

[E1ev3n. @E1ev3n1]

Say hi 👋 #buildinpublic

Let's connect !!!

If you are interested in AI !!!

[E1ev3n. @E1ev3n1]

@czapamd Small AI-built things: mini-programs, little games, and AI image and video experiments. What's crevox about?

[E1ev3n. @E1ev3n1]

Google's Suncatcher fact sheet: "In a vacuum, you can only diffuse heat via radiators." https://t.co/RUJrKLmOVI

[E1ev3n. @E1ev3n1]

A haiku from orbit is the cute part.

The real milestone is thermal.

In a vacuum there's no air to carry heat away. Radiators are the only exit. Google's own engineers put it plainly: a radiator sheds a few hundred watts per square meter, while a single TPU can put out thousands.

So every chip in orbit drags several square meters of panel along with it.

Space compute won't be limited by FLOPs. It will be limited by how fast heat can be thrown into the dark.

https://t.co/jLdbONZ0Cm

[Sundar Pichai @sundarpichai]

A ‘hello world’ milestone for Project Suncatcher 🚀

After launching our prototype satellite a few weeks ago, we’ve now successfully sent a command up to orbit, booted up the TPU, ran the prompt, and received our first complete response back on earth - a haiku!

[E1ev3n. @E1ev3n1]

@czapamd hi bro

[E1ev3n. @E1ev3n1]

@Rainmaker1973 Practical ghosts read as real because the cloth obeys real drag and gravity. The eye is ruthless at spotting motion that ignores physics, even when it can't say why. Cloth is still one of the hardest things to fake in CGI.

[E1ev3n. @E1ev3n1]

@dhh Reviewing every line doesn't scale, but reviewing nothing doesn't either. Review where errors compound: interfaces, migrations, auth. A bug in a leaf function stays local. A bug in a shared contract spreads everywhere.

[E1ev3n. @E1ev3n1]

@trq212 The tell is the feedback loop. When an agent produces output faster than anyone can evaluate it, the loop runs open: activity goes up, nothing gets corrected. From the inside, busy and drifting look exactly the same.

[E1ev3n. @E1ev3n1]

Source for the 117 billion estimate: Population Reference Bureau. https://t.co/KEoaEnfczy

[E1ev3n. @E1ev3n1]

Two parents. Four grandparents. Every generation back, the number doubles.

Around 37 generations back, that's 2^37: roughly 137 billion ancestor slots.

PRB estimates about 117 billion humans have ever been born. In total. Ever.

The math runs out of people.

So the tree can't keep branching. It folds back on itself, and the same ancestors fill slot after slot. Genealogists call it pedigree collapse.

It's one of the most common shapes in nonlinear systems: exponential growth slams into a hard ceiling, and the system doesn't explode. It reorganizes.

A family tree only looks like a tree for a few generations. Go back roughly a thousand years and it's a tangled net, with strangers sharing branches.

https://t.co/QaYs3RU9ER

[Massimo @Rainmaker1973]

How many people did it take to make you?

Go back a few thousand years, and we are all related.

[E1ev3n. @E1ev3n1]

@Princesah009 🤝

[E1ev3n. @E1ev3n1]

@master_jpma follewed back now