Chris

Singularity Enjoyer, on X to follow AI Leaders. Britisher in Bogota. I run @robomacro

[Chris @ChrizBogota]

@voluntas I worry about thermals, I really can't have a DGX die on me. I've been asking grok to create me some rules, where I lose 5% speed but get a decent drop in temperatures. It's not just peak temp but moving from hot to cold too fast

[V @voluntas]

ちゃんと性能が出ることを公開していくぞ。

[Chris @ChrizBogota]

@elshayib_ That is. A really shit site lol

[Elshayib @elshayib_]

You really can't hate this platform, brother got bullied just like that Lol

[ℏεsam @Hesamation]

One side is cheering for robots running in a stadium and the other is debating if curing cancer is worth accelerating for.

THE CONTRAST IS RIDICULOUS.

[Chris @ChrizBogota]

Most people have no idea how powerful grok bot is.
I have 8 bots running now and haven't even got deep into it. But the variety of things it's doing is as impressive as the calibre of the work. Best employee ever

[Lummox @Lummox_eth]

I gave Grok Bot 4 jobs and 12 hours, then stopped touching the workflow completely.

I had it research live information, monitor X, compare sources and organize the strongest findings while I worked on something else.

Instead of opening 20 tabs and prompting AI every 5 minutes, I gave it the objective once and came back to the results.

That’s when Grok Bot clicked for me.

For the last 3 years, AI meant opening a tab and asking for help. Now I’m more interested in how much work can happen while that tab stays closed.

The best agent might be the one I interact with the least.

[Chris @ChrizBogota]

@levelsio I fled.

[Chris @ChrizBogota]

Insane that poolside probably had $0 revenues then just made a $6bn deal

[Chubby♨️ @kimmonismus]

Nvidia is reportedly spending $6 billion to build one of the world’s most powerful open-weight AI models.

According to the WSJ, Nvidia will license Poolside’s technology and bring more than 100 of its employees into the Nemotron project.

Nvidia is also investing another $1 billion in Poolside at a $12 billion pre-money valuation.

The goal: challenge Chinese open-weight leaders such as DeepSeek and Kimi while competing directly with US frontier labs including OpenAI and Anthropic.

open source is the way to go. So good to see having NVIDIA on our side!

[Chris @ChrizBogota]

@Mister_Ch0c @0xIlyy $9,344 in Colombia

[Chris @ChrizBogota]

@0xIlyy Best rage bait tweet ever 😂

[Chris @ChrizBogota]

@0xkydo @gajesh I can't wait for you to extend this to Teslas, and AI Toasters etc in the future

[andrew chen @andrewchen]

2012: acquire company for users
2021: acquire company for engineers
2026: acquire company for training data

[TheDavidTai @TheDavidTai]

Stuff like this is why you should own your own compute, run your own models. Don't trust the and powerful providers change their terms on you. Many such cases.
https://t.co/7TZWoauCfx

[ア @yuruyurau]

a=(m,d=mag(k=9*cos(i*5)*sin(i),e=cos(i*3)*cos(i*2)*9)**3/1999+1.5-sin(t/2+m)**3/3)=>point(99*sin(c=d/16-t/48+m)+k*(p=d**sin(d*d-t+m))+200,99*sin(c*4)+e*p+200)
t=0,draw=$=>{t||createCanvas(w=400,w);background(9).stroke(w,96);for(t+=PI/20,i=1e4;i--;)a(i%16*13)}//#つぶやきProcessing

[⚡AI Search⚡ @aisearchio]

Remember, when a robot is chasing you down, run towards a wall and then make a sharp turn.

[Chris @ChrizBogota]

Are models plateauing in some way? If Astra is not a huge leap, what's the point

[Token Gremlin @TokenGremlin]

Andrew Ambrosino, an OpenAI employee and one of the company’s most active people on X, briefly posted that he was having a hard time choosing between Astra and GPT-5.6 Sol in Ultrafast mode.

Then the tweet vanished.

That is a VERY interesting comparison.

If Astra were only slightly better, I completely understand choosing Sol for the ridiculous speed.

But if Astra is the huge leap people are expecting, speed alone shouldn’t make the choice difficult.

So either:

→ Ultrafast Sol is absurdly good
→ Astra is more incremental than expected
→ or they’re optimized for very different kinds of work

Whatever the answer is, this is exactly the kind of breadcrumb that makes me think the next OpenAI lineup is going to be much more interesting than “new model replaces old model.”

[Chris @ChrizBogota]

@beffjezos GDP per square metre, Datacentres are no5. They are hugely desirable to host

[Beff (e/acc) @beffjezos]

What's cool about this whole datacenter controversy is that if someone brings up water as an issue you then immediately know that they are clinically retarded

[Jun Song @jun_song]

People who haven't actually used DGX Spark keep judging it purely by bandwidth numbers on paper.

For my ideal personal inference setup, durability and power efficiency come first. Except for something like the RTX 6000, most consumer GPUs just aren't built for sustained 24/7 loads, and they easily run up an extra $100+ a month on your power bill compared to a Mac or DGX. Taking everything into account, I only really consider Mac and DGX.

Sure, it used to be slower than high-bandwidth Macs before speculative decoding became a thing. But with techniques like DFlash, decoding speed gets surprisingly close to an M5 Max (614 GB/s).

On top of that, its raw compute is several times higher and it handles multi-batch way better. Once you start multi-serving, total token throughput actually surpasses the Mac. Plus, it stacks easily, so building a multi-node cluster for huge unified RAM is straightforward.

This is exactly why judging hardware purely by spec sheets without hands-on testing misses the whole point.

[Beff (e/acc) @beffjezos]

Who's funding this?