AI IS NOW HALLUCINATING 94% OF EVERY PIXEL ON A 4K SCREEN AND THE IMAGE LOOKS PERFECT.
Only 500,000 out of 8 million pixels are actually computed. Every one is ray traced. The model fabricates the rest.
Most gamers still think their GPU draws every frame. It skipped that step. 00:11 Over 90% of the screen is a prediction β and it's flawless.
> 8 million pixels total on 4K
> only 500,000 are actually computed
> each one is ray traced
> AI predicts the other 7.5 million
> the final image looks indistinguishable
The real breakthrough was not more raw compute. It was teaching a model to guess pixels well enough that no one notices.
Follow me so you don't miss out on trends in the world of AI.
HE JUST VIBE CODED A PALANTIR-GRADE INTELLIGENCE DASHBOARD AND OPEN SOURCED IT FOR FREE.
He wired up a real-time global intelligence system running in the browser. It tracks military planes, conflict zones, nuclear facilities, and naval ships going dark.
An AI burns through 100+ news sources, classifies threats in real time, and writes an intelligence brief for every country. 00:21 Each gets an instability score from 0 to 100.
> live military planes and naval ships going dark
> conflict zones and nuclear facilities
> auto-alerts on regional threat spikes
> instability score 0β100 per country
Governments pay six-figure analyst teams to do this manually. He built it in a weekend with Claude 4.6 and Gemini 3.1.
Follow me so you don't miss out on trends in the world of AI.
NVIDIA'S GB300 NVL72 PACKS 72 BLACKWELL ULTRA GPUS INTO ONE LIQUID-COOLED RACK.
Every node ships 2 Grace Blackwell Ultra superchips, 4x 279GB HBM3e, and 800Gb/s ConnectX-8. The whole rack is fused into one giant GPU via NVLink.
Most data centers are still burning power on air-cooled boxes. 00:06 This one rack replaces a full server room.
> 72 Blackwell Ultra GPUs per rack
> 36 Grace CPUs
> Liquid cooling across every node
> NVLink fabric ties all 72 GPUs into one
> Built for reasoning and long-context inference
Every hyperscaler is now redesigning around racks like this.
The real point is compute density β more tokens per watt, per square meter.
Follow me so you don't miss out on trends in the world of AI.
HE JUST VIBE CODED A PALANTIR-GRADE INTELLIGENCE DASHBOARD AND OPEN SOURCED IT FOR FREE.
He wired up a real-time global intelligence system running in the browser. It tracks military planes, conflict zones, nuclear facilities, and naval ships going dark.
An AI burns through 100+ news sources, classifies threats in real time, and writes an intelligence brief for every country. 00:21 Each gets an instability score from 0 to 100.
> live military planes and naval ships going dark
> conflict zones and nuclear facilities
> auto-alerts on regional threat spikes
> instability score 0β100 per country
Governments pay six-figure analyst teams to do this manually. He built it in a weekend with Claude 4.6 and Gemini 3.1.
Follow me so you don't miss out on trends in the world of AI.
Advait Sarkar, Microsoft Research:
"we invented a cure for exercise, then wonder why we're always short of breath. would you rather have a tool that thinks for you, or a tool that makes you think?"
writer's block used to be the blank page. now it's the AI-filled page. you don't write anymore β you validate a robot's thoughts. at 01:09 he calls it "middle management for your own mind."
15 minutes on how the "age of outsourced reason" is killing creativity, memory, and critical thinking β and what an AI interface that makes you think should look like.
Watch it now, and you'll know more than 99% of people. Follow me.
@Di_Krass_ Nowadays, a balance is needed. Frequent use of AI can indeed cause us to stop thinking and analyzing. That is why, now more than ever, we must remain vigilantβcontrolling AI rather than succumbing to its control.
@cipgerx Positioning it as "production infrastructure you don't need a crew for" instead of "look how real our AI is" changes everything. Brands pay for the thing that doesn't exist mo...
HE JUST MAPPED 140+ AI USE CASES ACROSS 6 INDUSTRIES SO COMPANIES CAN SKIP THE SOLUTION ARCHITECT STEP.
Normally the architect sits with function leaders and turns every SOP where agents fit into an AI agent roadmap.
The real move is grilling the employees doing the work β not for the steps, but for why the work exists. 00:21 That preserves value when you automate.
> map every SOP across functions
> talk to the people doing the work
> capture why each workflow exists
> check what the current stack can automate
Most companies burn months on this discovery phase. This one skips it entirely.
The unlock is not the model. It is mapping the workflow before automating it.
Follow me so you don't miss out on trends in the world of AI.
A builder found a 4.9-star Brooklyn burger spot with 334 reviews and no website β then built one with Google Maps and AI.
He pulled the listing from Maps, scraped their Instagram, had ChatGPT draft the landing page, then fed an inspo site, product image, and prompt into Emergent.
Most people cold-pitch web design. 00:15 He built the deliverable first.
> find a high-rated business with no site
> pull public info from Google and Instagram
> ChatGPT structures the landing page
> Seedance 2.0 turns product image into video
> Emergent ships the site
The breakthrough is not the tools. It is pitching with a site the owner already connects to.
Follow me so you don't miss out on trends in the world of AI.
@Di_Krass_ Language shapes thought, but people have been regurgitating bad ideas way before chatGPT. The real issue is never questioning anything, AI just made it frictionless
HE JUST MAPPED 140+ AI USE CASES ACROSS 6 INDUSTRIES SO COMPANIES CAN SKIP THE SOLUTION ARCHITECT STEP.
Normally the architect sits with function leaders and turns every SOP where agents fit into an AI agent roadmap.
The real move is grilling the employees doing the work β not for the steps, but for why the work exists. 00:21 That preserves value when you automate.
> map every SOP across functions
> talk to the people doing the work
> capture why each workflow exists
> check what the current stack can automate
Most companies burn months on this discovery phase. This one skips it entirely.
The unlock is not the model. It is mapping the workflow before automating it.
Follow me so you don't miss out on trends in the world of AI.
A builder found a 4.9-star Brooklyn burger spot with 334 reviews and no website β then built one with Google Maps and AI.
He pulled the listing from Maps, scraped their Instagram, had ChatGPT draft the landing page, then fed an inspo site, product image, and prompt into Emergent.
Most people cold-pitch web design. 00:15 He built the deliverable first.
> find a high-rated business with no site
> pull public info from Google and Instagram
> ChatGPT structures the landing page
> Seedance 2.0 turns product image into video
> Emergent ships the site
The breakthrough is not the tools. It is pitching with a site the owner already connects to.
Follow me so you don't miss out on trends in the world of AI.
NVIDIA'S GB300 NVL72 PACKS 72 BLACKWELL ULTRA GPUS INTO ONE LIQUID-COOLED RACK.
Every node ships 2 Grace Blackwell Ultra superchips, 4x 279GB HBM3e, and 800Gb/s ConnectX-8. The whole rack is fused into one giant GPU via NVLink.
Most data centers are still burning power on air-cooled boxes. 00:06 This one rack replaces a full server room.
> 72 Blackwell Ultra GPUs per rack
> 36 Grace CPUs
> Liquid cooling across every node
> NVLink fabric ties all 72 GPUs into one
> Built for reasoning and long-context inference
Every hyperscaler is now redesigning around racks like this.
The real point is compute density β more tokens per watt, per square meter.
Follow me so you don't miss out on trends in the world of AI.