NVIDIA just changed the game for Windows laptops.
Jensen Huang took the stage at Computex and unveiled the RTX Spark Superchip
It packs a 20-core Grace CPU, a full Blackwell RTX GPU with 6,144 CUDA cores, up to 128GB of unified memory, and hits 1 petaflop of FP4 AI performance.
All in a single package that fits into chassis as slim as 14mm and sips 45-80W.
The claims are wild:
100+ FPS at 1440p in Forza Horizon 6… on battery. Local 120B-parameter AI models. Full RTX and CUDA support.
And it’s designed from the ground up for personal AI agents that actually feel like smart teammates instead of clunky tools.
@HealthRanger Your podcast @HealthRanger was excellent yesterday where you covered the shortcomings of NVIDIA hardware. I too have seen many small accounts promoting the hardware lately.
What other hardware are you are looking at? Is it GMKtec EVOX2 or AMD Ryzen AI Halo? Any suggestions?
JENSEN HUANG ACABA DE ROMPER EL MODELO DE NEGOCIO DE OPENAI EN EL ESCENARIO.
Presentó una computadora de IA de $249 que elimina tu factura de $200/mes.
El video ya tiene 217,000 likes.
Se llama Jetson Orin Nano Super.
• 70 TOPS de rendimiento IA • 25W de consumo • más pequeña que una billetera
Y ejecuta Llama 3, Mistral, Gemma y DeepSeek
🔥 LOCALMENTE 🔥
Sin APIs. Sin límites. Sin enviar tus datos a la nube.
Hoy, un desarrollador promedio paga ~$200/mes a OpenAI para automatizaciones, agentes y asistentes de código.
La misma carga de trabajo en esta caja:
→ ~$2/mes en electricidad → hardware amortizado en ~10 semanas
La configuración da miedo de lo simple:
Instalas Ollama
Cambias UNA línea de código
Apuntas a localhost en vez de OpenAI
Listo. Todo sigue funcionando igual.
Y aquí está la parte importante:
Los modelos de 7B ya cubren el 80% de lo que la gente usa ChatGPT para:
• resumir • redactar • programar • Q&A sobre documentos • automatizaciones
Solo el 20% más pesado necesita Claude o GPT-4.
Resultado:
Tu costo mensual cae de $200 → ~$22.
Mientras tanto:
• las APIs suben de precio • los rate limits empeoranm • y cada vez más gente depende de infraestructura ajena
Los que monten IA local en 2026van a parecer adelantados a su tiempo en 2027.
📁 Guarda este hilo.
El cambio ya empezó. 👇
@KarelMercx@goldseek That provisional revenue is then adjusted (often several times) based on second, third provisional invoices that true up for future metal prices and assaying of the shipment on arrival at destination smelter/refiner.
@KarelMercx@goldseek That is correct. For any metal sold in concentrate form that is in-transit at reporting date, the company estimates the month that payment will be received and revenue for that shipment is recorded provisionally based on the estimated month of payment futures price.
Some of the Best Precious Metals Prices in the Industry with Bob Coleman.
Why investors pay more to buy and get less to sell, they need to shop around.
Crazy how we can kidnap Maduro and his wife to stand trial in the US but Trump and co. can't get a single meaningful arrest for COVID, voter fraud, Russia hoax, Epstein, Benghazi, J6, Biden corruption, Clinton corruption, Obama corruption, or literally ANYTHING.
@HealthRanger Hi Mike. I listen to your show almost daily and I am a big fan. I don't disagree.
The issue is this: How much did you pay to a third party that developed the AI tools to do the coding you are describing & how does that compare to the cost they paid to develop it? Thats the gap.