Many are asking whether NVIDIA at $5 trillion is too much. But if Jensen's numbers yesterday are remotely true, he guided $400 billion in revenues for 2026 (without counting on gaming GPUs or non-operational revenues), which at NVIDIA's current net margin (will fall but alas),

"We are being told that AI models are getting better, that AGI is close; yet we are getting worse and worse models for the same price. This tells us two things: we aren’t getting AGI soon, and that ‘fair’ is a charged word." - @TheTechOasis1

This is exactly right. @TheTechOasis1

Why Most AI Products are Scams by @TheTechOasis1

The amount of cringe-posting from the “I had a few hours to be play around with GPT-5 and it <insert exaggeration>” AI influencer crowd that’s inundating my timeline tells me the release is probably waaaay less impressive than I would have hoped. It’s probably great, it’s also

What if you could not only watch a generated video, but explore it too? 🌐 Genie 3 is our groundbreaking world model that creates interactive, playable environments from a single text prompt. From photorealistic landscapes to fantasy realms, the possibilities are endless. 🧵

Energy-Based AIs, Too Good to be True? by @TheTechOasis1

AI, Education, & Jobs. We Need To Scrap Everything. https://t.co/mT6e9dCNmP via @TheTechOasis1

Meet Kimi-Researcher - an autonomous agent that excels at multi-turn search and reasoning. Powered by k 1.5 and trained with end-to-end agentic RL. Achieved 26.9% pass@1 on Humanity's Last Exam, 69% pass@1 on xbench. 🔗 Tech blog： https://t.co/lNC0HKrndr

What If We Are All Wrong About AI? by @TheTechOasis1 https://t.co/K25ey43CXv #ArtificialIntelligence #Technology #DataScience #MachineLearning #Future

Apple’s Viral AI Paper. Reality or Fraud? by @TheTechOasis1

Been heavily testing MCP with local and proprietary models, and I'm worried about memory complexity. As long as the cache has high memory complexity, preventing larger context windows, even the simplest of MCP use cases saturate the context window before I can do anything (this

frontier ai today

MatFormers are very powerful alternatives to transformers. Similar to a regular transformer, but after training, you can split up the model to any size you like and get very strong performance that scales just like a regular transformer. So train once, get models of all sizes!

España tiene las segundas mayores reservas de uranio de la Unión Europea continental, como documenta el Libro Rojo del Uranio de la Agencia Internacional de la Energía. https://t.co/KqtktnmWlX Sin embargo, la extracción de minerales radiactivos como el uranio está prohibida por

The Second ‘ChatGPT Moment’ is Here by @TheTechOasis1

an AI geek-out extraordinaire for a Saturday morning... "Here’s NVIDIA and AI’s Future in 5 Minutes." by @TheTechOasis1

“Jensen Huang, NVIDIA’s CEO, gave a two-hour keynote yesterday covering NVIDIA’s future roadmap in the GTC conference, the ‘AI Woodstock.’ Three years’ worth of information was said, but using terminology and jargon that escaped most…” — @TheTechOasis1

@OpenRouterAI LLM usage leaderboard is here:

Deep Research, from both Google & OpenAI, marks the first time I am completely 99.9999% convinced that human level intelligence is truly unreachable using current artificial intelligence (AI) technology. Its time to rethink what AI is.