AI Post — Artificial Intelligence

Gemini 3.0 Pro is now rolling out in AI Studio. AI Post 🪙 | Our X 🥇

🧠 Google unveils Gemini 3 — a new leap in AI capabilities A system map of Google’s next-gen Gemini 3 has surfaced online — and early benchmarks suggest it outperforms today’s top models across the board. • Beats ChatGPT and Claude in math, coding, reasoning, and business-case tasks • Can generate full games, interactive websites, and complex simulations • Shows major improvements in tool use, planning, and multi-step execution Google traditionally rolls out new models for free on AI Studio and insiders say the official Gemini 3 release is expected today. AI Post 🪙 | Our X 🥇

Anthropic's CEO just warned that AI could eliminate 50% of entry-level jobs within five years. Junior consultants, paralegals, financial analysts, he's predicting 10-20% unemployment. According to Amodei, Claude writes 90% of the code at Anthropic. Their 300,000 business customers are increasingly using Claude not as assistance, but as the worker. Entry-level positions aren't just jobs. They're how you learn to do senior-level work. If AI absorbs that layer entirely, we're not just eliminating jobs, we're eliminating the path to expertise itself. What happens to entry-level white-collar workers by 2030? AI Post 🪙 | Our X 🥇

OpenAI CEO Sam Altman and his husband are funding the creation of genetically engineered babies By merging AI with CRISPR, they aim to “fix” humans while offshoring early embryo testing to evade oversight. Would y’all agree to edit your child using Sam's tool? 🤔 AI Post 🪙 | Our X 🥇

Jeff Bezos is reportedly throwing his money and time into an artificial intelligence start-up that he will help manage as its co-CEO. The AI company, called Project Prometheus, is coming out of the gates with $6.2 billion in funding This is the first time Mr. Bezos has taken a formal operational role in a company since he stepped down as chief executive of Amazon in July 2021. AI Post 🪙 | Our X 🥇

⚡️ Heretic: automatic uncensoring for modern LLMs A new tool called Heretic takes last year’s research on “refusal directions” in language models and turns it into a full, automated uncensoring pipeline. How it works: • Researchers found that LLM refusals come from one specific direction in the activation space • Heretic computes that direction by comparing harmful vs harmless prompts • Then it orthogonalizes attention + MLP projection weights to scrub that refusal vector out • An optimizer tunes parameters to minimize refusals while keeping KL-divergence low You run it once, wait ~45 minutes, and walk away with an uncensored version of almost any model, Llama, Qwen, Gemma, and even some multimodal architectures. No manual patching, no jailbreaks, no prompt tricks, it modifies the model itself. If this approach scales, open-source models like gpt-oss could end up far more flexible (and far less moderated) than anything offered commercially. AI Post 🪙 | Our X 🥇

⚠️ Peter Thiel dumps all his Nvidia shares — more “smart money” steps back Billionaire Peter Thiel has sold 100% of his Nvidia position, about 537K shares, which made up 40% of his entire fund. He rotated heavily into Microsoft, Apple, and partially Tesla, citing the same concern he’s repeated for months: AI valuations look like a bubble without real economic footing. And he’s not alone. What’s happening: • Thiel exits completely — calls NVDA “overheated” • SoftBank is reportedly selling its entire Nvidia stake • Michael Burry is shorting Nvidia with 14% of his portfolio When three investors with very different playbooks all move in the same direction, it’s not noise, it’s a sentiment shift. Not a crash call, but definitely a sign that the “AI can only go up” conviction is weakening. The real question now, is this the first crack in the AI trade, or just early profit-taking before the next leg up? AI Post 🪙 | Our X 🥇

🔔 Google is working on multi-agent systems to help you refine ideas with tournament-like evaluation. Each run takes around 40 minutes and brings you 100 detailed ideas on a given research topic. 2 new multi-agents are being developed for Gemini Enterprise: - Idea Generation - "Create a multi-agent innovation session" - Co-Scientist - "Drive novel scientific discovery with Co-Scientist" Co-Scientist 3-step workflow: - Tell Co-Scientist what you plan to research, point it to relevant data, and set your evaluation criteria. - A team of agents will generate ideas on your topic using their available data - The agents will evaluate the ideas against your criteria and rank them, tournament-style Google is not only automating research but also preparing a product that will enable others to do so. Full Scoop. AI Post 🪙 | Our X 🥇

🔔 Update to LeCuns departure from Meta • Yann LeCun is reportedly leaving Meta Platforms because he’s lost confidence in large language models as the path to human-level AI - he calls them a “dead end”. • He argues instead for so-called world models: systems that build an internal, causal understanding of the physical and action-based world (not just text) and can predict the results of actions. • His vision: machines that plan, reason, and act in hierarchies with measurable “energy functions” (i.e., cost/compatibility functions) rather than just generating next tokens - though it’s still very speculative, expensive, and long‐term. Source. AI Post 🪙 | Our X 🥇

xAI just released Grok 4.1 and Grok 4.1 Thinking beta models. AI Post 🪙 | Our X 🥇

AI Scientist Breaks Discovery Barrier This is just fantastic: Kosmos is presented as a breakthrough “AI Scientist” that can autonomously read thousands of papers, analyze tens of thousands of lines of code, and generate scientific insights that normally take humans months - sometimes even producing entirely new discoveries. It uses structured world models to reason over millions of tokens, letting it synthesize complex scientific domains with unprecedented depth. The launch is framed as a turning point: AI moving from passive assistants to active engines of real scientific progress. 2026 will be the year of scientific discovery and breakthroughs. Source. AI Post 🪙 | Our X 🥇

Sam Altman says AI will become so integrated into our lives that we stop talking about it Just like internet companies and mobile companies faded once everyone adopted those technologies. In 5-10 years, every product will be smart by default, "less hype, and more capability" AI Post 🪙 | Our X 🥇

This is freaking cool NotebookLM just launched its “Deep Research” mode - which can automatically browse hundreds of websites, synthesize the findings into a structured report, and include an annotated list of sources. This means that instead of piecemeal searches and manual organization, users can now let the tool take care of the heavy lifting for in-depth research. AI Post 🪙 | Our X 🥇

Researchers have created a microrobot to fight diseases inside the body "Scientists can control the device, which is less than two millimetres in width, through blood vessels to deliver essential drugs." Source. AI Post 🪙 | Our X 🥇

“Black Mirror” in real life — a startup launches AI copies of dead relatives 2Wai has unveiled an AI that imitates conversations with the deceased. In their promo video, a digital “mom” supports her daughter during pregnancy, and years later chats with her grown grandson, who tells her about the birth of his own child. Viewers weren’t impressed, many called it creepy and emotionally disturbing rather than a technological breakthrough. What do you think? AI Post 🪙 | Our X 🥇

Inside Microsoft's Fairwater datacenter. Their goal is to eventually scale AI workloads across datacenters using similar methods as employed to distribute high-performance computing and AI workloads across multiple servers today. AI Post 🪙 | Our X 🥇

Delivery robots in Moscow … now own the streets AI Post 🪙 | Our X 🥇

Jeff Bezos basically said that if the AI economy keeps driving energy demand higher then Amazon will eventually need data centers in space because Earth will not supply enough power. AI Post 🪙 | Our X 🥇

A robot tries golf and ... it's a hole in one So now the robots have beaten humans at golf too. Is there any sport where humans are still ahead? AI Post 🪙 | Our X 🥇

🔒 China unveils quantum-speed breakthrough China’s CHIPX lab claims to have built the first scalable, industrial-grade optical quantum chip that can run certain AI workloads up to 1,000× faster than Nvidia GPUs. The photonic chip packs over 1,000 optical components onto a 6-inch wafer, scales to “1 million qubits,” and can be deployed in weeks instead of months - though production remains limited at ~12,000 wafers per year. If the performance claims hold, it marks a major step in China’s race for quantum advantage, as Western firms like Nvidia push similar optical-quantum tech but haven’t yet shown comparable scale! Source. AI Post 🪙 | Our X 🥇