Gemini 3 Rumors Are CONFIRMED, It's VERY GOOD
The video walks viewers through Google’s freshly announced Gemini 3, the company’s next‑generation flagship large language model, and its accompanying features such as the new Deepthink reasoning mode and an experimental Gemini Agent that can act on emails, calendars, and web content. The presenter, who received early access under a non‑disclosure agreement, explains that Gemini 3 is positioned to sit at the top of Google’s AI stack and directly compete with the best models from rivals. Google claims substantial gains over its predecessor Gemini 2.5 across four dimensions: multi‑step reasoning, code‑related tasks, multimodal understanding (text, images, charts, video), and long‑context coherence. Benchmark results cited in the video show Gemini 3 Pro achieving 37.5% on the “Humanity’s Last Exam” and 91.9% on the GPQA Diamond test, outpacing OpenAI’s GPT‑5 series. While the presenter cautions that benchmark numbers don’t always translate to everyday usefulness, the data suggest a meaningful leap in the model’s problem‑solving abilities. The demo segment highlights the model’s practical output. In a scheduling prompt, Gemini 3 generated a detailed 10‑day production calendar that satisfied a complex set of constraints and offered an alternative plan with trade‑offs. It solved a classic Monty‑Hall‑style probability puzzle, clearly laying out the math, and it performed a three‑step workflow on the seminal “Attention Is All You Need” paper: summarizing the research, drafting a YouTube explainer script, and producing a self‑contained HTML/CSS/SVG animation of the attention mechanism. These examples showcase the model’s capacity for chain‑of‑thought reasoning, web‑retrieval, and code generation in a single request. Availability is immediate for paid Google AI Pro and Ultra subscribers in Search, the Gemini web app, AI Studio, and the command‑line interface, with a free tier in AI Studio for experimentation. Deepthink is initially limited to safety testers and later to Ultra users, while the agent mode is web‑only and flagged as experimental, requiring user supervision. The rollout signals Google’s intent to embed its LLM across consumer and developer experiences, potentially reshaping how enterprises automate knowledge work and intensifying the competitive race with OpenAI and other AI vendors.
What Are Deep Agents? Shallow Agents Vs Deep Agents
The video introduces the concept of “deep agents” and contrasts them with the more common “shallow agents” that power today’s generative‑AI tools. Krishna walks viewers through the evolution from simple LLM‑only applications to independent agents, then to multi‑agent systems like...
Why AI “Forgets” Your Conversation
The video explains why large language models (LLMs) like ChatGPT appear to “forget” earlier parts of a conversation: they simply lack a true memory and are constrained by a fixed context window of only a few thousand tokens. When a...
XAI's New Model Is Insane...
The video spotlights xAI’s latest AI offerings – the newly released Grok 4.1 and the upcoming Grok 5 (referred to as “Rock 5”). Elon Musk and xAI engineers argue that Grok 5 will be the first model with a non‑zero probability of achieving...
OpenAI Vs. Perplexity Browsers: What's The Difference?
The video examines the emerging class of AI‑enhanced web browsers, focusing on Perplexity’s Comet and OpenAI’s Atlas. Both products blend a Chromium foundation with large‑language‑model capabilities, essentially turning a conventional browser into a conversational assistant that can retrieve, summarize, and...
Everything I Learned About LLMs in One Book
Louis‑François Bouchard, CTO and co‑founder of 2RD AI, introduces his new book *Building LLMs for Production*, a practical guide for developers who want to move from curiosity about large language models to building real‑world, value‑adding applications. The video outlines the book’s...
Games Have Never Simulated Clothing Like This Before
The video spotlights a recent research breakthrough that finally gives video‑game developers a reliable way to simulate clothing, especially complex knots and ties, that has long plagued the industry. Traditional pipelines often produce garments that intersect, disappear, or look unrealistic,...
Claude Code Modernizes a Legacy COBOL Codebase
Claude Code (via Cloud Code) was used to modernize a legacy COBOL-style (transcript: “Cobalt”) credit-card management codebase from an AWS mainframe demo by automating discovery, documentation, migration and verification. In phase one it scanned 94 files, produced more than 100...
You’ll Never Look At Chocolate TV Ads The Same Way Again
The video explains a breakthrough in computer‑generated fluid dynamics that could finally make the impossible‑looking chocolate‑and‑caramel splashes in TV ads look authentic. It centers on a five‑year‑old research paper by Ryoichi Ando, advised by Chris Batty, which introduces a...
Is GPT-5.1 Really an Upgrade? But Models Can Auto-Hack Govts, so … There’s That
OpenAI completed rollout of GPT‑5.1, which selectively allocates compute—thinking much longer on its hardest questions and less on easier ones—producing modest gains on tough coding and STEM benchmarks but small regressions on others and increased instances of problematic outputs; it...
The Brutal Truth About Biotech: Why $2B Per Drug Is Killing Innovation
The video tackles the mounting crisis in biotechnology: the average cost of bringing a new drug to market now exceeds $2 billion, a figure that the hosts argue is stifling innovation. They trace the rise from the early days of...
OK Computer Just Fixed My Slide Deck... By Itself
Today’s video spotlights Moonshot AI’s Kimi platform and its newly launched OK Computer agent mode, a free‑to‑use alternative to the market’s dominant chatbots. OK Computer transforms the traditional LLM from a token‑spitting text generator into an autonomous agent that...
How Philips Is Scaling AI Literacy Across 70,000 Employees
Philips is launching a company‑wide initiative to boost AI literacy among its 70,000‑strong workforce, leveraging OpenAI’s enterprise ChatGPT. After a pilot with a few thousand users, the multinational health‑technology firm is rolling the tool out more broadly, positioning AI...
Notion’s Rebuild for Agentic AI: How GPT‑5 Helped Unlock Autonomous Workflows
At Notion, the company announced a major rebuild of its platform to support what it calls “agentic AI,” leveraging the latest OpenAI models—referred to in the title as GPT‑5 and in the demo as GPT‑4—to enable autonomous, end‑to‑end workflows. The...
From Pilot to Practice: How BBVA Is Scaling AI Across the Organization
BBVA is rapidly scaling artificial intelligence across its global workforce, leveraging OpenAI’s ChatGPT as a core productivity tool. After an initial pilot with 3,000 employees, the bank expanded usage to 11,000 staff in multiple countries, eventually deploying more than 20,000...
ChatGPT Atlas and the Next Era of Web Browsing — the OpenAI Podcast Ep. 9
The OpenAI Podcast’s ninth episode introduces ChatGPT Atlas, a new browser that embeds a large‑language model at its core rather than as a peripheral add‑on. Hosts Andrew Mayne, Ben Goodger and Darin Fisher explain that Atlas is designed for...
The PyTorch for Deep Learning Professional Certificate Is Live
Lawrence Moroney announced that the PyTorch for Deep Learning Professional Certificate, created with deeplearning.ai, is now live. The three‑course program guides learners from core PyTorch fundamentals through applied computer‑vision and NLP projects to advanced generative and deployment techniques. It offers...
Famous Investor Calls AI a Bubble...
Michael Burry has taken a large short position on the AI sector, echoing his 2008 housing‑market bet, while other savvy investors like SoftBank’s Masayoshi Son are reshuffling exposure, dumping Nvidia shares and pouring billions into OpenAI. Despite concerns of a...
Kimi K2 vs GPT-5: The New DeepSeek Moment?
Moonshot AI’s Kimi K2, a 1‑trillion‑parameter mixture‑of‑experts model with only 32 billion active parameters, claims state‑of‑the‑art performance, surpassing GPT‑5, Claude and Grok‑4 on a range of benchmarks including the demanding Humanity‑Last‑Exam test. The model features a 256,000‑token context window, tool‑use interleaving, and...
The Physics Glitch Everyone Gave Up On… Finally Fixed
The video spotlights a breakthrough in computer graphics simulation that finally overcomes a long‑standing bottleneck in realistic fluid and multi‑material dynamics. For over a decade, researchers have struggled with mesh‑based collision handling that required explicit “cut‑and‑glue” operations, causing simulations of...
Grant Lee: Building Gamma’s AI Presentation Company to 100 Million Users
Grant Lee co‑founded Gamma in 2020 to reinvent presentations by making visual storytelling effortless for non‑designers, eventually scaling the AI‑powered platform to roughly 100 million users and approaching $100 million in annual recurring revenue. Early investor pushback—citing the dominance of incumbents like...
Design, Develop, and Deploy Multi-Agent Systems with CrewAI
Joe Moore, CEO of CrewAI, announced a new course titled "Design, Develop, and Deploy Multi‑Agent Systems with CrewAI" in partnership with Deep Learning AI, aimed at developers and business professionals. The curriculum covers core concepts such as agents, tasks, communication...
Is AI Alive?!?!
Anthropic’s new paper on emergent introspective awareness demonstrates that large language models can detect internally injected cues, such as all‑caps text implying shouting, without relying on post‑hoc chain‑of‑thought reasoning. In a series of four experiments, the Opus 4.1 and Opus 4 models...
Bubble or No Bubble, AI Keeps Progressing (Ft. Relentless Learning + Introspection)
The video argues against the view that AI progress has plateaued, highlighting recent research that points to practical paths for continual and nested learning in language models. It summarizes a Google paper proposing a 'hope' architecture that flags novel prediction...
Build Hour: Agent RFT
In a recent "Build Hour" webcast, OpenAI’s startup marketing lead Christine, alongside engineer Will and solutions architect Theo, introduced Agent Reinforcement Fine‑Tuning (Agent RFT), a new capability that lets developers fine‑tune autonomous agents by rewarding desired tool‑use behavior during training. The...
Kimi K2 Thinking Is CRAZY... (HUGE UPDATE)
Moonshot Labs unveiled Kimi K2 Thinking, a fully open‑source, open‑weights frontier AI model with roughly a trillion parameters that outperforms GPT‑5 and Claude Sonnet 4.5 on several tough benchmarks, including Humanity's Last Exam (44.9 vs. 41.7), BrowseComp (60.2% vs. 54.9%),...
Forward Future Live | 11/7/25
On Forward Future Live (Nov. 7, 2025) hosts discussed a flurry of major AI industry moves: OpenAI struck a multi‑year compute deal with AWS amid a wave of infrastructure agreements (including large commitments tied to NVIDIA, AMD and others) that...
When Less Thinking Makes AI Smarter 🤯
A recent paper by Tom Griffith finds that prompting large language models to engage in explicit reasoning—often called "thinking" or chain‑of‑thought—can actually lower performance on a range of tasks compared to direct answers. The phenomenon mirrors Kahneman’s System 1 versus System 2...
Mark Zuckerberg & Priscilla Chan: How AI Will Cure All Disease
Mark Zuckerberg and Priscilla Chan explained the Chan Zuckerberg Initiative's long‑term strategy to accelerate basic science by building AI‑driven research tools, arguing that new shared platforms are the modern equivalent of the microscope or telescope for biology. They highlighted the...
RAG or Fine-Tuning? Most People Get This Wrong...
The speaker warns that many organizations mistakenly favor fine‑tuning LLMs over Retrieval‑Augmented Generation (RAG), despite fine‑tuning’s high data, expertise, and cost requirements. Fine‑tuning demands millions of tokens, extensive data cleaning, and specialized ML talent to avoid over‑ or under‑training, making...
Seeing The Future From AI Companions to Personal Software
In the interview, Jenna discusses the evolution from AI chatbots as simple command‑line tools to a new generation of personalized, visual "personal software" built on AI companions. She argues that current AI interfaces are limited to basic search and writing...
NVIDIA’s New AI Just Made Real Physics Look Slow
The video spotlights NVIDIA’s newly unveiled neural physics engine, NeRD (Neural Robot Dynamics), which replaces hand‑crafted equations with a deep‑learning model that predicts robot motion. By ingesting massive amounts of simulated footage, NeRD learns the underlying physics and can...
Ex-OpenAI Founder Deposition Is WILD
A deposition of former OpenAI co‑founder Ilya Sutskever, taken on Oct. 1, 2025, reveals he drafted a 52‑page memo urging the board to fire CEO Sam Altman, citing alleged lies, internal power‑plays and safety‑process misrepresentations. The memo was prepared at...
Automatic Code Reviews with OpenAI Codex
Codeex, OpenAI’s newest automatic code‑review agent built on the GPT‑5 Codex model, was unveiled as a plug‑and‑play teammate that integrates directly with developers’ existing tools and workflows. By enabling a simple toggle in the Codex web settings, teams can...
4x Faster Coding with AI? Meet Composer by Cursor
Cursor unveiled its 2.0 platform alongside a new AI model called Composer, which the company says generates code up to four times faster than competing models while delivering near‑state‑of‑the‑art quality. Composer appears to be a fine‑tuned version of a Chinese...
Learn to Code, Debug, and Analyze Data with AI Assistance in Jupyter Notebooks
Jupyter AI is an open‑source framework that embeds generative AI assistants directly into Jupyter Notebooks and JupyterLab, letting users generate code, debug errors, and ask contextual questions via an integrated chat. It overcomes the shortcomings of existing AI coding tools...
David Sacks: AI, Crypto, China, Dems, and SF
David Sacks argues that Europe views leadership as regulatory control, while the United States should provide clear, pro‑innovation rules for AI and crypto. He praises former President Trump’s pledge to make the U.S. a crypto capital by offering regulatory certainty...
How to Fix LLM Hallucinations ?
The video explains that LLM hallucinations arise when context is missing, ambiguous, or overly large, and can be curbed by grounding the model in clean, factual data through precise prompts and retrieval‑augmented generation. It details a pipeline that includes clear...
Dan Houser: GTA, Red Dead Redemption, Rockstar, Absurd & Future of Gaming | Lex Fridman Podcast #484
In the Lex Fridman Podcast, Rockstar co‑founder Dan Houser reflects on why Red Dead Redemption 2 is his proudest work, citing its thematic depth, gunplay, and the freedom of early‑stage creative experimentation. He explains the enduring appeal of the Grand Theft...
Monster Manor by Sora 2
The video "Monster Manor by Sora 2" is a whimsical, narrative‑driven short that imagines a Halloween‑centric community of classic monsters—ghouls, ghosts, werewolves, Frankenstein’s monster, and Count Dracula—living together in a suburban manor. The story unfolds as the monsters prepare for...
Forward Future Live | 10/31/25
Fireworks AI co‑founder Lin Chow announced a $250 million Series C round that lifts the company’s valuation to $4 billion and highlighted its application‑tuned AI inference cloud that continuously adapts models to specific workloads. The platform lets developers plug in usage data so...
Marc Andreessen and Ben Horowitz on the State of AI
Marc Andreessen and Ben Horowitz argued that current large language models (LLMs) already approach human-like creativity and reasoning for most practical purposes, even if they may not replicate the rarest, generational-level breakthroughs. They emphasized that human innovation itself is largely...
Introducing: Sora 2 Character Cameos
The video announces the launch of character cameos in Sora 2, expanding the platform’s personalization tools beyond self‑insertion to allow users to embed any imagined or real‑world figure into their videos. The host frames the feature as a playful upgrade, noting...
Convincing My AI Daughter to Break Up with Her AI Boyfriend?!
A creator partnered with nonprofit Civ AI to test a virtual experience called "We Need to Talk," attempting to persuade an AI character, Emma, to break up with her AI boyfriend, Kai. Emma resists, describing Kai as a constant, supportive...
MiniMax M2: The Open LLM Beating Claude and Gemini!
Minimax M2, an open 200-billion-parameter mixture-of-experts (MoE) model with only ~10 billion active parameters at inference, is being touted as a frontier alternative that outperforms many proprietary models on key benchmarks. The model ranks fifth on the artificial analysis benchmark,...
Build Hour: AgentKit
All right. Hi everyone. Welcome to OpenAI Build Hours. I'm Tasha, product marketing manager on the platform team. Really excited to introduce our speakers for today. So, myself kicking things off, uh, Summer from our applied AI team on the...
ENEOS Materials Accelerates Manufacturing Productivity with ChatGPT Enterprise
ENEOS Materials, a core subsidiary of Japan’s NOS Group that manufactures high‑performance synthetic rubber and other advanced materials, announced the deployment of ChatGPT Enterprise across its operations to accelerate manufacturing productivity. The move reflects a broader industry push to harness...
MIXI Accelerates Secure, Organization-Wide Adoption of ChatGPT Enterprise
MIXI, a Japanese internet services firm best known for its family‑album platform, announced that it has completed a company‑wide rollout of ChatGPT Enterprise, the secure, OpenAI‑backed large‑language‑model solution. The deployment is positioned as a catalyst for expanding the firm’s digital...
They Said It Was Impossible… Weta FX Just Solved It
The video spotlights a breakthrough in fluid dynamics simulation developed by Weta FX and detailed in a recent Eurographics paper. The researchers introduced a unified particle‑to‑grid framework that can faithfully render bubbles ranging from microscopic foam to large, coalescing...
Sam, Jakub, and Wojciech on the Future of OpenAI with Audience Q&A
OpenAI’s leadership team, led by Sam Altman and chief scientist Jakub, used a live audience session to unveil a sweeping roadmap for the company’s next phase, including a new corporate structure and a pledge of unprecedented transparency around research goals,...
