Today's AI Pulse
Pinterest inks $4 billion AWS cloud pact to boost AI visual search
Pinterest announced a $4 billion multi‑year agreement with Amazon Web Services, the largest infrastructure deal in its history, to power AI‑driven visual search and discovery through 2031. The partnership will expand use of AWS Trainium chips for large‑language and vision‑language models and migrate Pinterest workloads to a Kubernetes‑based architecture. The deal also increases reliance on AWS Graviton processors.
Also developing:
Proactive Agents for the Web with Devi Parikh - #756
In this episode, Devi Parikh, co‑founder and co‑CEO of Yutori, explains how proactive web agents can transform browser interactions by using visually‑grounded models that operate on screenshots instead of the fragile DOM, resulting in more robust handling of complex interfaces. She details Yutori’s evolving training pipeline—from supervised fine‑tuning to rejection sampling and reinforcement learning—and how its “Scouts” architecture coordinates multiple sub‑agents and tools to execute sophisticated web tasks. The conversation also covers the importance of ambient, background operation for these agents and outlines the roadmap from simple monitoring to full‑task automation on the web.
Hugging Face CEO Says We’re in an ‘LLM Bubble,’ Not an AI Bubble
Hugging Face CEO Clem Delangue told an Axios event that the current hype surrounds large language models (LLMs) rather than AI as a whole, warning that the "LLM bubble" could burst as early as next year. He argued that LLMs...
AI Deal Web Too Complex, Need Clear Visualization
I keep track of and talk about all these AI deals for a living, and even I have whiplash. Anyone have a great visualization of all the investments and deals between the labs, cloud, and AI infrastructure companies? eg Google owns...
ChatGPT Users Crave Spotify‑Wrapped Style Memory Recap
I need ChatGPT and Claude to start releasing an opt-in Spotify Wrapped ASAP. Memory is probably my favorite feature, but it needs a jolt of fun and delight. --- EXAMPLE: "You asked how to fix things 15 times. 10 of those were...
From Automation to $1.3B: Scribe’s AI Success Story
In this episode Jamie and Jaeden break down how Scribe achieved a $1.3 billion valuation by building AI‑powered workflow automation tools for enterprises. They explore how individuals can turn their expertise into AI consulting gigs and stress the growing importance of...
Gartner‑backed Vectara: Must‑see Demo for GenAI Enterprises
The most important thing to take away from Gartner's recognition of @Vectara in two of their Magic Quadrants: If you an enterprise building anything that leverages GenAI or Agentic AI, get a briefing and demo from Vectara on what they...
Gemini 3 Turns Any Image Into Interactive Experience
With Gemini 3, you can take an image of anything, like a board game, a floor plan, or a doodle, and make it an interactive experience. This is because of Gemini 3’s deep multimodal understanding. The model is able to analyze...
Perplexity Brings Getty Images Into AI Search — Here’s Why It Matters
The episode explains Perplexity’s new partnership with Getty Images, which integrates high‑quality, royalty‑free visuals directly into the AI‑driven search experience. It highlights how the collaboration gives users richer, context‑relevant images while creating fresh licensing and revenue streams for photographers. The...
Scaling AI Agents Drives Transformative Business Results
Fewer than 10% of orgs are scaling AI agents in any one business function. But they're the ones seeing transformative results. McKinsey & Company 's new State of AI report shows that while 62% of companies are experimenting with agents,...
Gemini 3 Shows Tier‑1 Performance, Yet Benchmark Gaming Persists
I played with Gemini 3 yesterday via early access. Few thoughts - First I usually urge caution with public benchmarks because imo they can be quite possible to game. It comes down to discipline and self-restraint of the team (who is...
The Future of Google Maps: AI Features Every User Should Know
In this brief episode, hosts Jaeden and Jamie break down the latest AI-powered enhancements to Google Maps, highlighting the integration of Gemini and Vertex AI to boost navigation accuracy and personalized recommendations. They explain how these features streamline user experience...
Poolside AI Acquires Fern Labs, Forming Full‑stack Frontier AI Powerhouse
news! @poolsideai has acquired @fern_labs, in which @airstreet was the first and sole investor - and a first investor into poolside as well. i met @ashedwardst in late 2024 before fern had a name or team. his technical depth, product taste...
AI Forces Wikipedia to Charge for Access—Here’s What It Means
Jamie and Jaeden dissect Wikipedia’s rollout of a paid API that forces AI companies to pay for access to its content, highlighting the platform’s move toward monetization amid rising operational costs. They discuss the controversy over charging for human‑verified knowledge,...
Lovable’s 8M-User Surge: The Future of AI App Building
In this episode, hosts Jaeden and Jamie dissect Lovable’s rapid growth to nearly 8 million users, highlighting how its AI‑driven app‑building platform is reshaping software creation. They discuss the platform’s core strengths, such as rapid prototyping and user‑friendly design, alongside security...
Sen. Warren Presses Trump Admin on Potential AI Bailout Plans
Sen. Elizabeth Warren wrote to White House AI adviser David Sacks and OSTP director Michael Kratsios demanding answers on whether the Trump administration intends to "backstop" major AI firms such as OpenAI with taxpayer money, citing OpenAI’s request to expand...
Google Gemini 3 Sparks Unprecedented AI Community Buzz
How big a deal is Google Gemini 3? Haven’t seen a release get so much attention across the community of 35,000 in AI as this one has. Use X Pro and lists. Way better to watch the whole community: https://t.co/9eRY65x3IQ (I have...
Gemini 3 Spreads Across Platforms Amid Cloudflare Outage
Are you trying Google Gemini 3 on Genspark? Or on Flowith? Or Cursor? Or just raw dogging it on Google AI studio? Just a few that have already added Gemini, my whole feed is Gemini this morning. Oh, and welcome...
Google’s Sundar Pichai Warns of “Irrationality” In Trillion-Dollar AI Investment Boom
Alphabet CEO Sundar Pichai warned that the AI investment frenzy, now worth over $1.4 trillion, shows signs of irrational exuberance reminiscent of the late‑1990s Internet bubble. While acknowledging that no company, including Google, is immune to a potential correction, he highlighted...
Gemini 3 Tops Benchmarks and Excels in Everyday Tasks
We’ve been intensely cooking Gemini 3 for a while now, and we’re so excited and proud to share the results with you all. Of course it tops the leaderboards, including @arena, HLE, GPQA etc, but beyond the benchmarks it’s been...
Training Vs. Inference: Break‑Even at 25 Million Queries
What should we focus on, (more) LLM training or inference scaling? (A question I got asked multiple times now, so here are some thoughts.) Training is usually very, very expensive, but it is a one-time cost. Inference-scaling is comparatively cheap, but...
Google's Gemini 3 Is Here: A Special Early Look
The episode breaks down Google’s rollout of Gemini 3, the latest large‑language model, highlighting its new multimodal capabilities, faster reasoning, and tighter integration with Google services. Executives Demis Hassabis and Josh Woodward explain the technical breakthroughs, safety measures, and the product’s...
Gemini 3 Unveiled
Google just released Gemini 3, its most intelligent AI model yet. I caught up with Demis Hassabis, CEO of Google DeepMind, to ask about: -Gemini 3 and Google’s AI strategy -Google's new Antigravity tool -Medical-grade AI Here's what he said:
Gemini 3 Launches as World’s Most Intelligent Multimodal Model
Today we’re taking a big step on the path toward AGI and releasing Gemini 3— our most intelligent model yet. With Gemini 3, you can bring any idea to life. It is state-of-the-art in reasoning, the best model in the...
Google Is Launching Gemini 3, Its ‘Most Intelligent’ AI Model Yet
Google launched Gemini 3 Pro today, its most advanced multimodal AI model that can process text, images and audio simultaneously. The model is being rolled out to all users in the Gemini app and to paid subscribers within Google Search’s...
ChatGPT‑5.1 and Gemini‑3 Ace Colorblind Test
ChatGPT-5.1-Thinking and Gemini-3 both pass the colorblind test. Claude and I both failed. https://t.co/VSdB5E3vz9
Replit Design Now Auto‑creates Videos for You
Replit Design will generate videos for you!
Your Favorite Microsoft 365 Apps Are Getting a Load More Copilot AI Agents - Word, Excel, Outlook, PowerPoint All Get...
Microsoft unveiled a suite of new Copilot AI agents for Word, Excel, PowerPoint, and Outlook at Ignite 2025, enabling users to generate content, charts, summaries, and presentations through natural‑language prompts. The agents engage in follow‑up questions to refine output and...
Transformer Sets New Translation SOTA with Comman A
The Transformer was built for translation back in 2017. Very proud of the team for setting a new SOTA with Comman A Translate!
Consistent Naming Crucial Even With Autocomplete Assistance
It doesn’t matter as much in the age of autocomplete as it used to, but the tiny mental friction of naming inconsistencies like “start_dim” vs “keepdim” still sticks out for me. Carefully pick names as if millions of developers may...
Microsoft’s Office Apps Are Getting Even More Free AI Features
Microsoft announced that a suite of free AI capabilities will be added to Outlook, Word, Excel and PowerPoint for all Microsoft 365 subscribers, with a preview slated for March 2026. The enhancements include an expanded Copilot Chat that can scan an...
Replit Launches Design: AI-Powered Beautiful UI Experience
Announcing Design, a new Replit experience focused on beautiful UIs. The first non-slop AI design experience, powered by Gemini 3.0 https://t.co/wssUgV9Zuk
AI Success Depends on the Right External Partners
AI project success often hinges on enlisting the right outside expertise. The best partners know your business, value outcomes, and deliver seamless collaboration. https://t.co/3eYBy5MKN6
Writer's AI Agents Can Actually Do Your Work—Not Just Chat About It
Writer, a San Francisco AI startup, launched Writer Agent, a unified chat‑plus‑automation platform that lets non‑technical employees create and run multi‑step business workflows across apps like Salesforce, Slack and Google Workspace without coding. Users can issue plain‑language requests, have the...
Own Hardware Pipelines Enable Aggressive, Low‑cost Inference Scaling
Wanted to say this is because the competition focused on cheaper models. But that eve doesn't seem true. What seems more plausible is that teams with strong training pipelines and their own hardware can push much more aggressive inference-time scaling at...
Collective Learning Models Aren’t Simple Superintelligence
Assuming every frontier model learns from every other, how is it different from straight-shot superintelligence?
Microsoft Remakes Windows for an Era of Autonomous AI Agents
Microsoft announced at its Ignite conference that Windows 11 will become an "agentic OS" by embedding native infrastructure for autonomous AI agents, including new platform primitives such as Agent Connectors, an on‑device registry for the Model Context Protocol, and a...
Seeking Deeper Analysis of Intriguing Solve Logs
Fascinating results. Can we get more eyes and analysis on the solve logs to try to understand: https://t.co/ovW1IG4vGL ?
Voice Takes Over: Win a Monokey Keyboard‑Free Device
https://t.co/GRTnUAe9ZD
Microsoft's Fabric IQ Teaches AI Agents to Understand Business Operations, Not Just Data Patterns
Microsoft unveiled Fabric IQ at its Ignite conference, a semantic intelligence layer that embeds ontologies into the Fabric data platform to map datasets to real‑world entities, relationships and operational rules. The technology upgrades Power BI’s 20 million semantic models into enterprise‑wide ontologies,...
Gemini 3 Pro Leaps Ahead in Multimodal Fidelity
Best outputs from Gemini 2.5 Pro, vs 3 Pro, this example nicely illustrates the fidelity jump with 3 Pro, through strong multimodal understanding and 3D reasoning https://t.co/0aM8oSgmFA
Gemini 3 Generates Voxel Art From a Single Image
Amazing test of Gemini 3’s multimodal reasoning capabilities: try generating a threejs voxel art scene using only an image as input Prompt: I have provided an image. Code a beautiful voxel art scene inspired by this image. Write threejs code as...
Google Unveils Gemini 3 Claiming the Lead in Math, Science, Multimodal and Agentic AI Benchmarks
Google unveiled Gemini 3, its latest proprietary AI model family—including flagship Gemini 3 Pro, Deep Think reasoning mode, Gemini Agent and the Antigravity developer environment—across Search, the Gemini app, AI Studio, Vertex AI and other developer tools. Independent benchmarks (Artificial Analysis, LMArena) rank Gemini 3 Pro as the global...
Google's Gemini 3 Launch Impresses the Community
Congrats to Google on Gemini 3! Looks like a great model.
Recreate Your Site’s Early‑2000s Look Using Gemini 3
highly recommend generating an early 2000's version of your website with gemini 3 https://t.co/J5fWuq6akB
How AI Data Platforms Are Shaping the Future of Enterprise Storage - Ep. 281
In this episode, NVIDIA’s Jacob Liberman explains how AI Data Platforms integrate GPUs directly with storage to accelerate enterprise workloads, turning data into a compute resource. He outlines the architecture of the GPU‑accelerated storage solution, its performance benefits, and how...
Using Gemini 3 with Codex Extension in VSCode
How can I use gemini 3 with codex extension in vscode? 😂
Gemini 3 Sets New Standard in Multimodal AI
Gemini 3 is here ... and it's INSANELY good. → best model in world for multimodal understanding → beats Claude sonnet 4.5, GPT 5.1 on *almost every* benchmark → wildly good at agentic tool use and vibe-coding → SOTA at complex reasoning → great at zero-shot...
Gemini 3 Pro Delivers More Intelligence per Token
The most interesting thing testing Gemini 3 Pro has been how *efficient* it is from tokens to tool calls The intelligence per token of models is increasingly rapidly even as prices fall, its quite something
Gemini 3 Hits 31.1% on ARC‑AGI‑2 Benchmark
Gemini 3 scores 31.1% on ARC-AGI-2. Impressive progress.
Waymo Expands to Five New Cities, Scaling Fivefold Annually
Waymo is adding 5 new cities: Miami, Dallas, Houston, Orlando. Waymo has been growing about 5x per year since it started scaling its service in 2023.