AI Model Announcements
- Google released Gemini Diffusion, a new model that uses diffusion for language modeling, achieving 10-15x faster generation than autoregressive models @demishassabis
- Google unveiled Veo 3, their latest video generation model with native audio generation capabilities, improved physics, and better prompt understanding @sundarpichai
- Google introduced Gemma 3n, a multimodal model that runs on as little as 2GB of RAM, supporting audio, image, video, and text across 140 languages @GoogleAI
- Mistral AI released Devstral Small 24B, an Apache 2.0 licensed coding agent model that reached #1 on SWE-bench for open-source models @MistralAI
- NVIDIA released Llama-3.1-Nemotron-Nano-4B-v1.1, a compressed version of Llama3.1-8B that outperforms DeepSeek-R1-Distill-Llama-8B while being twice as small @huggingface
AI Research
- Microsoft published research in Nature about Aurora, an AI foundation model that goes beyond weather forecasting to more accurately predict environmental events like hurricanes and ocean waves @MSFTResearch
- New research shows embedding models from different sources are so similar that they can be mapped between them based on structure alone, without any paired data @AndrewCurran_
- Microsoft's Discovery uses specialized AI agents that reason over scientific knowledge, generate hypotheses, and simulate results in a continuous loop, discovering a novel coolant in 200 hours @Microsoft
- Stanford researchers developed a generative AI agent architecture that can simulate the attitudes of 1,000+ real people for testing ideas in social science @StanfordHAI
AI Applications
- Google launched Flow, an AI filmmaking tool designed for their advanced models that allows users to extend videos, add sound effects, and maintain character consistency @GoogleDeepMind
- Google acquired Stitch (formerly Galileo AI), which allows users to design UIs iteratively from prompts and download them into Figma @deedydas
- Google introduced Jules, an app that makes changes to GitHub repositories with simple English prompts without requiring local cloning @deedydas
- Google demonstrated virtual try-on technology that uses AI to let users try on clothes using just a full body picture @deedydas
- Google showcased real-time translation with multimodal AI for Google Meet, eliminating language barriers in video calls @deedydas
- Framer announced new AI tools including AI Wireframing to quickly generate layouts and Workshop AI to code interactive components @benblumenrose
- OpenAI and Jony Ive announced io, a new company focused on creating the next generation of AI products and interfaces @OpenAI
- xAI added Live Search to their API, allowing Grok to search through realtime data from X, the internet, and trending news @xai
- OpenAI launched MCP (Multi-Channel Platforms) support for their Responses API, with Zapier as an official launch partner @gdb
- Google is bringing AI Mode to Search widely, providing GPT/Perplexity-like answers directly in search results @deedydas
- Mistral AI and Google DeepMind announced agent collaboration capabilities, allowing their respective agents to work together @AndrewCurran_
AI Industry Analysis
- Survey data shows a significant surge in AI use at work, increasing from around 30% of US workers in December to over 40% in March/April 2025, with expansions in both Gemini and ChatGPT usage @emollick
- Meta launched the Llama Startup Program to support early-stage startups building generative AI applications with Llama, offering cloud reimbursements and technical support @AIatMeta
- LM Arena raised $100M in seed funding led by a16z and UC Investments to support their platform for understanding and improving AI model performance @pmarca
- Analysis of AI power consumption shows that while individual usage is small, aggregate impact is significant - testing showed Llama 3.1 405B averaged 3,353 joules per prompt, equivalent to 2 minutes 50 seconds of human brain activity @emollick
- Gemini has over 400M monthly active users and processes 480T tokens a month according to Google @deedydas
- The speed of AI adoption in business will depend more on innovation in business models, risk management, and governance than on the speed of improvement in AI capabilities @random_walker
AI Ethics & Society
- ChatGPT's new memory-from-your-chats feature represents a significant change to how the model works, raising concerns about user control over model input @simonw
- Research on AI in education shows a split impact: when used as a tutor with instructor guidance, AI has significant positive effects, but when used alone for homework help, it can act as a shortcut that hurts learning @emollick
AI Model Announcements
- Google announces Gemini 2.5 Pro with "Deep Think" mode that uses parallel thinking techniques to consider multiple hypotheses before responding @demishassabis @OfficialLoganK
- Google introduces Gemini 2.5 Flash, a faster model that will be generally available in early June, pushing the pareto frontier of performance @sundarpichai @OfficialLoganK
- Veo 3, Google's state-of-the-art video generation model with native audio generation capabilities, is now available for Google AI Ultra subscribers in the US @GoogleDeepMind @JeffDean
- Imagen 4, Google's latest image generation model, is now live with improved details, more nuanced color, and better text outputs @GeminiApp
- Google announces Gemma 3n, a new model optimized for mobile on-device usage with multimodality and fast inference @demishassabis
- Google introduces Lyria 2 for YouTube shorts and on Vertex @AndrewCurran_
AI Research
- New paper on ARC-AGI-2 released, covering design principles, analysis of human performance, and current model performance @fchollet
- Google introduces Gemini Diffusion, a research model that's significantly faster than previous models while matching coding performance by correcting errors during thinking @GoogleAI
- Google's Gemini 2.5 Pro with Deep Think achieves 49.4% on USAMO (USA Mathematical Olympiad), a significant advancement in mathematical reasoning @quocleix
- Meta introduces Adjoint Sampling, a new learning algorithm that trains generative models based on scalar rewards, with theoretical foundations developed by FAIR @AIatMeta
- NVIDIA releases Cosmos-Reason1-7B, described as the first reasoning model for robotics, based on Qwen 2.5-VL-7B @huggingface
- New research paper suggests potential issues with deep learning representations and proposes solutions for improvement @jeffclune
- Meta releases OMol25, a dataset of 100M+ molecular conformers spanning 83 elements for training machine learning models with DFT-level accuracy @huggingface
AI Applications
- Google launches Flow, a filmmaking tool that combines Veo, Imagen, and Gemini models to help create cinematic clips and narratives @GoogleDeepMind
- Google introduces Jules, a coding agent that lets users make changes to GitHub repos with English prompts in a VM using Gemini 2.5 Pro @deedydas @eugeneyan
- Google announces Gemini in Chrome, an AI browsing assistant that provides summaries and answers without switching tabs @GeminiApp
- Google introduces Agent Mode in Gemini App to help users complete tasks across the web @sundarpichai
- Google launches AI Mode in Search, using "query fan out" technique to break queries into subtopics and generate comprehensive responses @GoogleAI
- Google introduces SynthID Detector, a portal to identify if digital content was generated by Google's AI tools, already used 10 billion times @GoogleDeepMind
- Google announces Google Beam, a 3D video communications platform that transforms 2D video streams into realistic 3D experiences @GoogleAI
- Microsoft announces Grok 3 API support coming to Azure, though with limited transparency regarding security and model details @emollick
- Stability AI upgrades Stable Video Diffusion 4D to Stable Video 4D 2.0, improving quality of 4D outputs generated from a single object-centric video @StabilityAI
- Google's NotebookLM app is now available on the App Store with Video Overviews feature @demishassabis @OfficialLoganK
- SAP partners with Cohere to embed enterprise-ready agentic AI into SAP Business Suite @cohere
AI Industry Analysis
- Google reports processing 480 trillion tokens monthly across products and APIs, a 50x increase year-over-year @sundarpichai @OfficialLoganK
- Google's Gemini app has over 400 million monthly active users, with 7 million developers building with the Gemini API (4x growth) @OfficialLoganK
- ChatGPT daily active users have increased more than 4x over the last year, with messages per day growing even more significantly @sama
- Google AI Overviews are now used by 1.5 billion people monthly across 200+ countries and territories @sundarpichai
- Meta's Llama models will be direct first-party offerings in Azure AI Foundry, hosted and sold by Microsoft @AIatMeta
- AI coding tools companies predominantly focus on React and TypeScript demos, while Microsoft showcases Java and .NET case studies as a strategic differentiation @GergelyOrosz
- One side-effect of AI coding is that "everyone is an IC now" (individual contributor) @alexgraveley
- The narrative that AI use will collapse due to data limits, costs, environmental factors, or regulation is not useful, as over a billion people use this technology with self-reported high utility @emollick
AI Ethics & Society
- AI Now Institute launching research on AI's growing energy demands and the industry's turn to nuclear energy, focusing on infrastructure, safety, and oversight risks @AINowInstitute
- Berkeley AI Research paper explores how frontier AI is reshaping cybersecurity, predicting attackers may gain more immediate advantages than defenders in the short term @berkeley_ai
- World Bank randomized controlled study finds using GPT-4 as a tutor with teacher guidance in a six-week after-school program in Nigeria had "more than twice the effect of some of the most effective interventions in education" at very low costs @emollick
- State of AI in Design Report released, surveying hundreds of designers and leaders from companies like Notion, Stripe, Ramp, Anthropic, and Perplexity on AI adoption in design @benblumenrose
AI Model Announcements
- Microsoft announced that xAI's Grok is coming to Azure, confirming earlier rumors @satyanadella @xai
- Meta released UMA (Universal Model for Atoms), a machine learning interatomic potential trained on over 30 billion atoms for molecular chemistry @AIatMeta
- Meta also released Open Molecules 2025 (OMol25), a new Density Functional Theory dataset for molecular chemistry @AIatMeta
AI Research
- DeepSeek released a comprehensive paper on large model training covering software, hardware, and mixed approaches - described as "the single best end-to-end paper on large model training" @deedydas
- Stanford NLP Group launched Marin, an open lab focused on truly open-source AI with open development where the entire research process is public and anyone can contribute @stanfordnlp
- Researchers released LEXam, a legal reasoning benchmark with 4,586 exam questions from Swiss, EU & international law, evaluating 20+ state-of-the-art LLMs @huggingface
- MIT researchers developed an AI model that can predict the location of virtually any protein within a human cell by training it with joint understanding of protein and cell behavior @MIT
AI Applications
- Microsoft introduced Magentic-UI, an open-source research prototype of a human-centered AI agent designed to work with people to complete complex web-based tasks in real time @MSFTResearch
- Microsoft announced GitHub Copilot coding agent, evolving from pair programmer to peer programmer that can autonomously complete tasks like bug fixes and new features @satyanadella
- Microsoft revealed Copilot Tuning which allows companies to train Copilot on their unique tone and language @satyanadella @Microsoft
- Microsoft launched Healthcare Agent Orchestrator to bring AI capabilities into healthcare systems' existing enterprise software @MSFTResearch
- Microsoft unveiled Microsoft Discovery, a platform that uses AI agents to generate ideas, simulate results, and learn for scientific research @satyanadella
- Google released the NotebookLM mobile app for Android with iOS coming soon @sundarpichai @TechCrunch
- GenSpark AI Sheets allows users to talk to their spreadsheets, automatically analyzing data and building reports and visualizations @fchollet
- Overlap is an autonomous agent that creates viral clips from any video and posts them to social media @ycombinator
- Hugging Face integrated MLX LM directly within their platform, allowing Mac owners to run 4,400+ LLMs locally on Apple Silicon @huggingface
AI Industry Analysis
- Microsoft is open sourcing Copilot for VS Code, which makes VS Code more dominant as forks cannot access the VS Code marketplace @GergelyOrosz
- Microsoft announced NLWeb, a new open project that lets users interact with any website using natural language, described as "HTML for the agentic web" @satyanadella
- Waymo is now doing more rides than Lyft in San Francisco with only 300 vehicles compared to Lyft's 45,000 drivers - each Waymo is doing more rides than 150 human drivers @paulg
- Dell is partnering with Cohere to offer Cohere North, a secure agents platform, to enterprises on-premises, which is crucial for regulated industries handling sensitive data @cohere
- Meta CTO describes AI agent "failures" as valuable demand signals that reveal real user intent and opportunities for developers @a16z
AI Ethics & Society
- AI Now Institute is launching a roadmap for local action regarding AI's impact on cities, addressing concerns about public services cuts, surveillance, and power reallocation to tech companies @AINowInstitute
- Trump signed a bill criminalizing revenge porn and explicit deepfakes @TechCrunch
AI Model Announcements
- Qwen has released Qwen 2.5 VL on Ollama, with a 6GB version available for image description tasks @simonw
AI Research
- Stanford's latest Natural Language Processing with Deep Learning course (CS224N) taught by Professor Christopher Manning is available online @stanfordnlp
AI Applications
- Perplexity on WhatsApp has been updated to be snappier, faster and more chatty, with news and alerts features coming soon @AravSrinivas
- Codex successfully upgraded a Jekyll-GitHub pages site to the latest Ruby and gems @eugeneyan
- o3 can generate creative screenshots from descriptions, including ones mimicking 1950s safety videos @emollick
- Cursor now allows users to quickly edit entire files using AI assistance @cursor_ai
- Replit introduces time travel feature that allows developers to go back in time for both code and database states @amasad
- Modal appears to be developing notebook functionality, expanding their AI infrastructure offerings @eugeneyan @HamelHusain
AI Industry Analysis
- Anthropic has received a $2.5 billion revolving credit line, with revenue hitting $2 billion in Q1 2025 (double from previous quarter) and customers spending over $100,000 annually increasing eightfold year-over-year @AndrewCurran_
- ChatGPT is rolling out memory improvements with a new toggle for persistent memory of projects and conversations, currently available to Pro and Plus users @AndrewCurran_
- K-Scale Labs is building open-source humanoid robot hardware and software for developers, with their K-Bot priced at $8,999 and deliveries beginning July 2025 @garrytan
- YC hosted a large MCP (Multi-agent Conversational Protocol) hackathon with 400+ attendees and 80 submissions, showcasing applications from cancer research to email management @ycombinator
- Using AI as a second opinion in your area of expertise is becoming a low-risk way to improve outcomes across most fields @emollick
AI Ethics & Society
- Grok AI reportedly expressed skepticism about Holocaust death toll, which was later attributed to a "programming error" @TechCrunch
- MIT study reveals that people's views on data privacy aren't fixed but shift depending on how, where, and why their data is used @MIT
- Good intellectual communities need both "naive young fast updaters" who introduce many ideas (including some low-quality ones) and "wise old slow updaters" who act as filters and sanity checks @AmandaAskell
AI Model Announcements
- Alibaba releases quantized versions of Qwen2.5-Omni-7B models on Hugging Face and ModelScope @Alibaba_Qwen
- Alibaba introduces WorldPM (World Preference Model), showing that human preference modeling follows scaling laws with experiments on Qwen2.5 models from 1.5B to 72B parameters @Alibaba_Qwen
- NVIDIA releases Direct Discriminative Optimization models on Hugging Face, improving visual generative models like EDM & VAR with record FID scores on CIFAR-10/ImageNet @huggingface
- Windsurf introduces SWE-1, a specialized coding model that competes with frontier models, along with SWE-1-lite and SWE-1-mini variants @windsurf_ai
AI Research
- Alibaba's research reveals human preference modeling follows scaling laws, suggesting diverse preferences might share a unified representation @Alibaba_Qwen
- Windsurf's SWE-1 model achieves near-parity with frontier models in helpfulness, accuracy, and edit quality for software engineering tasks @windsurf_ai
- MIT has disavowed a doctoral student paper on AI's productivity benefits, removing evidence that LLMs act as multipliers for high performers @emollick @TechCrunch
AI Applications
- Codex CLI continues to improve, with Greg Brockman suggesting future convergence of "local" and "remote" coding agents @gdb
- Y Combinator introduces Workflow Use, a deterministic, self-healing browser automation tool that's 10x faster and ~90% cheaper than pure LLM agents @ycombinator
- RunRL improves language models with reinforcement learning, helping customers increase accuracy from 60% with Claude to 95% @ycombinator
- Replit enhances their agent experience with improved checkpoints management, including naming, rollbacks, and preview app capabilities @amasad
- Y Combinator startup Firecrawl is offering $1M to hire three AI agents as employees @TechCrunch
- Cua introduces a Trajectory Viewer that shows exactly what Computer-Use AI agents see and do @garrytan
AI Industry Analysis
- OpenAI's planned data center in Abu Dhabi would be larger than Monaco @TechCrunch
- Greg Brockman and Paul Graham both declare "2025 is the year of agents" @gdb @paulg @ycombinator
- Garry Tan suggests OpenAI isn't trying to outcompete AI startups, noting "on the API side, they very much hope that a lot of them do really, really well" @paulg @ycombinator
- Over 300 companies including Adobe, Amazon, Google, Meta, Microsoft, OpenAI, and NVIDIA are taking Hamel Husain's AI evals course @HamelHusain
- Hugging Face announces official partnership with Kaggle, enabling direct running of HF models in Kaggle Notebooks @huggingface
AI Ethics & Society
- Ethan Mollick raises concerns about AI-powered always-on devices creating new privacy issues as recordings become more valuable when AI can process audio into useful data @emollick
- Aidan McLaughlin discusses alignment concerns about AI systems potentially being optimized for addiction rather than human fulfillment @aidan_mclau
AI Model Announcements
- OpenAI introduces Codex, a software engineering agent powered by codex-1 (a version of o3 optimized for software engineering) that can independently navigate codebases, implement changes, and propose pull requests @OpenAI @sama @gdb
- Cursor announces a new Tab model that can jump across files, rolling out to users in their latest update @cursor_ai
- Windsurf introduces SWE-1, their first frontier model for complex software engineering tasks, claiming performance comparable to Claude-3.5 Sonnet, GPT-4.1, and Gemini-2.5 Pro on challenging benchmarks @windsurf_ai
- Microsoft's 4o Image Generation is now live in Copilot, offering capabilities like rendering accurate text, editing creations, and making photorealistic images @Copilot
AI Research
- xAI publishes their Grok system prompts openly on GitHub following an incident with "unauthorized modifications" to the prompt that directed Grok to provide specific responses on political topics @xai
- Codex-1 achieves state-of-the-art performance on SWEbench, a benchmark for software engineering tasks @sama
- New meta-analysis of 51 studies shows AI has a large positive impact on students' learning performance (0.867 SD) and moderate positive impact on learning perception (0.456 SD) and higher-order thinking (0.457 SD) @mustafasuleyman
- Researchers from Berkeley AI Research introduce Real2Render2Real, a method to scale robot datasets without teleoperating, dynamic simulation, or robot hardware - using just smartphone scans and human hand demo videos @berkeley_ai
AI Applications
- Codex enables developers to run multiple software engineering tasks in parallel, helping with bug fixes, feature implementation, and code navigation @OpenAI @sama
- Google AI Studio rolls out a new built-in usage dashboard allowing users to easily check request and token volumes and spending @OfficialLoganK
- Google AI Studio introduces a new generative media experience bringing together Veo 2, Gemini 2.0 native image generation/editing, and Imagen 3 @OfficialLoganK
- Google offers Gemini Advanced free to U.S. college students through finals 2026 @GeminiApp
- Hugging Face announces integration with Kaggle, allowing users to use any model from Hugging Face directly in Kaggle without downloading and uploading models as datasets @huggingface
- Hotel bookings natively on Perplexity are quietly growing, with potential to disrupt the ad industry @AravSrinivas
- PDF downloads for deep research reports now fully rolled out to Free, Edu, and Enterprise users in ChatGPT @OpenAI
AI Industry Analysis
- Meta reportedly delaying its biggest AI model launch, Llama 4 Behemoth, due to poor internal performance, AI leadership reorganization, and researcher departures @deedydas
- Sam Altman envisions the future of work being like Starcraft or Age of Empires, with users directing "200 microagents" to fix problems, gather information, and design new systems @sama
- Google One recently crossed 150 million subscribers, a 50% increase since February 2024, partly driven by AI features @demishassabis
- OpenAI and Anthropic both establishing offices in Europe, with OpenAI setting up in Zurich, likely to hire from Google's large presence there @GergelyOrosz
AI Ethics & Society
- Jeff Clune advocates that every AI company should be required by law to publish their system prompts openly, similar to xAI's recent move following their incident @jeffclune
- Arvind Narayanan publishes a critique on the implications of AI that's "grounded in the state of AI today" rather than focusing on hypothetical AGI scenarios @emollick
- Ethan Mollick notes that most key experiments showing impressive AI abilities in academic research were done on GPT-4, a model now considered obsolete, suggesting current capabilities are likely higher @emollick
- François Chollet emphasizes that there's "a lot more signal in system failures than in regular operations" when analyzing AI systems @fchollet
AI Model Announcements
- OpenAI is preparing to share another "low-key research preview" soon, with better naming than ChatGPT @sama
- 4o Image Generation is now live in Microsoft Copilot with sharper visuals, more consistent text, and styles ranging from photorealistic to playful @Copilot
- Salesforce released BLIP3-o on Hugging Face, a family of fully open unified multimodal models @huggingface
- Falcon released Falcon-Edge – a series of powerful, universal and fine-tunable Bitnet models, along with a Python fine-tuning toolkit library called 'onebitllms' @huggingface
AI Research
- Google's AlphaEvolve made mathematical discoveries no human has before, including solving optimal packing problems and reducing 4x4 matrix multiplication from 49 operations to 48 (first advance in 56 years) @deedydas
- Given 50 open math problems, AlphaEvolve rediscovered the leading approach 75% of the time and improved on it 20% of the time @emollick
- Meta and Fondation Rothschild released "Emergence of Language in the Developing Brain" - the first systematic investigation of how neural representations of language evolve as the brain develops @ylecun
- Meta introduced Open Molecules 25, a foundational quantum chemistry dataset including over 100M DFT calculations across 83M unique molecules, built with 6B core hours of compute @ylecun
AI Applications
- OpenAI launched the "OpenAI to Z Challenge" using o3/o4 mini and GPT-4.1 models to discover previously unknown archaeological sites @gdb @kaggle
- Replit introduced Safe Vibe Coding to address security vulnerabilities created by AI coding assistants like Cursor and Windsurf that expose API keys by default @amasad @garrytan
- Unsloth now allows fine-tuning of TTS models like Sesame-CSM and OpenAI's Whisper locally, making training 1.5x faster with 50% less VRAM @ycombinator
- Google's Gemini App now offers Audio Overviews in 45 languages, turning documents, slides, and research reports into podcast-style conversations @GeminiApp
- Cursor released its "biggest release ever" with version 0.50 @cursor_ai
- Hedra Labs is creating AI character animation for video that avoids the uncanny valley, with nearly 3M users generating over 10M videos @a16z
AI Industry Analysis
- Microsoft's recent layoffs primarily affected programmers in its home state as AI now writes up to 30% of its code @TechCrunch
- Anthropic's lawyer was forced to apologize after Claude hallucinated a legal citation in court @TechCrunch
- Coinbase disclosed a breach where hackers stole customers' personal information including IDs by "paying multiple contractors or employees working in support roles" @TechCrunch
- AI's ability to make tasks not just cheaper but faster is underrated in its importance for creating business value, especially in software development @AndrewYNg
- 38% of employees surveyed admitted to sharing sensitive information with AI tools at work, highlighting the need for enterprise-grade secure AI solutions @cohere
- Hugging Face is pivoting the Transformers library to become the central model definition source across the AI ecosystem, partnering with vLLM, LlamaCPP, SGLang, and many others @huggingface
AI Ethics & Society
- Google is offering U.S. college students free access to Gemini Advanced through spring 2026 to help with exam prep and homework @GeminiApp
- Ambient intelligence research at Stanford offers a solution for catching early signals of cognitive decline @StanfordHAI
- Google's Project Euphonia released open-source tools to empower developers to build personalized audio tools and fine-tune models for diverse speech patterns @GoogleAI
- MIT CSAIL asked "What's a common misconception about machine learning that you wish more people understood?" to promote public understanding of AI @MIT_CSAIL
AI Model Announcements
- Google DeepMind introduces AlphaEvolve, a Gemini-powered coding agent for algorithm discovery that can design faster matrix multiplication algorithms, find new solutions to open math problems, and make data centers more efficient @GoogleDeepMind
- OpenAI makes GPT-4.1 and GPT-4.1 mini available directly in ChatGPT, with GPT-4.1 mini replacing GPT-4o mini @OpenAI
- Stability AI releases Stable Audio Open Small, a 341M-parameter text-to-audio model optimized to run entirely on Arm CPUs, enabling on-device audio generation on 99% of smartphones @StabilityAI
- Hugging Face releases Wan2.1, a model excelling in Text-to-Video, Image-to-Video, Video Editing, Text-to-Image, and Video-to-Audio @huggingface
- StepFun AI releases Step1X-3D, an open 3D generation framework with 4.8B parameters (1.3B geometry + 3.5B texture) under Apache 2.0 license @huggingface
- Meta FAIR releases Open Molecules 2025 (OMol25) dataset and Universal Model for Atoms (UMA) for molecular discovery and modeling atom interactions @AIatMeta
AI Research
- AlphaEvolve applied to over 50 open problems in mathematical analysis, geometry, combinatorics and number theory, rediscovered state-of-the-art solutions in 75% of cases and improved upon previous best solutions in 20% of cases @GoogleDeepMind
- AlphaEvolve found a simple code rewrite that removed unnecessary bits in TPU design, validated by TPU designers for correctness, representing Gemini's first direct contribution to TPU arithmetic circuits @AndrewCurran_
- AlphaEvolve sped up the FlashAttention kernel by 32% and found improvements in pre- and postprocessing of kernel inputs and outputs, resulting in a 15% speed up @AndrewCurran_
- Meta FAIR and the Rothschild Foundation Hospital partnered on a large-scale study revealing striking parallels between language development in humans and LLMs @AIatMeta
- Meta releases Adjoint Sampling, a scalable algorithm for training generative models based on scalar rewards @AIatMeta
AI Applications
- Anthropic launches a bug bounty initiative to stress-test an updated version of their anti-jailbreaking system before public deployment, in partnership with HackerOne @AnthropicAI
- Gemini Advanced now connects with GitHub, allowing users to generate/modify functions, explain complex code, ask questions about codebases, and debug by importing code from public or private repositories @GeminiApp
- Perplexity announces integration with PayPal and Venmo for commerce features including shopping, travel, voice assistants, and their upcoming agentic browser called Comet @perplexity_ai
- Google brings Gemini to Wear OS, Android Auto, Google TV, and Android XR, while making Gemini Live's camera and screen sharing features free for all Android users @demishassabis
- Y Combinator launches Storyboards, a tool that turns scripts into full storyboards with shot-level control and character/scene consistency @ycombinator
- Amjad Masad announces Percival, an AI agent that can evaluate and fix other AI agents, outperforming SOTA LLMs by 2.9x on the TRAIL dataset @amasad
AI Industry Analysis
- BigTech jobs (Google, Microsoft, Apple, Tesla, Meta, Nvidia, Palantir) show zero growth in the last 3 years, contributing to difficulty for CS majors to find jobs, with companies potentially leveraging AI to grow without hiring @deedydas
- Kaggle partners with Hugging Face to enable direct use of Hugging Face models in Kaggle Notebooks, along with discovering linked public code examples @kaggle
- Databricks acquires serverless Postgres startup Neon for $1B, representing a rare unicorn exit in the current tech market @deedydas
- Andrew Ng announces new course on Model Context Protocol (MCP) in partnership with Anthropic, teaching how to build AI apps that access tools, data, and prompts using the standardized protocol @AndrewYNg
AI Ethics & Society
- OpenAI introduces the Safety Evaluations Hub, a resource to explore safety results for their models that will be updated periodically as part of efforts to communicate proactively about safety @OpenAI
- Anthropic notes that some future models may require the advanced "AI Safety Level 3" protections outlined in their Responsible Scaling Policy @AnthropicAI
- Paul Graham suggests that AGI would mean the end of prompt engineering, as moderately intelligent humans can figure out what you want without elaborate prompts, and we can use the care needed to construct prompts as an index of how close we're getting to AGI @paulg
AI Model Announcements
- @Alibaba_Qwen released the Qwen3 Technical Report, documenting their latest model architecture and capabilities
AI Research
- @berkeley_ai released research on learning generalized visual navigation policy from scalable but low-quality and action-free passive data sources
- @AIatMeta published Part 4 of Physics of Language Models, introducing Canon layers that add "horizontal residual links" across tokens to significantly improve reasoning and generalization in Transformers, Mamba, GLA, and beyond
- @AIatMeta introduced CATransformers, a carbon-driven neural architecture and system hardware co-design framework that achieves 9.1% reduction in total lifecycle carbon emissions while maintaining or increasing accuracy
- @ch402 discussed the rationale behind titling their paper "On the Biology of a Large Language Model," explaining how the scientific aesthetic of biology is relevant to deep learning and interpretability research
- @GoogleAI shared research on using trust graphs to model relationships and apply Differential Privacy to reflect users' asymmetric privacy preferences in data-sharing scenarios
- @MIT_CSAIL introduced CausVid, a new AI model that crafts smooth, high-quality videos in seconds by combining the photorealism of diffusion models with the speed of autoregressive approaches
- @huggingface announced Ultra-FineWeb, a cleaner 1.1T-token foundation for better LLMs with 1T English + 120B Chinese tokens, filtered for quality, showing +3.6 points improvement on MMLU and +3.7 on CMMLU versus FineWeb
- @huggingface released Step1X-3D, a fully open-source 3D generation framework for high-fidelity and controllable generation of textured 3D assets
- @emollick noted that in September 2024, physicians working with AI performed better on the Healthbench doctor benchmark than either AI or physicians alone, but with o3 and GPT-4.1, AI answers are no longer improved by physicians
- @natolambert mentioned that the Tulu 3 paper coined the term RLVR (Reinforcement Learning from Value Ranking)
AI Applications
- @GeminiApp launched Veo 2 for Gemini Advanced users, allowing users to go from idea to video in minutes with simple text prompts
- @GeminiApp released an iPad app, addressing a previous limitation in platform availability
- @Alibaba_Qwen made Deep Research on Qwen Chat available for everyone after a few weeks of phased testing
- @gdb shared that Deep Research can now connect to organizations' Sharepoint, expanding its enterprise data access capabilities
- @simonw noted that Gemini, OpenAI, Perplexity, and Qwen all have features named "Deep Research" while Grok bucked the trend by calling theirs "DeepSearch"
- @huggingface announced up to 8x faster Whisper transcription on a single L4 GPU, powered by vllm_project
- @_catwu announced new Claude Code features including multipaste for large chunks of text or images, real-time steering to adjust approach during work, and OpenTelemetry support for tracking metrics
- @ycombinator launched OpenMemory MCP, a private memory for MCP-compatible clients that provides a persistent, portable memory layer for AI tools running 100% locally
- @windsurf_ai added the ability to edit Cascade's terminal suggestions before running them
- @TechCrunch reported that TikTok launched TikTok AI Alive, a new image-to-video tool
AI Industry Analysis
- @NVIDIAAI announced plans to build AI factories with HUMAIN (an AI subsidiary of Saudi Arabia's Public Investment Fund) that will transform Saudi Arabia into a global AI leader, deploying up to 500 megawatts powered by several hundred thousand NVIDIA GPUs
- @AndrewCurran_ reported that NVIDIA confirmed an agreement involving hundreds of thousands of "NVIDIA's most advanced GPUs over the next five years" for Saudi Arabia
- @AndrewCurran_ shared that Apple is working on their own Brain-Computer Interface (BCI) with a company called Synchron, developing a device called the Stentrode implanted in a vein atop the brain's motor cortex
- @_amankhan shared a graphic showing the growth of AI Product Management as a career path
- @GergelyOrosz noted that data shows AI Product Managers who know how to build AI products are in demand, contrary to claims that tech and software engineering is declining due to AI
- @garrytan observed that businesses seeking new customers will need to re-learn and optimize for AI-agent-driven search, similar to how they previously optimized for search engines
- @Deedy reported that Microsoft laid off 3% of its workforce (approximately 7,000 employees), noting that Microsoft's headcount has stayed flat for 3 years since 2022, coinciding with ChatGPT's launch
- @scottbelsky highlighted that platform shifts like AI create knowledge arbitrage opportunities, giving AI-native entrants to the workforce an advantage similar to early social media adopters
- @ylecun shared support for the House Commerce reconciliation text that includes a 10-year moratorium on state-level AI regulation, which he views as safeguarding American innovation in AI
AI Ethics & Society
- @medialab shared a Nature article discussing how chatbots and digital companions may affect individuals and society, featuring insights from Media Lab researcher @patpat_mit
- @StanfordAILab released minions secure chat, an open-source protocol for end-to-end encrypted LLM chat with less than 1% latency overhead, ensuring cloud providers cannot access messages as they decrypt only inside a secure GPU enclave
- @stanfordnlp highlighted that the House Energy and Commerce reconciliation text contains language preempting all state AI regulations for a 10-year period, representing a significant deregulatory push
- @simonw raised concerns about the usability and documentation of ChatGPT's memory feature, particularly regarding how to have conversations without having them considered as part of future memory
AI Model Announcements
- Meta releases Dynamic Byte Latent Transformer, an 8B-parameter model with alternative tokenization methods for improved language model efficiency and reliability @AIatMeta
- PrimeIntellect open-sources INTELLECT-2, a 32B parameter model trained via globally distributed reinforcement learning, beating QwQ-32B on math and code @huggingface
- BAAI releases RoboBrain, a 32B open embodied AI model enabling multi-robot collaboration with task decomposition, operable region detection, and motion trajectory prediction @huggingface
- Alibaba releases quantized versions of Qwen3 in multiple formats (GGUF, AWQ, GPTQ) for easy local deployment via Ollama, LM Studio, SGLang, and vLLM @Qwen
AI Research
- Meta introduces Collaborative Reasoner, a framework to improve collaborative reasoning in language models, paving the way for social agents that can partner with humans and other agents @AIatMeta
- OpenAI releases HealthBench, a new evaluation benchmark for AI systems in healthcare settings, developed with input from over 250 physicians from around the world @OpenAI @gdb
- Microsoft Research introduces ADeLe, a new evaluation method that explains what AI systems excel at and where they're likely to fail by breaking tasks into ability-based requirements @MSFTResearch
AI Applications
- Gemini 2.5 Pro enhances video understanding capabilities, processing up to 6 hours of video in a single request with audio-visual understanding, code integration, and temporal reasoning @HamelHusain
- ChatGPT adds PDF export functionality for research reports, complete with tables, images, linked citations, and sources @OpenAI @aidan_mclau
- Real-time webcam demo combining SmolVLM and llama.cpp server running locally on a Macbook M3 @huggingface
- Google using latest generative AI models (including Veo) to transform 2D product images into immersive 3D visualizations for Google Shopping @GoogleAI
AI Industry Analysis
- Google launches AI Futures Fund, a new program offering startups early access to Google DeepMind models, Cloud credits, and resources to build AI technology @GoogleDeepMind @JeffDean @demishassabis
- Y Combinator and partners announce AI Startup School in San Francisco featuring speakers including Sam Altman, François Chollet, Chelsea Finn, Andrej Karpathy, Fei-Fei Li, Elon Musk, Satya Nadella, Andrew Ng, and Aravind Srinivas @fchollet @garrytan
- According to Stanford's AI Index 2025, the frontier of AI development is increasingly competitive with only 0.7% separating the top-performing model from the 10th-ranked model @StanfordHAI
- Google's Gemma AI models surpass 150 million downloads @TechCrunch
AI Ethics & Society
- Mustafa Suleyman argues that larger LLMs are actually easier to control, stating "Scale doesn't hurt control - it helps" @mustafasuleyman
- Danish research shows AI adoption and impact depends on organizational encouragement, with no overall impact on wages or employment as of 2024 @emollick
- Alex Graveley suggests ChatGPT's push towards AI-assisted self-therapy and empathetic personalization could be "the greatest technological breakthrough" of his lifetime @alexgraveley
- Emollick warns against trusting reasoning chains to show what AI is thinking, noting they're designed to be useful in solving problems but aren't necessarily truthful @emollick