Google Opens Gemini Beta for Workspace
AI Product Management Certification
Today's top 25 insights for PM Builders from X and Blogs.
Google Opens Gemini Beta for Workspace
#1 𝕏
Sundar Pichai unveiled Gemini-powered Workspace upgrades—choose your sources to generate Doc drafts in seconds, build complex Sheets 9× faster, and auto-create on-brand Slide layouts with a simple prompt—and Drive now surfaces summarized answers atop search results; rolling o...
Also covered by: @Google AI, @Google AI
#2 𝕏
Logan Kilpatrick unveiled Gemini Embedding 2—a unified embedding model that brings text and multimodal capabilities into a single API, offering faster, more accurate retrieval and classification.
#3 𝕏
Hugging Face launched Storage Buckets: S3-like mutable storage optimized for high-throughput AI workloads (checkpoints, logs, agent traces), offering fast writes, overwrites and directory sync, all powered by Xet dedup to skip already-stored bytes.
#4 📝 OpenAI News
New ways to learn math and science in ChatGPT - A product announcement describing new ChatGPT features aimed at helping users learn math and science, offering improved instructional support and interactive learning experiences.
#5 𝕏
NVIDIA AI partners with @thinkymachines to deploy at least 1 gigawatt of Vera Rubin systems for frontier AI model training.
Also covered by: @Thinking Machines
#6 𝕏
Logan Kilpatrick introduced Gemini-powered Docs, Sheets, Slides, and Drive with AI Overviews, fully editable AI-generated slides, and new grounding sources, rolling out today to G1 Pro and Ultra users.
Also covered by: @Google AI, @Google AI
#7 𝕏
Kevin Weil 🇺🇸 says a single GPT-5.4 Pro run reportedly solved an @EpochAIResearch Frontier Math Ramsey-style hypergraph problem, then GPT-5.4 XHigh formalized it in Lean over a few hours.
#8 📝 OpenAI News
Improving instruction hierarchy in frontier LLMs - A research post announcing work on instruction hierarchy for frontier large language models, exploring how structuring instructions can improve model behavior and capabilities.
#9 𝕏
Sundar Pichai Our experimental AI system for mammography interpretation identified 25% more interval breast cancers, invasive cancers, and cases overall than conventional methods.
#10 𝕏
LlamaIndex 🦙 demoed audio-kb, a terminal-based audio note KB that uses LlamaParse to transcribe uploads or recordings, Google DeepMind’s Gemini Embedding 2 (3072d) for embeddings, and SurrealDB with HNSW for instant searchable voice memos.
#11 𝕏
Aravind Srinivas built a peer-to-peer file transfer web app on Perplexity Computer, using WebRTC with DTLS encryption, file chunking and Socket.IO signaling to send files directly without accounts.
#12 𝕏
Santiago argues that AI agents eliminate the need to hand-code orchestration and decision logic, making it much faster and easier for PMs to build and manage complex workflows.
#13 𝕏
Kevin Weil 🇺🇸 highlights Andrej Karpathy’s autoresearch agent autonomously tuning the nanochat model to achieve an 11% leaderboard improvement. He also uncovers the key parameter tweaks and architectural adjustments behind the performance boost.
#14 𝕏
Santiago recommends, when leveraging LLMs for code, building the simplest possible solution first and incrementally migrating or refactoring components only once specific pain points become clear.
#15 📝 Doug Turnbull
The tests are the code now - Argues that with AI-assisted coding, tests become the most important artifact for maintaining code quality. Emphasizes shifting attention to tests as primary maintainable representation of intended behavior.
#16 𝕏
claire vo đź–¤ shows how Jamey Gannon builds Midjourney mood boards by pulling vibe-matched images, leveraging visual language and skipping prompts to achieve consistent client work.
#17 𝕏
Mustafa Suleyman reports Copilot for Health is processing millions of AI-driven health queries, with users mainly tapping its symptom triage, medication lookup, medical research summaries and mental-wellness journaling features.
#18 𝕏
Lenny Rachitsky shares April Dunford’s guide to advanced B2B positioning—outlining four lessons to overcome hurdles like disagreement on competitive framing, product pessimism, and shaky value differentiation—to help PM teams cut through launch noise and win distribution.
#19 𝕏
Teresa Torres dives into “killing your darlings” by sharing how she intentionally sunset a revenue-generating product—cutting 40% of her income—to avoid stagnation, and outlines how team structures and org design can facilitate or hinder smooth product retirements.
#20 𝕏
NVIDIA AI outlines Jensen Huang’s five-layer “AI cake,” showing how a full-stack rebuild—from GPU hardware and software frameworks through models to real-world applications—is powering the new AI era.
#21 𝕏
Lenny Rachitsky predicts AI-powered autonomy will revolutionize farming, mining, and construction in the next 5–10 years. It’ll plug the labor gap left by retiring workers, especially aging farmers.
#22 𝕏
Demis Hassabis reflects on AlphaGo’s ten-year journey—defeating top Go players with deep neural nets and self-play (AlphaGo Zero, MuZero) and catalyzing breakthroughs like AlphaFold.
#23 𝕏
Google DeepMind builds on AlphaGo’s AI breakthroughs to automate mathematical theorem proving and accelerate scientific discoveries.
#24 𝕏
Guillermo Rauch celebrates 10 million weekly downloads of @aisdk’s npm i ai, showcasing a single package that plugs into any AI model.
#25 𝕏
Guillermo Rauch confirms that when used with Vercel AI Gateway, tokens are output raw with no markup.