Simon (@tokumin) header
Simon (@tokumin) profile picture

Products @GoogleLabs. More LaMDA moments.

latent space
Joined February 2007
Photo by Simon (@tokumin)Photo by Simon (@tokumin)Photo by Simon (@tokumin)Photo by Simon (@tokumin)Photo by Simon (@tokumin)Photo by Simon (@tokumin)Photo by Simon (@tokumin)
under promise. over deliver.
They undersold this, it feels rather magical. I entered a toothpaste website and had 2 ai agents having a podcast convo on “ if they are really telling the tooth 🦷 “ Worth trying it out.
1
56
35,370
Simon retweeted
Why we built this: A month ago I wanted to explore how users could accomplish a goal. I wanted an agent to sketch some quick mockups. It took my agent minutes, and I forgot and did something else. Now with Stitch Rapid it takes literally 3 seconds...
What if you could generate an entire product flow in just five seconds? Early design is rapid and iterative. To amplify your creativity, you need a synchronous design partner that keeps up with your train of thought. So we built a rapid design experiment in Stitch. It’s powered by the Gemini Decisions API to help explore what happens when early design iteration takes seconds instead of minutes ⚡️: 5️⃣ seconds from rough prompt to a multi-screen flow 2️⃣ seconds from tap to new variations 1️⃣ click to dive into Stitch for deeper UI polish ⚠️This is an early MVP to test how fast early ideation can feel, and we’ll be folding your favorite parts into the full Stitch experience soon. Look for the banner in Stitch to try it yourself or see how it works below 🧵👇
1
1
9
704
Honestly Gemini 4 Carbon >> Fable 5.2
50
8
1
364
12,332
Ok. Carbon is *really* good at code.
60
35
27
707
40,239
same
i think i like carbon more than fable
40
1,560
Simon retweeted
Cry. When we recorded the video Sundar used it was staging hardware and a pre-release model. Now in production the same flow is about twice as fast! (Everything 1x)
Today we’re launching the Gemini Decisions API, powered by our new Gemini Decision 1 model in @GoogleAIStudio. Instead of generating text token by token, it’s built for closed-set choices scoring up to 1,000 questions across text, images, audio, video + PDFs in a single pass. A few details: - 3 decision types: yes/no, multiple choice (up to 255 options) + rubric scoring - $0.10 / 1M input tokens with $0 billed for output tokens (~5x cheaper than Flash-Lite) - leads on several benchmark evals with lowest calibration error Google teams are already building with it, below is a look at Stitch Rapid designing a full product flow in seconds (1x speed)
1
1
9
481
Simon retweeted
Introducing Tabletop, a collaborative space from @GoogleLabs where you can create custom tools to build, plan, and do more with your friends, family, or club. Send an invite to your group and start a shared hub where everyone can contribute simultaneously for your next trip, potluck, or fantasy league. You can create live dinner polls, trip itineraries, expense splitters, and sports brackets tailored to your group.
47
72
18
665
118,874
Stitch Rapid: design at the speed of thought. Speed is killer feature and we've used Gemini Decision 1 to unlock a realtime iterative design experience in @stitchbygoogle that's unlike anything I've ever seen. 👉 stitch.withgoogle.com/rapid
Today we’re launching the Gemini Decisions API, powered by our new Gemini Decision 1 model in @GoogleAIStudio. Instead of generating text token by token, it’s built for closed-set choices scoring up to 1,000 questions across text, images, audio, video + PDFs in a single pass. A few details: - 3 decision types: yes/no, multiple choice (up to 255 options) + rubric scoring - $0.10 / 1M input tokens with $0 billed for output tokens (~5x cheaper than Flash-Lite) - leads on several benchmark evals with lowest calibration error Google teams are already building with it, below is a look at Stitch Rapid designing a full product flow in seconds (1x speed)
3
30
1,371
Simon retweeted
The kids were very excited for this 😂
Want to see how our team actually uses Tabletop? 👀 Founding engineer @romannurik gave us a tour of his family’s space— showing how Tabletop can be as useful, quirky, or wonderfully specific as you want it to be: 🪙 Interactive chore charts & piggy banks 🔍 Rainy-day scavenger hunt with live score tracking 🗺️ On-the-whim LEGO maps 🧝‍♀️ Fantasy book-themed NYC trip Pull up a chair and join the waitlist (US, 18+): labs.google/tabletop
6
2
29
1,911
make, share and use apps in group chats - Tabletop brings the live collaboration seen in Google Docs to anything! tabletop.google/006zqiWNqiP6…
🚨 NEW LABS EXPERIMENT 🚨 Introducing Tabletop 🎲— our newest experiment built to get your crew from "we should totally do that" to "we’re actually doing this." Drop links and files into one shared space, add your crew, and ask Tabletop for whatever you need to create custom spaces on the fly: 🗺️ Interactive road trip maps & dinner-decider wheels 🏆 Live pickleball brackets & merch voting boards 🪑 Drag-and-drop seating charts (so you don’t get stuck next to your boss) Pull up a chair and join the waitlist (US, 18+): labs.google/tabletop
13
824
Thanks, @melvinjohnsonp, and a tip of the hat to the incredible work your team did on science! These models will unlock amazing advancements in human health over the very short term. Keep that pedal to the floor :)
Another great private benchmark by @danielmckinn0n. Carbon shines on both rare disease detection and lab experiment prediction.
1
2
8
670
Big G remains 👑 of rare-disease evals. @GoogleDeepMind Gemini 4 Carbon and @AnthropicAI Fable 5.2 tie for first on @GamowLabs RareBench, solving a staggering 61.5% of problems. When I started this work, GPT 5.4 only eeked out 9%. Anyone saying AI will not transform healthcare is out of their mind at this point. Stay tuned for an updated, harder RareBench v2 now that this task is close to saturated. Gemini Carbon stands alone in first place on LabBench, which measures the model's ability to interpret real rare-disease laboratory data and predict the next experiment. A tip of the hat to both GDM and Ant for their tremendous work in this space. I am very confident that we are getting on a trajectory to solve rare disease in the next few years.
12
27
4
266
40,493
Today we’re rolling out Gemini 4 Carbon, our strongest model with frontier performance in cybersecurity capabilities and long-horizon agentic workflows. We’ve poured a lot of love into it, and it’s very good at working through hard problems.
90
119
18
2,318
356,834
Really like what Vincent and team are doing with Quest - I know this has been a true labor of love for him and the team. AI is 10x more useful when it really “gets you”, so focusing on solving that problem is smart and pretty unique.
Introducing Quest, the personal agent that studies you like a profiler. Quest is for believers-only. It requires radical transparency: full access to your life so your agent can craft your memory. If you give it a shot, you’ll unlock a personal advisor, always in your corner, helping you and encouraging you to start your quests. Quest is built different: 1. Memory from day one: 100+ wiki files created on you, all your relationships mapped. 2. Phone-Use: Quest has its own Android phone, and can control messaging apps like WhatsApp, Instagram or TikTok. 3. Fluid UI: Quest builds tailor-made iOS mini-apps around each of your quests. 4. Mothership of AI Assistants: Quest can manage your Instinct and Muse on your behalf. 5. Self-Improving Proactivity Harness: Quest reacts to anything, not just your email, and its proactivity evolves from your feedback and its own mistakes.
2
3
1
33
2,203
Simon retweeted
Everyone's busy bragging about AI coding. The real win is Gemini Notebook: you can literally talk to your notes and learn anything, and it's free.
This is now available to ALL users on mobile 🥳📱 Have you used it yet? Let us know what you think!
4
2
36
4,085
Organize the worlds information and make it universally accessible and useful
Excited to announce the addition of hundreds of textbooks to the Expert Intelligence system in Gemini Notebook! Most are offered without license fees, which makes this one of the largest pools of open knowledge in the world! blog.google/innovation-and-a…
4
3
35
2,455
The solution to this is to build your own tools with an RSI loop based on success metrics you define.
The thing with "AI workflows" is that they are constantly evolving. The way I used AI to get work done 2 weeks ago is different than it is today, and I know it will be different again next week - heck, it might even be different this afternoon. I find myself constantly learning, unlearning, and adapting...and there isn't a playbook on how to do it best right now. The result is that there are many different feelings about it: exciting, exhausting, intimidating, fun
1
1
8
1,022
I can picture all my students pulling out their phones in class to record with Gemini Notebook.
🚨 MAJOR MOBILE FEATURE 🚨 You can now record up to 1 hour of live audio directly in the mobile app. Capture lectures, team meetings, or your latest mad scientist rants and instantly use them as sources. Pro tip: Combine your live audio recordings with photos of your handwritten notes to create the ultimate personalized workspace
1
2
34
3,897