In partnership with
Did you know that we have LinkedIn, Instagram and X accounts
that you can follow?
In this issue:
🤝In Partnership: Get answers from your data in minutes
🤿Deep Dive: A new Google chip could etch part of Gemini's architecture directly into silicon
🤝Powered by: Stay informed in just a few minutes
🤿Deep Dive: Nvidia's tiny model for big robots
⚒Tool Snapshots: Tools for AI, no-code, and productivity
🖼AI Art: Examples of great and trending AI art
🤝IN PARTNERSHIP WITH CUBE
The BI workflow, now run by agents.
A question about your own business data still means a ticket, a queue, and a week of waiting.
Cube hands that workflow to AI agents: they model the data, explore it, and build the reports — your team guides and approves — so week-long analysis comes back in minutes. And because every agent works from one governed set of metric definitions, the answers come back right. 400+ companies run on Cube, including Brex, Wix, Patagonia, and Webflow.
🤿 DEEP DIVE
Google is Said to be Building Specialized Silicon Around Gemini's Model Architecture
Intelligence: Google is reportedly developing a chip called Frozen v2 that would etch part of Gemini's architecture into the silicon itself, potentially serving six to ten times more tokens per watt than current TPUs. The project is partly a response to a compute shortage severe enough that Google Cloud has turned down outside deals, with deployment targeted for 2028.
What standout is that Frozen v2 locks in Gemini's architecture but keeps the weights updatable, so the chip stays useful across future Gemini releases as long as the underlying architecture doesn't change.
An earlier version of the project, led by Google DeepMind chief scientist Jeff Dean, would have baked Gemini's actual weights into the chip, but was shelved for having too short a lifespan.
Google doesn't plan to produce Frozen v2 at TPU volumes, treating it more as a trial run for specialized silicon.
The chip would sit alongside Google's existing TPU line, which split into separate training and inference variants with its eighth generation.
A Toronto startup called Taalas has already demoed a chip with Llama 3.1 8B permanently wired into the silicon, claiming 17,000 tokens per second per user.
🤝POWERED BY THE FLYOVER
Cut Through Noise with The Flyover!
One Email with ALL the News
Ditch the Mainstream Bias
Quick, informative news that cuts through noise.
🤿 DEEP DIVE
Cosmos 3 Edge Brings Real-Time World Modeling and Robot Action to Edge Devices
Intelligence: NVIDIA released Cosmos 3 Edge, a 4 billion parameter open world model that helps robots and vision AI agents understand their surroundings and generate actions in real time on edge devices. It runs on hardware like Jetson Thor and ranks first among similarly sized models on the VANTAGE-Bench vision analytics benchmark.
Image credit: Nvidia
The standout feature is that Cosmos 3 Edge can work in both directions, predicting the visual outcome of an action or inferring what action caused a change, using a shared representation across vision, language, audio, and action.
The model combines two transformer towers, an autoregressive tower for reasoning and a diffusion tower for prediction and generation, that share multimodal attention layers.
On NVIDIA Jetson Thor, the model generates 32 actions per inference while achieving real-time control at 15 Hz.
NVIDIA is also releasing Cosmos 3 Edge Policy, a robot manipulation policy post-trained on the DROID dataset for pick-and-place tasks.
A separate Cosmos 3 Super 4-Step distillation checkpoint cuts diffusion steps from 35 to 50 down to just 4, delivering up to 25 times faster inference.
⚒ TOOL SNAPSHOTS
Futuristic tools within AI, no-code, and productivity
🎯 Lev8 - Finds the right people to reach so you spend less time hunting for leads.
🧩 ditto.site - Saves you from rebuilding the same UI from scratch.
🎙️ Routine AI - Lets you stay organized without constantly switching between apps.
🌐 Phantomstory - Gives your brand more opportunities to be discovered in AI search.
🖼️ Jockey by TwelveLabs - Makes finding old photos and videos feel as easy as searching the web.
🖼 AI ART
Examples of great and trending AI art
Images by WorkLucky4366
ℹ️ ABOUT US
The Intelligent Worker helps you to be more productive at work with AI, automation, no-code, and other technologies.
We like real, practical, and tangible use-cases and hate hand-wavy, theoretical, and abstract concepts that don’t drive real-world outcomes.
Our mission is to empower individuals, boost their productivity, and future-proof their careers.
We read all your comments - please provide your feedback!
Did you like today's email?
What more do you want to see in this newsletter?
