Underdog showcased its 32MB Woof model using an interactive, retro-themed browser demo where users can type to hear spoken words or tune different music styles. The software is part of a broader effort to deliver open, American-made AI that runs entirely on personal hardware to keep user information secure. It follows the startup's previous launch of Bark, a 167MB speech-to-text tool intended for local dictation and transcription. Underdog is also building a 27-billion-parameter laptop model alongside a waitlist application to compete with services like Muse and Instinct.
Key points
On-device speech models prioritize user privacy by eliminating the need to send audio dat…: On-device speech models prioritize user privacy by eliminating the need to send audio data to cloud services.
Compact architectures enable specialized functions like private voice assistants and cust…: Compact architectures enable specialized functions like private voice assistants and custom podcasts without heavy computing requirements.
Underdog is expanding its ecosystem with larger laptop-ready systems and alternatives to…: Underdog is expanding its ecosystem with larger laptop-ready systems and alternatives to established platforms like Muse and Granola.
Sources checked by our analysis pipeline: 1
AI companies across the world are racing to make bigger and better models, but the problem is all of them need massive amounts of processing power. On the other end, some researchers and startups are creating smaller models that can run on edge devices, like your phone or computer.
One such startup, Underdog, has released a game built with its new voice model called Woof that’s only 32MB large.
The game features a room with a retro tech theme, a typewriter, and a CRT monitor. There is a desk lamp, which makes a sound when you click its switch. The intro warns users about “intrusive” cloud providers in the town — ostensibly a jab at cloud-based models.
You can interact with the typewriter and music tuner.
You can type anything on the typewriter using your keyboard, and the game will speak it back. You can also rotate the tuner to generate different kinds of music.
All of this works within the browser through Underdog’s speech generation model, which, the company says, can run on most devices without any internet connection.…
The retail giant is expanding conversational commerce by turning its AI assistant into a direct messaging contact saved on user phones. Customers can submit product photos, request gift suggestions, or automate grocery restocking without opening a dedicated browser or shopping app. This update mirrors a broader tech trend prioritizing lightweight text interfaces for complex digital services.
The Wall Street investment bank updated internal guidance to restrict advanced programming assistants exclusively to dedicated software developers and quantitative strategists. This pivot cuts off thousands of frontline traders and bankers who had spent months leveraging the platforms to analyze market moves and model trading flows. The decision reflects growing industry scrutiny around software costs, security risks, and actual productivity returns.
The latest general availability release delivers native video generation upgrades targeting enterprise media workflows through the Gemini API and Gemini for business. Creative teams gain expanded camera path controls alongside multi-turn editing features designed to adjust visual components without restarting production. These additions lower operational expenses by combining drafting resolutions with a steep reduction in generation rates.