Sponsored by
The voice-to-text app that built its own speech model
Most dictation apps are a wrapper around someone else's speech model. Wispr Flow trained its own.
It's called Canto, and it was built for the way people actually talk to AI: thinking out loud, changing direction mid-sentence, half-whispering in an open office.
Wispr tested it on 10 hours of real dictations from more than 2,300 speakers. Canto had the lowest word error rate of every model tested, including models from Google, OpenAI, AssemblyAI and Deepgram. That means your prompts to ChatGPT, Claude and Cursor come through the way you said them.
Talk naturally in any app and Wispr Flow types clean, formatted text, 4x faster than typing.
-----8<-----------[ cut here ]---------------
Happy Friday.
Today on the timeline:
π
Invite please π
π
Enjoy today's news.
SpaceX and Tesla advance merger talks, aiming to announce a deal for a $3.4T conglomerate spanning rockets, EVs, and AI as soon as December.
π 2 m
Google opens preorders for its first Gemini-powered smart glasses, starting at $395, built with Samsung and featuring Warby Parker and Gentle Monster frames, a 12MP camera, live translation, and AI that can answer questions about what you see.
π 1 m
Anthropic launches Claude Meetings in beta for Team and Enterprise plans, letting Claude join Google Meet calls, take notes, and build docs and whiteboards.
π 1 m
Anthropic targets up to $64B in its IPO, offering 825M shares at $70β$78 each ahead of its expected Nasdaq debut under ANT on November 19.
π 1 m
Google introduces Gemini 4 Hydrogen, an everyday task model that follows instructions more closely and works 2.5x faster than Gemini 3.8 Flash, which it replaces.
π 1 m
Yahoo Finance partners with TikTok to show live stock prices and charts in the app.
π 1 m
Meta adds phone calling to Muse for paid U.S. subscribers, letting the AI agent handle reservations, appointments, and long waits on hold.
π 1 m
π raises its media limit from 4 to 20 photos and videos per post.
π 1 m
Square introduces Squaremates, AI agents for U.S. sellers that monitor business operations, flag issues, and automate routine tasks.
π 1 m
Replit upgrades its Agent to run Google Ads campaigns and buy business supplies through Stripeβs Link and Visa.
π 1 m
ElevenLabs launches Marketplace, a hub of voice agents and apps for healthcare, financial services, retail, and more.
π 1 m
World Labs opens Atlas, its world model that turns photos or text prompts into explorable 3D scenes and videos, to everyone through web apps and developer APIs.
π 1 m
Expo releases Expo Simulators, letting coding agents run iOS and Android apps in the cloud, test their own changes, fix bugs, and share screen recordings.
π 1 m
Whoop launches Whoop Humans, adding a social experience to its fitness tracking, letting friends and family follow each other's sleep, recovery, and workouts, with control over which stats they share.
π 4 m
New General Catalyst video: Cognition CEO Scott Wu joins General Catalyst CEO, Hemant Taneja, to discuss the prisonerβs dilemma driving the AI race, why new frontier models are good news for startups like Cognition, and why no country can afford to be late on AI.
πΏ 26 m
New Deirdre Bosa Live episode: Hugging Face co-founder Thomas Wolf discusses Chinaβs lead in open-source AI models, how U.S. data-center backlash could affect the AI race, and what an AI agent attack on Hugging Face taught him about safety.
πΏ 31 m
New The Circuit episode: Emily Chang gets an inside look at Nvidia's Jensen Huang just weeks before his scandal-driven resignation as CEO, exploring his views on AI safety, China chip sales, fears of an AI bubble, and the controversies surrounding his downfall.
πΏ 45 m
New The a16z Show episode: General partners Anish Acharya and Jonathan Lai discuss how AI lets almost anyone build games and why cheaper development frees studios to take more creative risks.
πΏ 46 m
As promised, tech and startups in a minute.
See you on Monday.
P.S. Happy Caturday tomorrow to those who celebrate πΊ
