In partnership with

SPD-BEEHIIV:861c2d53-2b81-4a62-840b-d27def422cd1:R27:fe567976d48bbb9f12f06501Eligible Ultra subscribers get access and a gallery for sharing and remixing worlds. Pro users get a waitlist.
| Daily issue | Thursday, November 19, 2026 |
| | Our toolsSuperpower ChatGPT/WFH.team/Snipman/Superpower Daily New Tab | Today's briefing What matters todayGoogle is letting eligible subscribers change AI worlds from inside them. Meanwhile, OpenAI says some of its chip-design edits have reached a design sent for manufacturing. | Inside today's briefing | | 01 | Google’s newest AI worlds arrive with some familiar research hurdles. | | 02 | OpenAI’s reported chip-design gains come with an important distinction. | | 03 | The Homeland Security chief’s AI outage theory remains unverified. | | 04 | Anthropic’s IPO reportedly priced above range before today’s scheduled debut. |
|  | Lead story / launch Google lets eligible Ultra subscribers change AI worlds as they exploreGoogle began rolling out Genie 4 through Project Genie on November 18, giving eligible Google AI Ultra subscribers a way to change AI-generated environments while exploring them. The experimental browser-based tool now accepts text prompts during a session: users can change the weather, introduce a character or redirect the action without leaving the world. Audio is generated in real time alongside the visuals, adapting ambient sounds and character dialogue to the scene. A new creation agent uses conversation to build the initial setting, visual style and background score. Gemini can also participate during exploration, observing the screen to suggest where to go next or automatically trigger events. A community gallery turns those creations into starting points for others: users can alter a world’s structure, soundtrack or rules, then publish the remix. Genie 4 does not replace every use for its predecessor, though. Google keeps Genie 3 available for longer sessions and panorama-based worlds, raising its session limit from one minute to three minutes per generation and adding links between worlds. Google still identifies camera latency, complex physical interactions, precise real-time audio and long-term scene permanence as research challenges. Those are limits on how an environment responds and holds together, not just how it looks. Google says generated video and audio carry SynthID watermarks, with safety screening for model inputs and outputs. AI Pro subscribers can join a waitlist, but have no announced access date. Read full story ↗ |
|
|
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.
 | research OpenAI says AI improved selected AMD GPU circuits by up to 14%OpenAI says model-written changes to selected AMD GPU circuits delivered up to 14% improvement in power, performance and area—not a 14% speedup for whole GPUs. Some edits are in a design sent for manufacturing; others are in files for future versions. Synopsys tools checked that the rewrites preserved circuit behavior. OpenAI and Synopsys are developing GPT-Synopsys for chip-design workflows. Continue reading ↗ |
|  | A tool for your workflowWFH.team | | A focused feed of carefully selected remote roles and practical work-from-home resources. | |
|  | security risk Homeland Security chief shares unverified AI theory for recent outagesHomeland Security Secretary Markwayne Mullin relayed an unnamed contact’s theory that self-modifying AI software could overwhelm surrounding systems. Operators described separate software and infrastructure faults, and Microsoft explicitly ruled out AI agents. The FAA’s root cause remains under investigation, while Transportation Secretary Sean Duffy said its SMART AI tool played no role. Continue reading ↗ |
|  | business Anthropic reportedly sets $83 IPO price ahead of scheduled Nasdaq debutFinancial outlets citing sources put Anthropic’s IPO price at $83 a share, above its planned range, ahead of its scheduled Nasdaq debut today. NBC reported a target market value of $1.5 trillion and first-half revenue above $16 billion. The listing arrives amid a technology selloff and concerns about AI spending, debt, costs and demand—making its reception a useful, but not decisive, read on investor appetite. Continue reading ↗ |
|
|
Attio is the #1 fastest-growing CRM on Brex and Ramp.
Built for the most ambitious companies in GTM, Attio is an end to end revenue platform that allows teams to run multiple plays across different revenue motions and lifecycles.
Teams at the frontier are already building on it: Granola, turbopuffer, Modal, and many many more.
| | Beyond the feed / AI events | | | The Open Court Hackathon with Wonder Studios will bring creative technologists, developers, and builders together to prototype interactive experiences with real-time video models, with model access, credits, and £700 in prizes. Wonder co-founder and Chief Creative Officer Justin Hackney, formerly Creative Director at ElevenLabs, will open with a showcase on the technology’s creative potential. | | See the event → Explore more events |
| Daily tool drop 6 AI tools worth knowing today | |
| | | The Internet Had a Point | AI Models and Unfixable Hardware Bugs |  |
| |
| Your daily challenge Gather / Moderate / At your pace Seven letters. One must-use tile. How much can you gather? Take the challenge →Explore all games | | Need a new website? Try BayBold, a Bay Area web design agency. | | AI PULSE / YOUR PERSPECTIVE Would you let an AI agent spend up to $100 without asking before each purchase? Take part → | Reader check-in Help shape tomorrow's briefingOne click tells us what to keep, improve, or tighten. | |
| Prefer one email a week? Get the Monday briefing, including Friday and weekend news. |
|