In partnership with

SPD-BEEHIIV:93565e2f-6dcb-4c9a-8794-edcd13323a3a:R20:5690054f1ddead7ebfbf8a3fOpenAI says the upgrade can stretch subscriber allowances, while its fastest mode comes with a usage tradeoff.
| Daily issue | Wednesday, November 18, 2026 |
| | Our toolsSuperpower ChatGPT/WFH.team/Snipman/Superpower Daily New Tab | Today's briefing What matters todayOpenAI has new models and a lower price for Astra. Today’s workbench also follows Gemini’s shift from answering questions to handling errands. | Inside today's briefing | | 01 | OpenAI says its new models can stretch subscribers’ usage allowances. | | 02 | Anthropic’s billing beta removes an API-key handoff—with a security caveat. | | 03 | Google’s free Gemini tools still leave some final clicks to you. | | 04 | Physicist Suvrat Raju has paused new PhD intake after AI’s benchmark success. |
|  | Lead story / launch OpenAI releases new models and cuts Astra’s standard prices by 20%OpenAI released GPT-6.2 Astra and Sol on November 17, making both available through its developer API, ChatGPT Work and Codex. Astra’s standard input and output rates are 20% lower than GPT-6 Astra’s. Sol’s pricing stays at GPT-6.1 Sol’s level. Beyond the rate card, OpenAI says both models use fewer output tokens to complete demanding tasks, potentially letting subscribers do more before hitting usage limits. The company still recommends Astra for the hardest projects. In Codex, it says Astra is better at changing and testing existing software; early testers reported cleaner code, more focused tests and fewer revision rounds. Sol remains the cheaper option for everyday coding and writing. OpenAI says it finished within half a point of Astra on Cognition’s FrontierCode 2.0 at about one-fifth the cost, with both models using maximum reasoning effort. The test evaluates changes to real codebases, not broad equivalence between the models; it does not show they are interchangeable across tasks. The fastest mode brings a separate tradeoff. OpenAI says Ultrafast generates tokens at up to 4.5 times the standard rate but consumes more allowance—not a promise that whole projects finish 4.5 times faster. The $200-per-month Pro plan gets Ultrafast for Sol; the $500 plan gets it for both models. Eligible Enterprise and Edu workspaces can enable it for both through an administrator. 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.
 | launch Anthropic launches a billing beta for outside Claude tools without shared API keysAnthropic’s linked billing beta lets supported tools charge your existing organization without receiving its API key. Customers can use existing spending commitments, set per-platform limits and disconnect tools in Console. The catch: requests run in the tool provider’s organization, so your IP restrictions, encryption-key settings and data-retention rules do not carry over. Continue reading ↗ |
|  | A tool for your workflowWFH.team | | A focused feed of carefully selected remote roles and practical work-from-home resources. | |
|  | launch Google rolls out free Gemini tools for appointments, bills and shoppingGoogle began rolling out free Gemini task tools across six countries, with advanced features starting in the U.S. Users can delegate appointment calls, bill reviews and shopping, with spending caps and approval controls for critical actions. Some sites still hand the final checkout step back to the user. Google says tasks run in isolated virtual machines with defenses against unsafe links and prompt injection. Continue reading ↗ |
|  | benchmark Physicist pauses new PhD intake after AI scores 49 out of 50 on research testTheoPhysEval co-creator Suvrat Raju says he has stopped taking new PhD students while assessing their prospects. The benchmark authors report that GPT-6-Astra, running in Codex with search and calculation tools, solved 49 of 50 unpublished problems at an average cost below $3 each. That tests checkable research calculations, not independent major discoveries. Physicists interviewed by Understanding AI question how trainees will build intuition if agents do the calculations. 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 | | | Non-coders and curious professionals will learn to turn a website idea into a working, publicly deployed prototype using ChatGPT Codex and Vercel, with hands-on guidance to build their own projects. No coding experience is required, just bring a laptop and an idea. | | See the event → Explore more events |
| Daily tool drop 6 AI tools worth knowing today | |
| | | The Internet Had a Point | Meme from the stashThe AI Writing Gym Rush |  |
| |
| Your daily challenge Relay / Easy / 3–5 min Start with L. End with D. Can you use every letter? Take the challenge →Explore all games | | Need a new website? Try BayBold, a Bay Area web design agency. | | AI PULSE / YOUR PERSPECTIVE In the past 30 days, how has using AI affected your working hours? 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. |
|