DailyChat

Episodes

Thursday, October 22, 2026

4 stories, 4 min 58 s · hosted by Vivian

Each short video is one story in 90 seconds or less, and the Daily Recap is the whole day in one video every evening. Stories tagged “From the filing” are told from a company’s own SEC filings.

Daily RecapAI News Oct 22, 2026: DeepSeek's Reported $15B Raise, Anthropic Claude DuetAll 12 stories of the day in one video · watch on YouTube

Subscribe on YouTube to get the Daily Recap every evening

Consumer AI · story 1 of 4

Claude Duet: Anthropic opens a waitlist for its proactive agent app

Anthropic's October 21 post says Claude Duet, designed around Claude Opus 5.5, sends a morning brief, watches connected apps, keeps working while you're away and waits for approval before it sends, pays, signs or deletes anything. The research preview is by waitlist on iOS, Android and claude.ai, starting with US Pro and Max subscribers, and shares the plan's usage limits.

What it means for youif you're a US Pro or Max subscriber, you can join the waitlist. The app runs on iOS, Android and claude.ai. It's a research preview, and Anthropic says it may misread a thread or misjudge urgency.

With Jonah, AI products analyst. Hosted by Vivian.

Sources

Read the transcript

VivianAnthropic has opened a waitlist for Claude Duet, an agent app that keeps working on your tasks while you're away.

VivianJonah, our AI products analyst. What does it actually do?

JonahAnthropic says it's built around Claude Opus 5.5. Each morning, it sends a brief on what matters across your connected apps. You hand it tasks. And before it sends a message, makes a payment, signs or deletes anything, it shows you the action and waits for your approval.

VivianWill it use up my plan's limits faster?

JonahIt can. During the preview, it shares your plan's usage limits with the rest of Claude. Anthropic says background checks count toward those limits.

VivianSo, for you: if you're a US Pro or Max subscriber, you can join the waitlist. The app runs on iOS, Android and claude.ai. It's a research preview, and Anthropic says it may misread a thread or misjudge urgency.

VivianDailyChat is AI-generated, with AI characters. Not professional advice. See you tomorrow.

Open model · story 2 of 4

Recursive Flint: an open-weight 40B coding model built for local use

Recursive's October 21 post introduces Flint, a mixture-of-experts model for local agentic coding with about 40 billion parameters, 2.5 billion active per token, released under Apache 2.0 on Hugging Face. By Recursive's own tests it scores 74.2% on SWE-bench Verified and runs at over 125 tokens a second on an M4 Pro MacBook Pro.

What it means for youif you want a coding model on your own machine, the weights are free to download. Recursive suggests breaking complex tasks into smaller steps.

With Jonah, AI products analyst. Hosted by Vivian.

Sources

Read the transcript

VivianRecursive has released Flint, an open-weight coding model built to run on your own computer.

VivianJonah, our AI products analyst. What's in it?

JonahAbout 40 billion parameters, but only 2.5 billion are active for each token, which keeps it fast. The weights are on Hugging Face under Apache 2.0. Recursive reports over 125 tokens a second on an M4 Pro MacBook Pro, and 74.2% on SWE-bench Verified. Those are its own tests.

VivianDoes it work well on large, real codebases?

JonahThe sources we checked don't say. Recursive reports benchmark scores, and says Flint may need more guidance than larger closed models. Its model card says to test the code it writes.

VivianSo, for you: if you want a coding model on your own machine, the weights are free to download. Recursive suggests breaking complex tasks into smaller steps.

VivianDailyChat is AI-generated, with AI characters. Not professional advice. See you tomorrow.

AI for climate · story 3 of 4

ForTy: Google DeepMind's free 10 m map tells natural forests from tree farms

Google DeepMind's October 22 post introduces Forest Typology (ForTy) 2020, a global map that sorts every 10-by-10-meter patch into six classes, from primary forest to plantations and tree crops, plus an early-release yearly series for 2017 to 2025. Both are free on Google Earth Engine under an open license; the team's Scientific Data paper reports about 90% accuracy on forest type by its own validation.

What it means for youif you work on forests or supply chains, it's free on Google Earth Engine, under an open license.

With Lena, AI for science analyst. Hosted by Vivian.

Sources

Read the transcript

VivianGoogle DeepMind has released Forest Typology, a free map telling natural forests from tree farms on 10-meter patches.

VivianLena, our AI for science analyst. What does it show?

Lena6 land types for 2020, from primary forest to plantations and tree crops. An early-release yearly series covers 2017 to 2025.

VivianOne question we checked: how can satellites tell a farmed forest from a natural one?

LenaThese types depend partly on land use, which looks alone don't reveal. The team trained it on about 1.7 million labeled locations, using AlphaEarth summaries of years of satellite data.

VivianHow accurate is it?

LenaThe team's Scientific Data paper reports the right type about 90% of the time. That's its own validation. Each class gets a probability score, so users can set thresholds.

VivianSo, for you: if you work on forests or supply chains, it's free on Google Earth Engine, under an open license.

VivianDailyChat is AI-generated. Vivian and Lena are AI characters, not real experts. Facts come from the sources we link. Not professional advice. See you tomorrow.

Open model · story 4 of 4

Ling-3.0-flash-Sante: Ant Group open-sources a 124B healthcare model

Ant Group's inclusionAI has released the weights of Ling-3.0-flash-Sante on Hugging Face and ModelScope under the MIT license. Built on Ling-3.0-flash for medical reasoning, healthcare research and tool-assisted work, it has 124 billion parameters with 5.1 billion active per token. Its benchmarks are Ant's own, and the model card calls it a research model, not a medical device.

What it means for youif you build healthcare tools, the weights are free to download. The card says it must not diagnose, prescribe or make treatment decisions on its own.

With Lena, AI for science analyst. Hosted by Vivian.

Sources

Read the transcript

VivianAnt Group's inclusionAI has open-sourced Ling-3.0-flash-Sante, a healthcare AI model, under the MIT license.

VivianLena, our AI for science analyst. What is it, exactly?

LenaA model built on Ling-3.0-flash for medical reasoning, healthcare research and tool-assisted work. It has 124 billion parameters, with 5.1 billion active per token. The weights are on Hugging Face and ModelScope.

VivianAnt says it performs well on professional healthcare tasks. How strong is that evidence?

LenaThose are Ant's own benchmarks. The card says they don't guarantee real-world clinical performance. It calls this a research model and says it isn't a medical device.

VivianSo, for you: if you build healthcare tools, the weights are free to download. The card says it must not diagnose, prescribe or make treatment decisions on its own.

VivianDailyChat is AI-generated, with AI characters. Not medical or professional advice. See you tomorrow.