In partnership with
6-min read
Could AI help with work that once took weeks?
Physicist Matthew Schwartz says Claude reproduced one paper’s results in around 20 minutes. Human experts still guided the research and checked the conclusions.
Today: What that experiment showed, Atlas’s new robot hands, Runway’s interactive video preview, FLUX 3 image controls and Microsoft’s faster voice models.
Let’s get into it 👇
Claude helped a physicist tackle complex calculations
Physicist Matthew Schwartz used Claude to build BootLoops, an open-source toolkit connecting calculations across scientific fields.
Image source: Claude-shaped science \ Anthropic
He says Claude reproduced one paper’s results in around 20 minutes, while his original code took weeks to write. BootLoops later calculated 30 physics integrals, including 15 he reports were previously uncomputed.
Human experts guided the research and checked conclusions. Schwartz cautions that correct calculations aren’t always scientifically meaningful. BootLoops is his independent project.
Some teams never seem to stop moving. They're on Attio, the agentic CRM.
Every customer signal is captured in one shared context layer, always current and compounding. Agents and workflows build pipeline, chase every buying signal, and move deals forward, an always-on revenue engine running alongside your team.
With Attio, you’ll get:
Leads automatically prioritised and routed to the right rep
Expansion and risk signals caught the moment they land
Follow-ups written in your voice, already there when you arrive
Teams like Parallel, Turbopuffer, and Wordsmith build on Attio. Are you one of them?
Could AI predict when a brain tumor will return?
Researchers developed FastGlioma using tumor tissue data from 60 patients and tested it with 20 more. It showed promise in predicting glioblastoma recurrence—but the small study has not shown that using its predictions improves care or survival.
Want it weekly?
🔥Big news in AI
🔹Atlas gets new hands for handling real tools
Boston Dynamics unveiled new Atlas robot hands with four fingers, 13 degrees of freedom and touch sensors across the fingertips and palm. The design supports manipulating objects and gripping tools while pressing their triggers. Built for AI training in simulation, the hands omit a fifth finger to reduce cost, complexity and potential failures.
🔹Runway previews video you can interact with
Runway Labs introduced Project Continuum, a research application exploring real-time video interfaces. Users can interact with animated overlays, resize scenes that adapt automatically and watch coding agents’ steps as they happen. The early preview also explores turning existing media into interactive worlds—offering a glimpse of how computer interfaces could evolve.
🔹FLUX 3 gives you more control over AI images
Black Forest Labs introduced FLUX 3 Image, supporting precise edits, bounding-box layouts, up to 4K output and 10 reference images. The company says edits preserve pixels outside the requested changes. Commercial weights support fine-tuning and self-hosting, with open weights coming soon. API use is 50% off through October 8.
The best in influencer marketing. And you’re invited.
Get ready for Return on Influence Festival '26. An entire day dedicated to influencer marketing with speakers running some of the best programs in the world. October 21, 2026, 10 am - 5 pm ET.
🛠️ Try This
Give Claude a real task—and the context to do it
This guide shows how to give Claude clear context, specific tasks and useful rules.
Learn to turn meeting notes into action lists, improve email drafts, analyze documents and organize recurring work with Projects, Artifacts, Skills and connectors.
Practical examples show how to refine responses with follow-up prompts and check the results before using them.
Microsoft’s new models can turn speech into text as you talk
Microsoft launched MAI-Transcribe-2-Streaming and two new voice models to help developers build faster, multilingual AI conversations.
The transcription model supports 60 languages, producing initial text in just over 100 milliseconds. MAI-Voice-2.1 supports 23 languages while keeping the same recognizable voice; its Flash variant targets faster, high-volume use.
Introductory transcription pricing is $0.54 per audio hour through year-end. Voice generation costs $22 per million characters, or $15 for Flash. Potential uses include customer support, multilingual assistants and tutoring.
Tired of news that feels like noise?
Every day, 4.5 million readers turn to 1440 for their factual news fix. We sift through 100+ sources to bring you a complete summary of politics, global events, business, and culture — all in a brief 5-minute email. No spin. No slant. Just clarity.
You Don’t Need Coding Experience to Start Building With AI
Build with Andrew helps beginners turn an idea into a working app with AI. This issue also includes Microsoft’s learning path on generative AI’s business value and a technical course on building assistants with on-device memory.
Inside: Remote AI jobs, cash-prize hackathons, Runway’s creative platform and new hiring research.
Want it every Sunday?
Make your next AI experiment a useful one.
Pick one task this week. Give AI the context, review the result and improve it.
Join 76,000+ readers getting practical AI news, tools and walkthroughs from The AI Entrepreneurs.
