Microsoft says its AI diagnosis system got 83% of complex medical cases right in a study published in Science on 22 October, against 36% for general doctors using textbooks and the internet. The cases were simulations, not real patients, and a hospital trial comes next.
Microsoft's AI beats general doctors on hard cases, in a simulation
MAI-DxO runs several AI agents like a panel of doctors that ask questions, order tests and debate the diagnosis, Microsoft says. On emergency-department cases it scored 80% against 47% for physicians. TechCrunch notes that specialists, who usually handle cases this hard, weren't in the comparison, and that much of the improvement comes from more powerful underlying models. Microsoft will next test it with patients at Beth Israel Deaconess Medical Center in Boston.
Why it matters: adding agents that question each other raised accuracy above the model working alone, a setup worth copying wherever a wrong answer is expensive.
Meta's revenue is up 26% with 15% fewer people
Meta reported on 21 October third-quarter revenue of $64.6 billion, up 26% on last year, while headcount fell 15% to 66,775. It now expects to spend $135–145 billion on capital this year and says its AI investments are already bringing in extra ad revenue.
Why it matters: one of the largest employers in tech is growing with fewer people and credits AI for part of it, so expect the same question about hiring versus AI tools to reach smaller firms.
Legora raises $550 million to build agents for lawyers
On 21 October, Legora announced a $550 million Series E at an $8.8 billion valuation, led by Atreides Management. It has more than $200 million in annual recurring revenue and over 140,000 legal professionals using it at 2,400+ firms and in-house teams. The money goes into new agent products for intake, horizon scanning and legal research.
Why it matters: any office that runs on documents and checklists, such as CA practices and compliance teams, is next in line for this kind of tool.
Claude proves open-source code won't crash, under set conditions, and finds hundreds of bugs
Anthropic's Frontier Red Team, working with Theorem, had Claude write proofs that 239 entry points across 123 open-source libraries can't crash on inputs under 64 MB, under stated assumptions, Anthropic says. Along the way Claude found and patched hundreds of bugs. Anthropic warns that verified doesn't mean bug-free, and the proofs can take hundreds of millions of tokens to produce.
Why it matters: checks that used to be too expensive for most software are getting cheaper, so ask your software vendors how they now test the code they ship you.
OpenAI, Anthropic, Hugging Face and Alibaba hand agent projects to one foundation
The Linux Foundation's Agentic AI Foundation took on four projects on 22 October: Agent Skills from Anthropic, Open Responses from OpenAI, OpenEnv from Hugging Face and OpenSandbox from Alibaba. It now hosts 10 projects, up from three at its launch in December 2025.
Why it matters: shared standards make it easier to switch AI vendors later, so prefer tools that support them.
Goa launches an AI policy with a Konkani language model
Chief Minister Pramod Sawant and IT Minister Rohan Khaunte launched the Goa AI Policy 2026–2029 on 22 October, The Goan reports. It aims to train and certify 50,000 Goans in AI by 2030 and proposes a Konkani language model with Bhashini. The IT Department also signed an MoU with BITS Pilani.
Why it matters: state policies are where training schemes and local-language tools for regional businesses will come from, so Goa firms should watch for the scheme details.
Worth watching: how Microsoft's system does with real patients in Boston, and what support for local businesses Goa puts behind its policy.
Arthat AI
Working on something similar at your company?
Arthat AI builds these automations for Indian businesses every week. If you want to talk through whether this fits your situation, we'll give you a straight answer — even if the answer is "don't build it."
