Daily Brief - SiliconNoon (Page 19)

The Arena

Microsoft folds coding and always-on agents into one Copilot app

Microsoft finally shipped the consolidated Copilot it promised in June — and the surprise is that an app sold to knowledge workers now writes and hosts internal apps. Two other stories worth your time: ICLR's second anonymity incident in two cycles, and Berlin's AI street cameras, which are not judging anything yet. Microsoft unveiled the Copilot "super app" it has been teasing since June: one app with three tabs — Home, Code and Autopilot — that merges Copilot Chat with Cowork and rebrands the

The Frontier

Peking University's 591-run RL study finds uniform sampling wins

A Peking University group built the comparison platform, ran the experiments, and reported the boring answer: with everything else held constant, the data policy is not what moves the score. The sharper finding is that the benchmark summary you choose can flip which method looks best. Peking University's Data-Centric AI group ran 591 reinforcement-learning experiments to test whether cleverer data policies beat plain uniform sampling, and could not find one that reliably does. The paper, DataFl

The Guardrails

Medicare's AI pilot pays contractors a share of every denial, records show

The administration's Medicare experiment pays private companies when they refuse care — and its own actuaries wrote that down months before a Senate hearing where the man running the program said they don't. Medicare's WISeR pilot pays its contractors a share of the money they save by denying care, and the administration's own actuaries said so in writing. The program — Wasteful and Inappropriate Service Reduction — launched in January in six states: New Jersey, Ohio, Oklahoma, Texas, Arizona a

The Everyday

Goncourt drops the prize novel accused of being AI-written

France's literary season just produced its first casualty of AI, and nobody involved can actually prove authorship either way. The Académie Goncourt withdrew Thélyson Orélien's novel C'était ça ou mourir from its selection on Friday 25 September, four days after the book won the Prix du Roman Fnac. The academy said the decision was driven by "the plagiarism of which its author has been guilty over time" and by "the convergent result of several analyses by credible researchers and journalists" c

The Stack

Alibaba's CUDA alternative opens to outside developers

Alibaba spent its Apsara conference showing off silicon. The more consequential announcement came the day after, when it opened the software that decides whether anyone can actually use it. T-Head, Alibaba's chip subsidiary, published a fresh round of open-source releases for T-Head SAIL on September 23 — the CUDA-style stack that sits between PyTorch and its homegrown Zhenwu accelerators — one day after unveiling the Zhenwu V900, which Alibaba CEO Wu Yongming called the most compute-capable AI

The Guardrails

The Pentagon wants $30 million for an AI lie detector

The US military is polygraphing its own officers over press leaks, and now it wants machine learning to improve the instrument. Also today: the creator of Ruby on Rails retires from hand-writing code, and Gemini starts phoning businesses for you. The Pentagon has asked Congress for $30.3 million over five years to build an AI-powered successor to the polygraph, to be run by the agency that screens federal employees. The line item sits in the Defense Department's FY2027 budget justification for

The Arena

A Chinese firm built a city traffic model on satellite imagery

A Shenzhen-listed location-data company decided the sensor city traffic was missing was never on the ground. Its answer went on show in Hangzhou this week. Merit Interactive (每日互动), a Shenzhen-listed location-data firm, has released MUMO — Modeling of Urban Mobility — a city traffic model that fuses satellite remote sensing with ground-level traffic data. The company announced the model at the fifth Global Digital Trade Expo, which runs September 23–27 in Hangzhou and drew more than 2,000 exhib

The Stack

Copilot made Okta's engineers faster but not more productive

A year-long field study inside Okta finds the individual gains are real and the organisational ones are not there. Plus: the Jev team ships a code reviewer built on the assumption that nobody reads a 230-file diff. A peer-reviewed field study inside Okta found that GitHub Copilot cut engineers' working hours and sharply lifted their motivation and perceived skill — while producing no statistically significant change in the code the company shipped. The paper, published in Communications of the

The Guardrails

Shanghai lets banks put self-trained models in front of customers

Shanghai's banking regulator just published the rules it will enforce when a bank's own model talks to a customer — and Moonshot turned the browser into something an agent can remember. Shanghai's financial regulator has issued 16 measures that, for the first time, sketch a route for banks and insurers to put generative AI in front of customers. The document — 沪金发〔2026〕19号, signed September 18 and published on the National Financial Regulatory Administration's Shanghai bureau site on September

The Frontier

Consumer AI chatbots missed skin diagnoses in up to 98% of photo cases

A new audit ran nine consumer AI assistants through 13,797 skin-disease questions. The photograph was the problem. Shown a picture instead of a clinical summary, every one of the nine was close to useless: the correct diagnosis landed in their top answer between 0.4% and 7.0% of the time. The audit, published in Diagnostics by researchers at the Plastic Surgery Hospital of the Chinese Academy of Medical Sciences, took 511 published cases of cutaneous vascular tumours, malformations and related

The Everyday

China's AI video boom: 221,900 shows, 1,055 hits

Chinese cities are paying studios to make AI video at industrial scale, and the audience has not kept up. Also: a Chinese startup opens a crowdsourced robot-training market, and the FT prices the neolab wave. Chinese local governments are competing to subsidise AI video production, and the output has already outrun the audience: 221,900 new AI shows launched on Douyin in the first half of 2026, and only 1,055 of them passed 100 million views, the benchmark the industry uses for a hit. That rati

The Arena

The AI buildout is 3.6% of GDP a year — more than any US boom

A new Brookings paper prices the US AI buildout at $10.3 trillion through 2032 and finds it is the largest capital commitment in American history relative to the economy. Separately, an engineer left Google rather than help make AI cheaper. The US AI buildout is projected to consume $10.3 trillion between 2025 and 2032 — about 3.63% of GDP every year for eight years, a share larger than any previous American infrastructure boom, including canals, railroads and electrification combined. The figu

See all