Editorial - SiliconNoon (Page 43)

The Frontier

OceanBase tops Berkeley's data-agent benchmark with GLM-5.2

Two benchmark results and one hype check to start the week: a Chinese database outfit took the top of a Berkeley-run data-agent leaderboard on a Chinese model, a frontier model reportedly broke a century-old naval cipher, and the robot industry's own trade body published the least flattering number of the year. OceanBase, the distributed database business Ant Group spun out, has taken first place on the Data Agent Benchmark with 90.62% accuracy — the first submission to clear 90%, and it was bu

The Arena

Amazon bars Meta's Muse agent from shopping its site

Two agentic-commerce stories landed within hours of each other: Amazon enforced its own store rules against a rival's shopping agent, and Huawei put a coding model in front of every HarmonyOS developer. Amazon cut Meta's Muse shopping agent off from Amazon.com over the weekend, and told users why in a popup: continued access by an unauthorized AI agent violates the retailer's Conditions of Use. Amazon's objection is less about competition than about consent and disclosure — per GeekWire, Meta n

The Frontier

Deep Dive — Latent-space pretraining survived 5.7 trillion tokens

For as long as large language models have existed, one rule has been treated as structural rather than optional: predict the next token, and let structure emerge from the statistics. A technical report posted to arXiv on September 9 breaks that rule at a scale where earlier attempts fell apart — and the numbers attached to it are why the open-source community has spent two weeks arguing about a base model nobody can chat with. The paper is NCP-ArchPreview, from the Intern-NCP Team at Shanghai A

The Everyday

South Korea reopens a military-service route for AI researchers

Seoul reopens a path it closed in 2013 to keep AI researchers inside its biggest labs — and the first intake is already undersubscribed. South Korea will let AI researchers satisfy their military service obligation inside large corporate laboratories from 2027, reviving a route the government shut off in 2013. Eligible science and engineering researchers with master's or doctoral degrees serve three years of research instead of conventional service — the same structure that never stopped at uni

The Frontier

Kev's open Jev alternative now trains on Qwen3.5 for about $95

The decision-model category Jev created a week ago now has a full open implementation, a cost figure that undercuts every "frontier model" story this month — and a benchmark table that still puts Jev on top. Jared Palmer ported Kev, his open decision-model family, onto Qwen3.5 bases and published the 0.8B, 4B and 9B checkpoints under Apache 2.0 — and the experiment log shows the whole Qwen3.5 effort cost about $95 of rented H100 time, plus three cents of Jev API calls. Kev is not a chatbot: eac

The Frontier

AI 101 — What is multimodal AI?

Multimodal AI is a model that handles more than one kind of input — text, images, audio, video — inside the same system, rather than one model per sense stitched together. "Modality" is just the academic word for a channel of information. Text is a modality. A photograph is a modality. A voice recording is a modality. Multimodal AI reads more than one of them at once and reasons across them. That last part is the whole trick. You have had software that reads images for years — optical character

The Frontier

Tsinghua open-sources RPent, an agent layer for real robots

Two open-source releases today point the same direction: the interesting work is no longer in the model itself, but in the scaffolding around it. A Tsinghua-led team published the layer that sits between a frontier model and a robot's motors — and a leaderboard to show it lifts a frozen policy. A solo developer, meanwhile, trained a small continual-learning model on a laptop GPU and documented what it forgets. A Tsinghua-led group open-sourced RPent, a framework that turns a general model into

The Arena

Alibaba names Dayiheng Liu head of its Qwen LLM team

Alibaba finally has a name on the door of its flagship model line — the day before that name takes the Apsara main stage. Alibaba has appointed Dayiheng Liu, one of its senior AI researchers, to lead the Qwen large language model team, ending months of ambiguity about who runs the company's most strategically important product. The Information reported the appointment on Monday, citing two employees with knowledge of the matter, and said it brings clarity to the Qwen team's leadership after sev

The Everyday

Apple's Siri AI home hub is in employee testing, eyeing October

Apple's first AI-era hardware is quietly living in its own employees' houses, and Zhipu is selling Chinese enterprises a privacy switch with a few visible holes in it. Apple is testing its long-delayed home hub in employees' homes and aiming to ship it between October and early 2027, according to Mark Gurman's Power On newsletter. The device, known internally as J490, is a roughly square screen that either sits on a stand — Gurman describes it as looking like a HomePod mini cut in half — or mou

The Everyday

Jev's first mass deployment is a WeChat plugin — and its Chinese is weak

Jev's pitch was always that most AI calls in real software don't need a paragraph — they need a decision. In China that idea has found its first mass-market use, and the honest limits are already showing. TypeSafe's decision model has gone viral in China as a WeChat plugin, where it reads incoming messages and returns calibrated probabilities about what the sender actually means. Open-source projects now do the plumbing: one Windows app screenshots the chat area, runs OCR over the message bubbl

The Guardrails

White House ordered Anthropic to pull Fable — and SoftBank is funding OpenAI with junk

Politico's tick-tock of the 19-day standoff is out, and the same week SoftBank went to the junk-bond market to keep its OpenAI stake growing. Here is what both say about how the AI build-out is actually governed and financed. Politico has reconstructed the 19 days the White House spent trying to get Anthropic to pull Fable 5 — and it ends with the government doing it for them. The magazine piece, published Sunday, walks through the spring sequence: Amazon chief executive Andy Jassy raised jailb

The Arena

Altis Labs raises $25M to make an AI scan read a trial endpoint

A cancer-trial startup wants its model's reading of a CT scan to be the number a drug is judged by — and Apple's Siri home hub is already in employees' hands. Altis Labs raised $25 million to make an AI reading of a CT scan the thing a cancer trial is measured against. The round, announced September 17 and co-led by OrbiMed and Qiming Venture Partners USA, funds a company whose product is not a drug and not a chatbot: IPRO, a model that reads routine CT scans taken during oncology trials and re

See all