Deep Dive - SiliconNoon (Page 8)

The Arena

Meta's $10B Anthropic tab outs the AI race's awkward economics

For months, Mark Zuckerberg has cast Anthropic — and OpenAI — as the closed, centralized dangers his open-source movement exists to fight. His own spreadsheets tell a flatter story. Sources told The New York Times that Meta's internal projections assumed it could spend as much as $10 billion a year on Anthropic's models, at the same time Zuckerberg was publicly trashing the lab as a rival. The report, published Aug. 27, frames the pair as the industry's most cosy frenemies: competitors in the se

The Stack

Deep Dive — Nvidia's quiet takeover of the AI stack

Nvidia just posted a $96.2 billion quarter, guided to another year of ~70% growth well into fiscal 2028, watched its stock climb 7%, and — hours later — entered serious talks to buy Hugging Face for over $13 billion. Read separately, those are two feel-good headlines about an AI industry that can't stop spending. Read together, they are something more pointed: the company that powers almost every AI datacenter is moving to own the layer where models are distributed, hosted, and served — not just

The Frontier

GLM-5.3-Flash proves frontier intelligence is now a Chinese-chip economics story

The model that quietly ruled the OpenRouter leaderboards all week finally acknowledged its identity today, and the technical detail buried in the announcement is more consequential than the reveal itself. Z.ai confirmed that GLM-5.3-Flash — launched this morning — is the same engine that ran anonymously as "ox-alpha" and became the week's most-used open model. That much we expected, having watched sleuths chase the attribution from Mystery model Ox Alpha tops GPT-5.6 — sleuths say it's Zhipu's G

The Arena

Deep Dive — DeepSeek's leaked P&L shows frontier AI can run on a shoestring

A lab that spent the year as everyone's favorite underdog just handed the industry its first real balance sheet, and it reads like an argument about what AI should cost. DeepSeek generated roughly $70.7 million in revenue over the first seven months of 2026 while losing $106 million, according to figures The Information reported this week, citing sources. The company has not confirmed the numbers, and no filing backs them yet — treat them as sourced, not filed. But if they are even roughly righ

The Arena

Jalapeño's first benchmarks put OpenAI on Nvidia's turf

OpenAI's decision to design its own inference chip stopped being a rumor and became a competitive threat today. Fresh off a June announcement, the company published the first real measured results for Jalapeño, the accelerator it co-built with Broadcom — and the numbers, if they hold, are the kind that supply-chain analysts and Nvidia's investor-relations team both lose a little sleep over. The headline "work per watt" gains we flagged in our morning brief are worth sitting with, because the dee

The Frontier

Deep Dive — Marin's 535B run puts open model training on livestream

A 535-billion-parameter model is training right now with its code, data mixture, loss curves, and even its failures posted in public as they happen. The Marin project started the run this week and wants the world to watch the experiment unfold rather than read about it in a retrospective paper three months from now. It is the starkest test yet of a question the open-weights community has never fully answered: can transparency at frontier scale be real, or is it just branding? We covered the lau

The Guardrails

Taiwan's AI-server indictment exposes the gap in chip export controls

Prosecutors in Keelung charged nine people — including an Nvidia employee and two Super Micro staff — with helping high-end AI servers reach China, in a scheme that leaned on forged paperwork and multi-stop transit. It is the cleanest public window yet into how the chip black market actually moves hardware, and it underscores that export controls only matter if the people holding the paperwork enforce them. What happened, and who's charged Taiwanese prosecutors on Monday indicted nine defenda

The Frontier

Deep Dive — AI co-reader caught 15 liver cancers radiologists missed

Alibaba's research arm, DAMO Academy, has published a medical-imaging AI model in Nature Medicine that reads the contrast-enhanced CT scans radiologists already order and flags liver cancers the human eye missed. In a two-month deployment inside a routine clinical workflow, the model helped clinicians find 15 malignant tumors — most of them around a centimeter in diameter — that had been overlooked, and changed treatment plans for patients as a result. We should be careful about what this is an

The Guardrails

Training AI on books is fair use — piracy is the crime

Copyright law officially turned 50 years old with zero updates, and the AI industry's biggest unresolved legal question still hinges on a statute written in an era of photocopiers, not neural networks. TechCrunch's explainer this weekend crystallizes where the fight actually stands after a year of landmark rulings: courts keep telling AI companies that training on copyrighted books is legal, even as they hand authors record payouts. The result is a legal landscape that rewards one thing and puni

The Arena

Deep Dive — Nvidia's $6 billion plan to beat China at open weights

For a company that historically made its money selling the shovels, Nvidia just bought a pickaxe. The chip giant is taking the extraordinary step of licensing the model-building software of the Paris startup Poolside and hiring 109 of the people who used it — a deal people familiar with the matter say Nvidia intends to turn into one of the world's most powerful open-weight AI models, built specifically to answer Chinese heavyweights like DeepSeek and Kimi K3. The number attached to that ambition

The Stack

MCP's hardest problem: proving which agent asked for what

The Model Context Protocol — the open standard that lets AI assistants plug into tools and data sources, the plumbing under nearly every agent you've met — published a fresh roadmap on Friday. Across its five priority areas, one stands out less for its novelty than for how long it has been quietly dodged: agent identity. Today MCP can tell you which person handed a tool to a model. It cannot yet tell you which agent is on the other end, what it is allowed to touch, or whether the model carrying

The Frontier

NVIDIA's AVO agent clears every ARC-AGI-3 public level

NVIDIA says its long-horizon agent system, AVO, completed all 183 levels across the 25 environments of the ARC-AGI-3 public set with a perfect 100.00 score. The result is real and independently checkable against the official scorecard — but the more interesting claim is buried underneath it: the system lifted Claude Opus 5 from a roughly 30% model baseline to 100% by changing the harness, not the model. What actually happened ARC-AGI-3 is the latest installment of François Chollet's attempt t

See all