Monday, November 9, 2026

TL;DR

Nvidia's board is investigating CEO Jensen Huang after he admitted an affair with a Megaspeed executive US intelligence says was passing information to Beijing. Anthropic committed $3.4 billion to bypass the grid with its own gas generators, and Goldman Sachs pulled Claude Code access from most of its traders and bankers.

Signals

The bottleneck in AI-assisted coding is moving from writing code to reviewing it, and practitioners are building habits around that specifically rather than waiting for a better model.  #

Neil Kakkar, in Gergely Orosz's Nov 3 newsletter, ships 200+ PRs a month at the AI-native startup Tano - 11x his rate at PostHog - with a 6x lower revert rate, mostly merged with no human review and gated instead by an automated Claude Code agent. He separates the gain from the cause: a pre-AI engineer with the same tools would likely see only 1-2x, since review queues are the real constraint. Simon Willison's Nov 9 post independently collapses agent-written PR text behind a details/summary tag, keeping his own short write-up as the visible, human-written part. Gergely Orosz (guest essay by Neil Kakkar), How software engineering is changing: a tale of two eras · Simon Willison, Hide your slop PR descriptions behind details/summary

Beyond Willison and Orosz, it was a quiet week - Zvi Mowshowitz and Jack Clark both published, but nothing else cleared the bar for a second signal.

News

ACTARTEX's exposed instance count nearly doubles to 1,465  #

Censys found 1,465 internet-exposed consoles of the ARTEX pentesting framework on Nov 5, up from the 800+ it reported a week earlier; 32 expose attacker post-exploitation kits directly. Take any public-facing instance offline now. Infosecurity Magazine, on Censys's findings

WATCHNvidia's board investigates Jensen Huang over an alleged Chinese intelligence relationship #

Huang told staff Nov 8 he'd had an affair with Megaspeed executive Alice Huang, whom US intelligence says was passing information to Beijing. The board says it has no evidence yet that secrets were compromised; Huang stays CEO. CNBC, Nvidia investigates CEO Jensen Huang's relationship with alleged Chinese spy

WATCHAnthropic commits $3.4 billion to gas generators, skipping the grid entirely  #

The Financial Times reports Anthropic agreed in late September to pay OnePWR Solutions $3.4 billion for 1.7 gigawatts of heavy-duty generators to power its data centers. Reuters could not independently verify the report; neither company commented. CNBC, on the Financial Times' reporting

WATCHGoldman Sachs pulls Claude Code and GitHub Copilot from most bankers and traders #

Bloomberg reports Goldman restricted the tools in October to roughly 12,000 engineers, citing a productivity and risk review - a reversal from CIO Marco Argenti's July praise of trader 'vibe coding.' Strats keep access. FStech, on Bloomberg's reporting

WATCHPost-midterm Democrats line up concrete AI oversight priorities #

Reps. Castor and Casar and Sens. Warner and Wiener are naming AI safety and data-center oversight as January priorities for the new Congress; a YouGov poll found 81% of voters back candidates who favor AI safeguards. Nextgov, What the midterm results mean for federal tech policy

The long view

A year ago, scrutiny of Nvidia was a supply story: how many chips it could ship, and to whom. On Nov 8, Jensen Huang told staff he'd had an affair with a Megaspeed executive US intelligence concluded was passing information to Beijing, and Nvidia's board opened an independent investigation with outside counsel - while saying it has no evidence yet that trade secrets or national security were compromised. If this holds, the company gating the AI industry's chip supply now faces a leadership-stability question too, with a congressional briefing already set for Nov 18.