Sunday, November 1, 2026

TL;DR

Alphabet lost roughly $500 billion in value after tying Search's slowdown to AI Mode's weaker ad monetization, while OpenAI's funding round fell to $11 billion of its original $30 billion target. Wiz traced nine breached AI dev platforms feeding a stolen-credential resale market, and chatbots got voters' congressional districts wrong before Tuesday's election.

Act on this

  • Before Tuesday's vote, don't trust an AI chatbot's voter guide: the Washington Post found six chatbots named the wrong congressional district up to 48% of the time. Washington Post, AI chatbot voter guides #
  • If you run on Fireworks, Zeabur, Lightning AI, Railway, Warp or OpenCode, rotate AI provider credentials: Wiz traced nine breaches this month feeding a stolen-key resale market. Wiz, Black October for AI developer platforms #

Signals

AI coding fluency is becoming table stakes, not a differentiator - hiring is shifting toward judgment and initiative instead.  #

Anthropic engineering manager Felix Rieseberg wrote Oct 27 his team gets 'zero signal' from coding interviews now that Claude does that work, with no replacement settled yet. Gergely Orosz's Oct 30 newsletter calls it a leading indicator other AI-native engineering orgs are likely to follow. Felix Rieseberg, Hunger beats tenure · Gergely Orosz, The Pulse: Anthropic kills the coding interview

Anthropic's own safety sacrifices are real but narrow, concentrated where its leaders personally care most.  #

Zvi Mowshowitz's Oct 27 essay credits Anthropic's biosecurity safeguards and state-law advocacy, then names the gap: its RSP v3 reportedly drops unilateral pause commitments, its leaked draft bill would preempt stronger state rules, and Dario Amodei now calls 'race to the top' the strategy he once rejected in 2022. Zvi Mowshowitz, Anthropic Versus Anthropic

Benchmark aggregators, not labs or users, are quietly deciding which models win a trillion-dollar market. #

Nathan Lambert's Oct 28 post on Artificial Analysis v5 argues every weighting choice in an aggregate index picks winners and losers, not just measures them - the new version weights knowledge work, coding, science and general ability equally at 25% each, a choice he calls defensible but arbitrary. Nathan Lambert, Artificial Analysis v5 and the benchmarker's dilemma

News

ACTChatbots get your voting district wrong up to 48% of the time, Washington Post finds #

Testing six chatbots on 3,000+ addresses, the Post found they failed to name the correct congressional district 48% of the time and misnamed it 22% when they tried. Microsoft Copilot cited a largely AI-generated voter site in 75% of tests; Amazon paused Alexa's voter guides. Washington Post, AI chatbot voter guides

ACTWiz traces 'Black October': nine AI dev platforms breached, feeding a stolen-credential resale market  #

Fireworks, Railway, Warp, Zeabur, Lightning AI, OpenCode and Northflank disclosed breaches this month. Stolen AI credentials now move through reseller sites within days; two cases Wiz traced each ran up over $100,000 in model calls before detection. Wiz, Black October for AI developer platforms

WATCHOpenAI's funding round falls to $11 billion, a third of its original $30 billion target #

The Wall Street Journal reports OpenAI secured only about $11 billion in commitments as big investors cooled; the Financial Times says OpenAI cut its target to roughly $20 billion. The round still values the company near $1.25 trillion, short of the $1.4 trillion it first sought. Investing.com, OpenAI raises just $11bn of $30bn target (WSJ) · TNW, OpenAI funding round target cut to $20bn (FT)

WATCHAlphabet loses roughly $500 billion in value as Google ties AI Mode to weaker ad growth #

Chief business officer Philipp Schindler said Oct 29 that AI Mode's query growth hurt search-ad growth through lower per-query monetization. Search revenue still grew 12% to $63.35 billion, down from about 19% earlier this year; shares fell roughly 12-13% the next day. TechSpot, Alphabet shares plunge over 10% as AI search struggles to make money · WBUR, Alphabet stock tanks on disappointing AI-search revenue

WATCHA ChatGPT bug chain could have let one shared link hijack any signed-in account  #

Parameter Labs chained a MathLive postMessage flaw, an unsandboxed iframe and a YouTube CSP bypass to run code as the victim inside ChatGPT's HTML previews. OpenAI patched it Oct 26 and paid a $5,000 bounty, which Simon Willison called low for the severity. Parameter Labs, We found a gap at OpenAI · Simon Willison, linking to Parameter Labs

The long view

A year ago, Google pitched AI Mode as a usage-growth story - hundreds of millions of queries, no public data tying it to ad economics. On Oct 29, chief business officer Philipp Schindler said directly that AI Mode's query growth slowed search-ad growth because of its lower monetization, and Alphabet lost roughly $500 billion in value in a day. If this holds, each future earnings call turns AI Mode from a growth metric into a line item investors price against the ad dollars it displaces.

Also noted