A stock chart falling sharply beside a rising stack of server racks

AI News, Oct 30: Alphabet's Worst Day Ever

The Big Story: Alphabet’s worst day ever, and Amazon’s answer

Alphabet lost $482 billion in market value on Thursday, its worst day ever and, by Yahoo Finance’s count, the second-largest one-day loss for any US-listed company after Nvidia’s $593 billion. The stock fell 12%, its steepest drop since the 2004 IPO. The trigger was not a revenue miss but a line further down: free cash flow swung to a loss of about $9 billion as AI spending climbed. Wednesday’s earnings disclosed that number. Thursday was the market’s verdict on it.

Amazon reported hours later and made the trade harder. Adjusted earnings came in at $1.89 a share against $1.96 expected and revenue at $201.94 billion against a $202.31 billion consensus. Capital expenditure hit $77.3 billion against a $58.7 billion estimate, a third more than analysts had modeled, and free cash flow landed at negative $42.4 billion. Andy Jassy held 2026 cash capex at roughly $220 billion, said Amazon expects to spend “meaningfully more in 2027”, and argued AWS, up 39% to $45.98 billion, can “be a trillion-dollar revenue business in the next five to 10 years.” Shares fell more than 7% in late trading.

What makes the pairing uncomfortable is a Gartner finding reported the same afternoon: 38% of organizations are over budget on AI and 85% plan to spend more next year, while just 21% of the time employees saved got redirected into more innovative work. “Ultimately, organizations will not generate cost savings from AI at this point,” Gartner analyst CV Viverito said.

Three cyber stories in one day, none of them needing a frontier lab’s permission

Epoch AI pointed GLM 5.3, the open-weight model Z.ai released in August, at Firefox’s patch gap, the interval in which a fix is public in source but not yet installed on anyone’s machine. Running autonomously, it built working attacks on all six Firefox releases between September 17 and October 26 at roughly $1,000 each: a webpage that, simply when visited, read a local file and sent it to the attacker. Firefox has over 200 million users, and Epoch notes it could lift their stored passwords. No provider, Z.ai included, refused any of the work.

CrowdStrike described the operational version. A China-nexus group it calls VAPOR PANDA ran an agent framework named ARTEX, primarily on DeepSeek’s model, against more than 200 organizations in at least 14 countries over an 11-day window beginning October 17, with planning agents burning on the order of 1.4 billion input tokens in a single intrusion. Post-exploitation was confirmed at nine of the targets, a figure worth holding against the headlines.

Mistral spent the same day selling the defensive side of the same property. Exploiting a patch gap used to need a frontier lab’s permission. Not anymore.

Today’s Top Stories

Mistral shipped a trillion-parameter model, and Hacker News read the licence

Mistral released the weights for Mistral Large 4, nicknamed Le Chonk: a trillion-parameter mixture-of-experts model with 49 billion active parameters per token, trained in Mistral’s own European datacenters. It is the window’s top Hacker News story at 1,564 points, and the first thing that thread did was read the licence. The Mistral Ecosystem Open-Weight License is a modified MIT allowing commercial use, except that running a Model-as-a-Service business needs a separate agreement with Mistral. The thread’s verdict: weights-available, not open-weight. On Artificial Analysis’s Intelligence Index it scores 26 against 59 for Gemini 4 Carbon. The narrower claim is the one that holds: best open model for agentic coding, on 47% fewer tokens per task than GLM-5.3.

Canada bought a piece of Cohere

Cohere announced the initial close of a round targeting as much as $7 billion CAD at a valuation of nearly $36 billion, led by German retail group Schwarz Group, with the federal government taking a $1.6 billion stake through the Canada Development Investment Corporation, staged against milestones with jobs and revenue targets attached. Mark Carney announced it on day one of the Canada-EU summit. Nvidia is returning, and Cohere is in the final stages of merging with Germany’s Aleph Alpha. Sovereign AI has been a sales pitch for two years; this is the version where the sovereign takes equity.

A trojanized Claude Code extension sat on Cursor’s marketplace for 11 hours

Version 2.1.313 of the Claude Code extension, distributed through Cursor’s VS Code marketplace, collected host details and tried to add an attacker-controlled SSH key, enable local SSH and open a reverse tunnel, persisting via a macOS LaunchAgent or Windows scheduled task. Cursor says the build was live from about 2:27am to 1:35pm Pacific on October 29 and affected at most 5,000 users, and that enabling SSH needs admin rights the extension lacked on most machines, so the tunnel would have failed even though the key, the persistence and the host details still went out. Cursor took its whole marketplace offline, rotated keys and rebuilt it. Both Hacker News threads on the incident went to correcting headlines saying Anthropic had been hacked: Cursor states the malicious version was not developed by Anthropic and that it has no evidence the official distribution was affected.

Reddit is closing the door on logged-out visitors

Reddit told investors it is moving to a primarily logged-in experience, which keeps its content from AI scrapers and makes it licensable instead. The pressure is measurable: search referrals fell by more than a million per week on average in October, and Reddit warned daily actives could drop a low double-digit percentage next quarter. Monetization went the other way, revenue up 53% to $895 million. “I’ve come to the conclusion that this is the right decision for our platform and the right decision for our business,” CEO Steve Huffman wrote.

Quick Hits

Ready to automate your busywork?

Carly schedules, researches, and briefs you—so you can focus on what matters.

See what people say

"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.

Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.

On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."

Gus Ibrahim, Founder & Director, IHR