In partnership with
A Single Typo Shouldn’t Cost You a Sale
A shopper types “running shos” and gets nothing back. Another types “red waterprof trail runners under $150” and hits the same wall. Both knew what they wanted.
Traditional site search relies on exact keyword matching, leaving shoppers stranded at “0 results found” screens.
Gladly understands what shoppers mean, typos and all. Instead of a dead end, they get products that fit what they asked for and a conversation that keeps them shopping.
Anthropic’s amended S-1 provides more detail on the company’s risks and opportunities.
Anthropic’s amended IPO filing, reported by Bloomberg, puts third-quarter revenue at nearly $17 billion, up from $1.4 billion a year earlier. The same document discloses more than $895 billion in future spending commitments, almost all of it for computing power, with about $70 billion due in the next twelve months. The quarter’s $23.3 billion net loss is mostly an accounting artifact of the company’s rising valuation, but the operating line tells a real story: Anthropic turned its first operating profit in Q2, then slipped back into the red in Q3 on heavy short-term compute costs. The listing, expected before Thanksgiving under the ticker ANT, could rival SpaceX’s record $86.2 billion raise, with investors reportedly eyeing a valuation of $1.4 trillion or more. Anthropic reported it generated $33.24 billion in sales for the nine months ending September 30th, 2026.
The detail to watch: price cuts on newer models and customer discounts shaved about $600 million off the quarter’s revenue growth, as OpenAI and cheaper open-weight rivals push prices down. That is the central wager of this IPO in miniature: whether all that compute turns into revenue faster than competition erodes what each token is worth.
Meanwhile, The Financial Times reported that credit agencies are expected to give Anthropic a high "junk" rating following its IPO, as relayed by TechCrunch. In plain terms, that is the top rung of below-investment-grade, which typically means borrowing costs run higher than for blue-chip issuers. For a lab spending at frontier scale, pricier debt is a real constraint. If your workflows lean on a single provider, the lab's financing is now part of your vendor risk assessment. This is a key issue that other frontier labs are watching, as they weigh their own financing plans. The “price of money” is one of main levers (if not the biggest lever) weighing on the future of the AI trade now.
Anthropic's S-1 describes two futures: miracles or existential risk.
In its S-1 filing, Anthropic frames its stock as a bet on either an "age of miracles" or "catastrophic or existential risks to humanity," according to The Verge. Risk-factor sections are usually written in dry legal boilerplate, so seeing a frontier lab describe its own stakes this starkly to public investors is unusual. It is also a useful primary source for anyone who wants to know how these companies talk about danger when securities law is listening. It’s useful to read it as both a disclosure and a piece of positioning.
OpenAI reportedly lowers its funding target.
TechCrunch reports that OpenAI has lowered the target for its new funding round, because the raise is reportedly not living up to expectations. The company is the sector's most visible name, so its fundraising is read as a proxy for investor mood across AI. A trimmed target does not mean trouble on its own, since the reporting is thin on specifics. It does suggest that the current appetite for very large checks to AI firms has tempered, even for the category leader. Clearly, all eyes are on the Anthropic IPO, currently expected in November.
an official Claude extension in Cursor was backdoored for nearly 12 hours.
Ars Technica reports that a Claude extension from Anthropic, distributed through Cursor's marketplace, carried a backdoor for nearly 12 hours, putting as many as 5,000 developers' machines at risk. At the time of reporting, Anthropic had not yet warned customers. Extensions run with broad access to code, credentials, and file systems, so a compromised one is a direct line into a developer's environment. Any team using AI coding tools should inventory its installed extensions, check who publishes them, and decide who gets to approve updates.
OpenAI, Google, Microsoft, and Anthropic back a shared opt-out signal.
The Verge reports that OpenAI, Google, Microsoft, and Anthropic say they will adopt a common policy-token signal, letting publishers state their AI training preferences once and have them read by multiple companies. Today, opting out often means a patchwork of crawler-specific rules that each company interprets on its own. Currently, a site that wants to opt out of training has to block each company’s bot by name. A single signal would help anyone who publishes online, from solo writers to large media archives. For those unfamiliar: every website can have a .txt file called robots.txt, which tells automated crawlers what they can and cannot access on their site. The policy token is a specific line in that file. Interestingly: the token separates training from search; it allows a website to tell all AI content scrapers that it wishes to “opt out” of training, distinguishing it from search indexing and archiving. The open question is enforcement: a standard only matters if crawlers honor it and someone can tell when they don't. OpenAI, Microsoft, and Google will begin adopting the common policy-token signal over the coming months; Anthropic is said to be rolling out the signal immediately.
Nvidia weighs cosigning Nscale's West Virginia data center.
Nvidia is considering acting as cosigner on financing for an Nscale AI data center in West Virginia, TechCrunch reports. The chipmaker that sells the hardware would, in effect, lend its credit to the buyer. That is a sign of how investors have grown nervous about AI infrastructure spending, and of how much support the buildout needs to keep going. When the supplier backs the customer's loan, demand and financing become entangled in ways that could be worrying for students of markets history. I cover Nvidia’s place in the AI financing universe here in my podcast episode, which you can watch here: https://www.youtube.com/watch?v=onSdiQDkAs0&t=891s
Marvell Technology agrees to acquire Lumentum Holdings in all-stock deal.
Marvell Technology has agreed to buy Lumentum Holdings in an all-stock deal valuing the optics maker at $115 billion, Bloomberg reports. Leaving aside the SpaceX-xAI merger, that makes it the largest tech acquisition ever on an equity basis. Lumentum makes the lasers that let data centers move enormous volumes of data using light instead of copper wiring, which has quietly become one of the tightest bottlenecks in building AI infrastructure. Marvell, which already designs custom AI chips for Amazon, Microsoft, and Google, is a Lumentum customer and bought photonics startup Celestial AI earlier this year; Nvidia took a $2 billion stake in Lumentum in March.
The market’s first reaction was a wince: both stocks fell on the news, Lumentum by as much as 12%. The deal isn’t expected to close until the first half of 2028, a long runway in an industry where today’s bottleneck can be tomorrow’s commodity. The bet is that whoever controls how data moves between chips will matter as much as whoever makes the chips.
judge lets NAACP's Clean Air Act suit against xAI proceed.
A judge has allowed the NAACP's Clean Air Act lawsuit over pollution from xAI's data center to move forward, TechCrunch reports. xAI is now owned by SpaceX. The ruling does not decide the merits, but it keeps the case alive and puts the facts under legal scrutiny. Expect courts to become a recurring venue where the environmental costs of AI infrastructure are tested, one facility at a time.
Salesforce reportedly pays $375M for defense startup Rune.
Salesforce is reportedly buying Rune, an AI logistics startup, for $375 million, TechCrunch reports. Rune would bolster Missionforce, Salesforce's defense platform, and the deal extends a run of acquisitions by the company. Enterprise software vendors that built their businesses on sales pipelines are now buying their way into military logistics. Watch for more of this as AI makes planning and supply-chain software a more attractive defense product.
Apple Intelligence reaches China with ios 27.3.
Apple will bring several Apple Intelligence features to China through an iOS 27.3 beta on November 4th, with a full release in December, The Verge reports. The new Siri AI is not part of the package. It is a major market expansion for one of the largest consumer AI products, though a partial one. How much of the feature set arrives later is the detail to track.
worth your time.
Simon Willison's writing on the "lethal trifecta" for AI agents is a clear framework for the first story above. His argument is that risk spikes when a system combines access to private data, exposure to untrusted content, and the ability to communicate externally. It gives you a quick test for any assistant or extension your team adopts: which of the three legs does it have, and can you remove one?
The human mind is the original generative machine.