Scoop
AI tool news · rumor vs. reality

The AI Wire ●

Rumors tracked. Announcements verified. Updated daily.

Lead story
ANNOUNCED

Anthropic prices its IPO at $83 a share, above its range, in an upsized offering — trading starts Thursday as ANT

Anthropic priced its initial public offering at $83 a share on Wednesday, November 18, above the $70 to $78 range it set on November 10, according to Reuters, Bloomberg and CNBC reports citing people familiar with the matter. The offering was also enlarged, to 980 million shares from the planned 825 million, which at $83 comes to about $81.3 billion in gross proceeds before underwriting fees.

At that price Anthropic is valued at roughly $1.5 trillion on a fully diluted basis, up from the $965 billion attached to its last private round in May. The SEC declared the registration effective on November 18, the offering is expected to close on Friday, November 20, and the final prospectus is still to be filed; Anthropic declined to comment to Reuters.

Shares are expected to start trading on the Nasdaq on Thursday, November 19, under the symbol ANT, with Goldman Sachs and Morgan Stanley leading a 22-bank underwriting syndicate. Alongside the public offering, seven technology companies — AMD, Broadcom, Google, Micron, Microsoft, Nvidia and TSMC — have agreed to buy a combined $26 billion of non-voting shares in a separate private placement at the IPO price.

The debut makes Anthropic the first major frontier AI lab to go public, and the second mega-IPO focused on AI this year after SpaceX's in June. It caps the roadshow the Wire has tracked since November 10 — see that entry for the original $70 to $78 terms.

Update, November 19: ANT opened for trading at $111.50 a share at 11:49 a.m. ET, 34% above the $83 IPO price, after pre-open indications climbed from about $105 to $111. The opening print put Anthropic's fully diluted valuation near $2 trillion, up from about $1.5 trillion at the IPO price, and co-founders Dario and Daniela Amodei rang the Nasdaq opening bell. The offering is expected to close on Friday, November 20; first-day closing figures were not yet final at sweep time.

Close update, November 19: ANT closed its first day at $97.17, up 17.07% from the $83 IPO price and about 13% below the $111.50 opening print. The shares peaked at $115.10 just before noon ET and slipped as low as $92.72, with more than 846 million shares changing hands — the most active name on the Nasdaq — on a day when the main US indexes finished more than 2% lower. CNN put Anthropic's fully diluted value at about $1.76 trillion at the close, down from roughly $2 trillion at the open. The offering is expected to close on Friday, November 20.

2026-11-19

RELEASED

OpenAI starts testing ChatGPT dots with Plus users and brings dots to all Business Standard accounts

OpenAI has begun testing ChatGPT dots with a small number of ChatGPT Plus subscribers, expanding the always-on proactive assistant beyond its launch tiers of Pro, Business Premium and Enterprise. Thibault "Tibo" Sottiaux, who leads engineering for Codex and ChatGPT, announced the test on November 19 and said he is seeing more positive feedback on how proactive dots have become after many improvements.

At the same time, OpenAI is bringing dots to all ChatGPT Business Standard users. Dots, announced in September and powered by GPT-6 Astra with GPT-6.1 Terra added earlier this month, have their own cloud computer, work across connected apps through plugins, and keep making progress on projects even when the user is not actively interacting with them.

It is not yet clear whether Plus and Business Standard users will get the same dots usage allowances as the higher tiers, or whether additional restrictions will apply. On the launch tiers, dots conversations do not count toward regular ChatGPT usage limits for Pro and Business Premium subscribers, though complex tasks sit under separate allowances and work delegated to ChatGPT Work or Codex counts toward those products' allowances.

RUMOR Rumor still unresolved

Rumor: Meta will price its Muse Charm gadget at $300, with a year of free Verizon 5G — Meta has announced no price

Meta plans to price Muse Charm, its pocket-sized AI device, at $300 and offer buyers up to 12 months of free 5G data from Verizon, The Information reported on November 19, citing a person familiar with the matter. Reuters and TechCrunch carried the report the same day; Meta has not announced any price.

Charm carries Meta's Muse agent into a physical device users can keep in a pocket or wear around the neck, talking to it, asking questions and giving it commands. Mark Zuckerberg unveiled the device at Meta's Connect event in September and said then that Meta expected to ship it in time for the holiday shopping season in December, with the broader Muse experience including real-time voice and avatars — but gave no price.

Asked for comment, a Meta spokesperson told TechCrunch: "We're charmed by the interest in our AI devices, and we'll have more to share soon" — a non-denial that confirms nothing about the $300 figure or the Verizon offer. Reporting around the story also says Meta's separate smartwatch launch has moved to 2027 as the company focuses on bringing Charm to market; treat the price, the data plan and the watch timing as unconfirmed until Meta announces them.

RELEASED

Google brings computer use to the Gemini app for Mac: Gemini can work across your apps and the web in the background

Google announced on November 19 that computer use is rolling out in the Gemini app for Mac. Gemini can now navigate Mac apps and websites in the background to finish multi-step tasks across local files, desktop apps and the web, working inside an isolated background window with a live preview right in the chat.

The rollout starts in English in the United States. TechCrunch reports it requires the Gemini app for Mac and a Google AI Plus, Pro or Ultra subscription, and Google did not say when it will expand to other markets or platforms.

It is Google's answer to the computer-use agents rivals have been shipping for desktop: instead of only answering in a chat window, Gemini can now act on the machine itself, with the live preview as the way to watch what it is doing.

2026-11-18

RUMOR Rumor still unresolved

Rumor: the Trump administration is weighing a 'name and shame' rule forcing US firms to disclose Chinese AI use — Reuters

Reuters reported on November 18, citing three people familiar with the talks, that the Trump administration is considering requiring US companies to publicly disclose their use of Chinese AI models, posting the information on their own websites. The Treasury Department and the White House Office of Science and Technology Policy have discussed the idea.

The point, per Reuters, is discouragement without a ban: a 'name and shame' approach meant to steer companies away from Chinese models without imposing restrictions that could anger industry or Beijing. Airbnb is the named example — it uses Alibaba's Qwen to power its AI customer service agent — and House China Select Committee chairman John Moolenaar welcomed the report.

Nothing has been announced or enacted. Reuters said it was not clear whether the idea is gaining traction inside the administration or would move into formal rulemaking, and the White House, Treasury and Alibaba did not respond to requests for comment. This stays a rumor until an official proposal appears.

ANNOUNCED

Palo Alto Networks agrees to buy Applied Compute for $4.5B in stock, betting on secure open-weight AI for the enterprise

Palo Alto Networks announced on November 18 that it has signed a definitive agreement to acquire Applied Compute, a platform for post-training and inference of open-weight AI models, in an all-stock deal valued at $4.5 billion. Applied Compute was founded in 2025 by former OpenAI researchers and raised $350 million earlier this month at a $3.25 billion valuation.

The pitch is secure enterprise adoption of open-weight models: Palo Alto says the platform will help companies adapt and run open models alongside closed frontier models, with security built in. Co-founder and CEO Yash Patil will become Palo Alto's co-chief AI officer when the deal closes, and the research team will join with him.

The deal needs regulatory clearance and is expected to close in Palo Alto's fiscal third quarter of 2027. It was announced alongside fiscal first-quarter results showing revenue up 39% to $3.44 billion, which Palo Alto also used to raise its full-year outlook.

ANNOUNCED

OpenAI: its coding agents now finish 46% of researchers' 2–4-hour tasks unaided, up from 23% in March — and the median researcher runs $2,400 a day in agent usage

OpenAI published “Research acceleration: The jagged frontier” on November 18, its second report on how coding agents are changing research inside the company. Its headline measure is zero-intervention success: whether an agent finished a delegated task with no person stepping in.

For tasks estimated at two to four hours of human work, that success rate rose from 23% in March to 46% in October; for four-to-eight-hour tasks it rose from 13% to 35%. The gains are uneven — agents succeeded an estimated 79% of the time monitoring training runs but only 25% of the time writing inference and serving code — and OpenAI warns the weaker categories could become bottlenecks to overall research progress.

Usage has surged alongside: the median researcher's daily agent usage was $2,400 at API list prices in October, the top tenth hit $23,000, and logged experiments per active researcher ran at 1.68 times the 2025 baseline. OpenAI stresses these are internal estimates, that researchers changed which tasks they delegated as agents improved, and that the report tracks indicators relevant to recursive self-improvement without reading as a commitment to pursue it fully.

ANNOUNCED

OpenAI signs Groupe Figaro: Le Figaro journalism will surface in ChatGPT, with ChatGPT Enterprise to follow inside the newsroom

OpenAI and Groupe Figaro announced a strategic partnership on November 18 that will make Le Figaro's current reporting and archive discoverable in ChatGPT as summaries and excerpts linking back to the original stories. The coverage spans news, business, culture and lifestyle across Le Figaro and associated titles such as Madame Figaro and Gala, produced by more than 600 journalists.

Le Figaro keeps full editorial independence and control over its journalism under the deal. Groupe Figaro also plans to introduce ChatGPT Enterprise across various teams to support day-to-day work, guided by its AI charter and editorial standards.

Figaro is the third French media group to sign with OpenAI, after Le Monde in 2024 and LVMH's media division on November 16. Neither side disclosed financial terms, a launch date for the ChatGPT integration, or which plans and countries receive it first; Le Figaro says ChatGPT draws 27.4 million monthly unique visitors in France.

ANNOUNCED

“Essentially unusable”: GPT-6.2 Astra and Sol draw scathing early reviews over autonomy, even as benchmarks tie Fable 5.2

OpenAI's GPT-6.2 Astra and Sol, released November 17, have drawn unusually harsh reviews from the early testers OpenAI itself seeded. Investor Matt Shumer called the models “essentially unusable” and initially skipped his customary long review; developer Theo Browne called Astra “the worst model OpenAI has ever released.”

The complaints are about judgment rather than raw capability: testers say Astra constantly stops to ask permission on innocuous steps, stalls or stops early on long end-to-end builds, and produces hard-to-vary output with a noticeable house style. Browne described a TypeScript-compiler rewrite where Astra worked seven days without landing an improvement, then four more days without a change when redirected. Wharton professor Ethan Mollick had a mixed experience — a working Blender and Unity space-elevator simulation, but a model he found “very uneven & weird” and harder to steer than rivals.

The benchmarks tell a different story: Astra scores 62 on the Artificial Analysis Intelligence Index, second only to Anthropic's Fable 5.2 at 65, and tied Fable 5.2 atop Epoch's index. OpenAI responded that the models changed substantially in the final days before launch, which may limit how far the early tests apply. The timing is awkward — the reviews landed the same week Anthropic priced its record IPO, and more than one tester joked that Astra was an advertisement for it.

RELEASED

ChatGPT gets Live Dictation: streaming transcription in the app, with the 10-minute limit gone

OpenAI updated dictation in the ChatGPT app on November 18 with Live Dictation, which streams your speech into text as you talk instead of waiting for you to finish and then transcribing the whole recording. OpenAI announced the change on X, saying dictation in ChatGPT is getting faster and more reliable.

The update also removes the old 10-minute dictation limit, according to Atty Eleti, who works on voice features at OpenAI. For anyone who talks to ChatGPT at length — drafting, brainstorming, or dictating notes — the limit removal is the practical headline: long voice messages no longer need to be split or rushed.

RELEASED

OpenAI launches ChatGPT for Accounting: an Accounting plugin plus QuickBooks, Xero, NetSuite and more, a day after ChatGPT for Sales

OpenAI launched ChatGPT for Accounting on November 18, pairing a first-party Accounting plugin with partner plugins for Intuit QuickBooks, Xero, Oracle NetSuite, BILL, FreshBooks, Rillet and Campfire. It is the second job-specific ChatGPT package in two days, following ChatGPT for Sales.

The plugin prepares draft journal entries and invoices from emails and receipts, suggests account classifications, reconciles bank transactions and drafts month-end reports and audit preparation. Anything unclear is flagged for a human, and every final entry waits for human approval before it is posted to the books.

Availability depends on plan and workspace settings: ChatGPT Business members can install the plugin without prior admin approval, while Enterprise workspaces need an administrator to approve it. Xero expanded its own integration the same week, bringing month-end close and client portfolio reviews into ChatGPT for US users first. OpenAI has not announced separate pricing for the package, and says Business and Enterprise data is not used to train its models by default.

Update, November 19: FreshBooks, one of the eight launch partners, formally announced its ChatGPT for Accounting connection, stressing that it is read-only — ChatGPT can retrieve live FreshBooks data to answer questions but cannot change anything in a FreshBooks account. Xero also expanded its own integration on November 19, initially in the United States only, covering month-end close review, client portfolio review and quarterly review preparation.

ANNOUNCED

OpenAI says its models found up to 14% gains in AMD's flagship GPU designs — and some rewrites are already headed for manufacturing

In a research update published November 18, OpenAI says its models rewrote SystemVerilog RTL — the code describing a chip's circuits — on AMD's flagship GPU intellectual property and found improvements of up to 14% across power, performance and area, measured against baselines AMD's engineers had already heavily optimized. OpenAI characterizes the work as roughly a year of progress achieved in less than five months, with Synopsys tools verifying that each rewrite preserved the block's behavior.

AMD has already included some of the model-written changes in a design sent for manufacturing, and added others to production design files for future versions. On blocks of its next-generation flagship CPU, AMD separately verified single-digit percentage gains from model-proposed edits.

The same approach was used on Jalapeño, OpenAI's first custom inference chip: model-proposed changes cut standard-cell area by about 13% in the vector unit and about 10% in the matrix unit, with the gains supported by tests on real silicon, and AI assistance helped move the chip from initial design to tapeout in nine months. OpenAI also reports 9,881 scored evaluations on two open-source RISC-V cores, PicoRV32 and Ibex, where GPT-6.2 Astra performed best — while noting the caveat that all five models tested scored better when designing from a written spec than when improving existing code, partly because they could leave out features the originals had.

The update is a research-and-engineering account, not a product launch: there is no generally available chip-design assistant. It does extend the GPT-Synopsys work the Wire covered in October — OpenAI is training GPT-Synopsys to drive Synopsys' design tools the way expert chip designers do, under a multi-year agreement.

RELEASED

Meta's Muse gets Try On: the shopping agent will show you how clothes look on you before you buy

Meta's Muse agent added a Try On feature on November 18, announced by Chief AI Officer Alexandr Wang and the Muse account. While you shop for clothes in Muse, the agent offers to show you a preview of how the item would look on you, generated from the exact product listing it found rather than a generic outfit.

David Singleton of Meta Superintelligence Labs walked through it with a holiday sweater he picked this week. Muse finds products through its retailer connections or its own browser when a shop has no direct connection, a separate system called Sentinel reviews its actions, and every purchase still needs the shopper's approval, whatever the price.

The preview is a look, not a fit guarantee: Meta has not spelled out which countries or stores support Try On, or what photo a shopper needs to supply. It lands a day after Muse's Loblaw grocery launch in Canada, in the shopping use case Meta has singled out for Muse's early growth — the company said on its October earnings call that Muse had passed 10 million users.

RELEASED

Meta's Muse lands in Canada for shopping: Loblaw is the first Canadian retailer on the agent, and Muse Voice rolls out too

Loblaw Companies announced on November 18 that it is making its product catalogue available through Muse, Meta's AI-powered personal agent — the first Canadian retailer to enable the capability. As a partner connector, Canadians can go from store selection to checkout at participating Loblaw banners using Muse.

Shoppers can use Muse to plan meals, build shopping baskets and check out at participating banners without leaving the agent. Alongside the Loblaw partnership, Meta announced that Muse Voice is rolling out in Canada on mobile devices, the web and Mac.

Muse runs on Muse Secure VM, a cloud-based system designed to keep passwords and payment details outside the agent's view. The announcement does not identify which Loblaw banners are participating, and it gives no detailed breakdown of delivery fees or how checkout works in each case.

RELEASED

Infor brings Claude into manufacturing operations: Industry AI for Claude can act on purchase orders and production, with human approval

Infor and Anthropic have partnered to put Claude to work on the factory floor through Infor Industry AI for Claude, a new plugin and connector that lets manufacturers ask plain-language questions and carry out approved actions without leaving the conversation. The connector runs on the open Model Context Protocol.

Claude can access operational data in Infor, review recommendations and — once a user reviews and approves the action — update a purchase order, reallocate stock or shift a production line. Claude is now Infor's preferred AI assistant for its products, and Claude models are the preferred models behind Infor's Industry AI line.

The integration is available now to select Infor Velocity Suite customers who hold their own Claude licence. It is a select-customer release, not a general rollout to all Infor users.