AI Daily Briefing
Every public feed is filtered for AI relevance, reports of the same story are grouped, and the day's top events by independent sources are picked.
2026-11-22
Content is auto-organized by AI from public feeds for reference only. Verify important facts at the original source.
Today's AI conversation centers on how the sector re-evaluates public-market valuations and business fundamentals, as model advances intersect with company conditions and regulation. Anthropic's market performance and the SEC scrutiny of Nvidia, alongside new model releases and policy shifts, have spread attention across multiple fronts, reflecting a more grounded view of the industry's current state.
2026-11-22 AI briefing
- NO.01Rising1134↑ 8%
Anthropic Reaches $1.6 Trillion Market Value in First Trading Week After IPO
Anthropic listed on Nasdaq on November 19 at $83 per share, rising 17.1% on debut, before shares fell 8.5% the next day to $88.91 amid a broad selloff in AI and chip stocks. Its IPO raised $93.5 billion, with total offering proceeds reaching $119.5 billion after the over-allotment option was exercised. First-week valuations are reported differently, at $1.6 trillion and $1.8 trillion, leaving the market value in a wide range.
华D
C
T
B
C
+15华尔街见闻, DEV Community · AI, plus 19 more21 sources·42 reports
- Anthropic Falls 15% on Day Two, Nearly Breaking Its IPO Price华尔街见闻 · 11/22, 09:25
- Anthropic's $81.3 Billion IPO Falls More Than 11% on Day TwoDEV Community · AI · 11/22, 05:56
- Anthropic Raises $93.5B in Record IPO as Tech Stocks Post Worst Week Since 2020cnBeta · 11/22, 00:00
- Anthropic ends first trading week valued at $1.6tn as VCs hold $25bn stakesThe Next Web · 11/21, 23:20
- Bloomberg: Anthropic's record IPO fails to lift AI market sentimentBloomberg Technology · 11/21, 22:50
- Anthropic's record IPO lands as tech posts worst week on Wall Street since 2020CNBC · 科技 · 11/21, 21:00
- NO.021004Flat
Anthropic Mathematician Says Claude Formalized the Generalized Riemann Hypothesis
Anthropic mathematician Levent Alpöge says he and Claude formalized the proof of the generalized Riemann hypothesis in Lean without extra assumptions, releasing the work in 14 parts while withholding the 250,000-plus-page document. Stanford professor Ravi Vakil separately published a generalized Riemann hypothesis proof by postdoc Alessandro Masullo, describing it as entirely human-made. Neither claim has undergone peer review or independent verification, and Terence Tao says checking will take time.
TT
T
S
D
L
+15The Guardian · AI, The Next Web, plus 19 more21 sources·27 reports
- Anthropic mathematician claims Claude helped prove the Riemann hypothesisThe Guardian · AI · 11/22, 00:39
- Anthropic mathematician says he and Claude proved the Riemann hypothesisThe Next Web · 11/21, 21:12
- Anthropic Says Claude Proved Riemann Hypothesis; Stanford Releases All-Human Rival Proof Hours LaterThe Decoder · 11/21, 19:19
- After Anthropic's Riemann Hypothesis Claim, a Human Mathematician Publishes EarlySolidot · 11/21, 19:09
- Anthropic Mathematician Claims Claude Proved the Riemann HypothesisDEV Community · AI · 11/21, 18:04
- Riemann Hypothesis Claimed Proven Twice, One Proof Claude-DrivenLatent Space · 11/21, 13:02
- NO.03Rising592↑ 19%
Thinking Machines Lab Open-Sources Inkling 2 Mini with 32GB Model Card Recommendation
Thinking Machines Lab open-sourced the multimodal Inkling 2 Mini on November 20 under Apache 2.0, a looped Transformer with 26.4B total parameters and roughly 20GB of weights designed for local use. The official demo ran on a 48GB MacBook Pro, while the model card recommends Apple silicon with 32GB or more and 32GB of VRAM for other GPUs. Sources disagree on active parameters: Artificial Analysis reports about 5.4B, while TestingCatalog and 智东西 cite 3.4B.
掘T
M
T
D
V
+4掘金, The Next Web, plus 8 more10 sources·18 reports
- Inkling 2 Mini tested: 26.4B model loops middle layers 3x in 20GB laptop memory掘金 · 11/22, 08:26
- Thinking Machines releases Inkling 2 Mini, an open-weights model that runs on a laptopThe Next Web · 11/21, 23:49
- Thinking Machines Releases Inkling 2 Mini: a 26B Looped MoE That Runs on a 32GB MacMarkTechPost · 11/21, 20:13
- Thinking Machines releases Inkling 2 Mini, a looped model that runs on a laptopThe Decoder · 11/21, 18:33
- Thinking Machines Open-Sources Inkling 2 Mini, a 26.4B Looped MoEDEV Community · AI · 11/21, 18:03
- Thinking Machines Lab releases open-weights Inkling 2 Mini, model card suggests 32GB of memoryV2EX · 11/21, 15:56
- NO.04Rising486↑ 23%
SEC probes Nvidia accounting practices as shares post seven-week slide
The SEC is reviewing Nvidia's accounting practices, with one focus on the company's decision to remove Singapore from its geographic revenue breakdown, according to people familiar with the matter. The DOJ is also examining Nvidia's $12.9B Hugging Face acquisition and broader antitrust issues. Shares fell over 14% on Friday, marking a seventh straight trading day and week of declines, with weekly losses reaching 23% and market value shrinking by nearly $900B, stoking concerns about the sustainability of the AI boom.
IN
C
3
D
+4IT之家, NPR · 科技, plus 7 more9 sources·11 reports
- SEC probes Nvidia over potential accounting issues as shares halve from October peakIT之家 · 11/22, 07:07
- AI Stocks Took a Sharp Hit This Week, Raising AlarmsNPR · 科技 · 11/21, 18:00
- Nvidia Hit by SEC Accounting Probe as Market Value Drops Nearly $900B This WeekcnBeta · 11/21, 17:58
- Nvidia posts longest weekly losing streak since 2008, shedding about one Amazon in valuecnBeta · 11/21, 17:13
- Nvidia shares slump as SEC reportedly probes potential accounting issues36氪 · 11/21, 14:20
- Nvidia sheds $491B in a session as SEC reportedly examines accountingDEV Community · AI · 11/21, 08:53
- NO.05319Flat
Trump says he will block U.S. firms from building data centers abroad
Speaking at the White House, President Trump said he would block U.S. companies from building data centers overseas and impose a "massive tax" on those that do, according to CNBC and Axios. The remark, made after a trilateral summit with the leaders of China and Russia, targets the overseas buildout of AI compute infrastructure by American tech firms.
TS
B
N
T
C
+1Tom's Hardware, SiliconANGLE, plus 5 more7 sources·7 reports
- Trump wants to bar US companies from building data centers abroadTom's Hardware · 11/22, 00:29
- Trump plans to tax US companies that build data centers abroadSiliconANGLE · 11/21, 08:01
- Trump Threatens to Ban US Firms From Building AI Data Centers OverseasBBC News · 科技 · 11/21, 03:51
- Trump Threatens to Bar US Tech Firms From Building Data Centers AbroadNYTimes · 科技 · 11/21, 02:51
- Trump Threatens 'Tremendous Tax' on AI Data Centers Built Outside the USThe Verge · AI · 11/21, 02:50
- Trump Vows to Block US Companies From Building Data Centers AbroadCNBC · 科技 · 11/21, 02:35
- NO.06Rising288↑ 7%
GPT-6.2 Astra and Sol now generally available, with deployment comparisons
OpenAI's GPT-6.2 Astra and GPT-6.2 Sol became generally available on November 17, 2026, with Astra upgrading GPT-6 Astra and Sol building on GPT-6.1 Sol. Earlier analysis reported a 20% per-token price cut for Astra and said Sol closed nearly to Astra on a major coding benchmark at roughly one-fifth the per-task cost. The latest report cautions that the underlying material carries only Foundry testing language and provides no benchmark figures.
DT
量
I
+1DEV Community · AI, The Decoder, plus 3 more5 sources·7 reports
- GPT-6.2 Astra vs GPT-6.2 Sol: Which Should C# Developers Deploy?DEV Community · AI · 11/22, 05:00
- OpenAI Releases New Prompt Templates for GPT-6.2 Astra and SolThe Decoder · 11/21, 21:32
- GPT-6.2 Astra cuts token prices 20%, but budget Sol nearly matches itDEV Community · AI · 11/21, 20:52
- GPT-6.2 Astra doubles embodied task success rate over GPT-6 Astra量子位 · 11/21, 14:35
- OpenAI Cuts GPT-6.2 Astra Chain-of-Thought Extraction to 0.2%IT之家 · 11/21, 09:45
- OpenAI Cuts Model-Copying Attack Success Rate From 54% to 0.2%DEV Community · AI · 11/21, 05:45
- NO.07222↓ 9%
Gemini 4 Neon now live across Google platforms with expanded pricing
Google DeepMind's mid-tier Gemini 4 Neon is now available across the Gemini API, AI Studio, the Gemini app, and GitHub Copilot, positioned between flagship Carbon and lean Hydrogen with 1 million token context and four reasoning tiers. Artificial Analysis reports an Intelligence Index of 61, yet figures vary by thinking level, and coding agent scores differ across sources. Developer calculations suggest Neon becomes the cheaper option only after Carbon and Argon price changes take effect on December 1.
DX
G
I
掘
DEV Community · AI, X:TestingCatalog (@testingcatalog), plus 2 more4 sources·5 reports
- Gemini 4 Neon Costs More Per Token but Less Per TaskDEV Community · AI · 11/22, 01:51
- Gemini 4 Neon pricing: cheaper than Carbon only after Dec 1DEV Community · AI · 11/21, 22:50
- Gemini's model selector reportedly shows a periodic-table lineup of Gemini 4 modelsX:TestingCatalog (@testingcatalog) · 11/21, 19:16
- Google launches Gemini 4 Hydrogen robotics model and Antigravity Robotics pluginGoogle Developers Blog · 11/21, 04:10
- Google Releases Gemini 4 Neon With Faster, Cheaper TasksInfoQ 中文 · 11/21, 02:41
- Google releases Gemini 4 Neon, topping the agentic coding leaderboard掘金 · 11/20, 08:28
- NO.08219Flat
UN Talks on Autonomous Weapons End Without Accord, Diplomats Agree Only to Keep Discussions
Diplomats meeting in Geneva failed to agree on rules for autonomous weapons, settling only for an agreement to begin exploring options and a two-year timeline for proposing regulatory approaches. The Red Cross had called for legally binding rules within one year, but that push was not adopted. Humanitarian groups warn that military AI is advancing far faster than diplomatic efforts.
TN
T
The Verge · AI, NPR · 科技, plus 1 more3 sources·3 reports
- UN talks on rules for 'killer robots' end without agreementThe Verge · AI · 11/22, 04:03
- U.N. talks on autonomous weapons end without binding rulesNPR · 科技 · 11/21, 22:18
- UN talks on controlling autonomous weapons collapseThe Guardian · AI · 11/21, 21:15
How is this different from the industry updates board?About the briefingHide
The hot board is a rolling 48-hour window that keeps changing; the daily briefing is frozen once a day so you can look back at what happened on a given date. Both use the same events and the same heat rule: each independent source counts once, decaying with a 24-hour half-life.Open the industry updates board →
Summaries come from public feeds and remain the property of the original authors; verify important facts at the source. If you believe a summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.
- Surge
- Discussion rising fast
- New
- First report within 6 hours
- Rising
- Still gathering discussion
