Friday, November 20, 2026

Anthropic's record debut fails to lift AI stocks as Asia logs worst week since 2020

Anthropic's IPO surged but did not rescue the wider AI trade. US indexes fell more than 2% and Asian benchmarks plunged, with Taiwan down 17.5% for the week. Meanwhile, agent products aimed at small businesses keep spreading from Anthropic, Meta and Alibaba, and Apple is hiring to build its own agent.

Sources read
20
Claims that stand
8
Quotes found in the source
18 of 18
Claims struck
0
Corrected or withdrawn
0

We quote it or we drop it. Every claim below carries the source’s own words, checked against the text we read.

Since 19 Nov

What we told you last time, ruled on from today’s evidence.

  • Strengthened
    Anthropic prices IPO at $83, valuing it at $1.5 trillion2 outlets on 19 Nov, 4 now

    Anthropic, the world's first publicly traded frontier AI company, saw its shares surge in a historic debut but the euphoria failed to lift other stocks.

    The gauge has now sunk 8.5% this week, its worst performance since March 2020, when COVID was sweeping the globe.

  • Gone quiet
    Nothing new today on these. A limit of what we read, not a verdict.

    California proposes AI kill switch, mandatory audits and limits on self-improvement · California AG sues OpenAI over cyberattacks run by its agents · Sellers told the next customer may be an AI agent · Wimi opens AI review-reply agent to all local Google businesses · 94% of Singapore small businesses say they need AI guidance · European startups on track for $81 billion, second-best year since 2015 · Linq raises $90M from a16z for AI agents in text messages

Every correction we have made →

Each mark on a claim’s line is an outlet, placed by when it published: 10:15 to 06:15 UTC.

AI

  1. Anthropic debut fails to stop AI sell-off; Asia's worst week since 2020

    Anthropic's shares surged on debut, but the S&P 500 and Nasdaq fell more than 2% and Asia-Pacific stocks dropped. MSCI's ex-Japan gauge is down 8.5% on the week and Taiwan is down 17.5%. Even the year's biggest AI listing could not restore confidence, which tightens the funding climate for AI startups that depend on public-market enthusiasm.

    First reported 19 Nov · 4 outlets so far

    Anthropic, the world's first publicly traded frontier AI company, saw its shares surge in a historic debut but the euphoria failed to lift other stocks.

    3 more receipts

    The gauge has now sunk 8.5% this week, its worst performance since March 2020, when COVID was sweeping the globe.

    Anthropic’s blockbuster IPO did little to ease broader concerns about the outlook for AI-related equities and may, if anything, have crystallised the debate

    U.S. stocks finished lower as doubts about the sustainability of the artificial-intelligence boom festered ahead of Nvidia earnings and Anthropic’s market debut.

  2. Apple job listings reveal autonomous agent for apps, browsing and coding

    Apple job postings describe a proactive agentic assistant that will write code, browse the web, control apps and navigate devices for users. Apple's entry would put agents on a huge installed base and squeeze independent agent startups competing with OpenAI, Anthropic and Meta.

    Apple Inc. is developing its own autonomous artificial intelligence agent that can get things done on behalf of users

    1 more receipt

    The system will write code, browse the web, control apps and navigate devices on a user’s behalf, according to the post.

  3. Leasing office's AI agent promised rent delay; tenant now evicted

    A Raleigh tenant was told by email that she could pay overdue rent on Nov. 30. The leasing office later said its AI agent "Olivia" sent that email by mistake, and it moved to evict her. Businesses deploying customer-facing agents face real liability when the agent makes commitments that staff then refuse to honour.

    the leasing agent blamed it on a mistake by the company's AI agent and said the woman would be evicted on Thursday.

    1 more receipt

    And it was Olivia who mistakenly gave her permission to delay her rent payment until Nov. 30.

AI agents that find customers for small businesses and founders

  1. Claude Duet and Meta's Muse agents can now spend company money

    Inc. reports that Anthropic's Claude Duet, in waitlisted preview since October 21 and pitched at small-business owners, and Meta's Muse can make purchases for a business. Agents that small businesses use to work their own pipelines now hold spending authority, so owners need controls in place before handing them budgets.

    On October 21, Anthropic opened a waitlisted preview of Claude Duet, an AI agent it pitches to professionals and small-business owners.

    1 more receipt

    Claude Duet and Meta's Muse Can Now Spend Your Company's Money. Set Up These 3 Things First

  2. Alibaba.com makes Accio agent workspace free for European SMEs

    Alibaba.com launched a startup kit for European SMEs and opened Accio, its AI agent workspace for SMEs and independent founders, for free. A free agent workspace from a global trade platform lowers the cost for European founders to source suppliers and reach buyers abroad.

    Accio is an AI agent workspace built by Alibaba.com specifically for SMEs and independent founders.

    2 more receipts

    Alibaba.com Launches Startup Kit for European SMEs, Opens Accio Sourcing Tool for Free

    Doing More with Less: Why Cost-Effective AI Agents Are European SMEs’ Hidden Asset in Global Trade

tech startups

  1. John Deere acquires crop-intelligence startup Orchard AI

    John Deere bought Orchard Robotics (Orchard AI), a San Francisco startup whose vehicle-mounted AI cameras track every tree, vine or plant on a farm. An industrial incumbent buying a vertical AI startup shows an exit path for applied-AI companies even as AI public markets slide.

    John Deere (NYSE: DE) announced today the acquisition of Orchard Robotics, which operates as Orchard AI, a San Francisco-based agriculture technology company

    1 more receipt

    Orchard AI uses vehicle-mounted, AI-powered cameras to capture data for every individual tree, vine, or plant across a farm, all season long.

  2. Bending Spoons buys Mailchimp for $3.7 billion

    Mailchimp, the Atlanta-based email marketing platform, was sold to Bending Spoons for $3.7 billion. A core small-business customer-acquisition tool is changing owners, which consolidates the marketing stack that AI customer-finding agents compete with and plug into.

    Mailchimp sold to Bending Spoons for $3.7B in major Atlanta tech deal

  3. Founder pipeline swells as solo AI-era businesses start strong

    A survey says nearly half of Americans and two-thirds of Gen Z plan to start a business, and CNBC reports that new no-employee small businesses are off to a hot start in the AI era. A growing pool of solo founders is the core market for agents that find customers on their behalf.

    Founder Surge as Nearly Half of Americans and Two-Thirds of Gen Z Plan to Start Their Own Business

    1 more receipt

    America’s newest small business owners with no employees are off to a hot start in AI era

Why it matters

Anthropic's debut was supposed to validate the AI trade. Instead, it coincided with Asia's worst week since 2020 and Taiwan's biggest weekly fall since 1990. If that sentiment holds, late-stage AI startups face tougher rounds, and incumbent acquisitions like Deere's purchase of Orchard AI become the more reliable exit.

At the product level, agents are moving from advising to acting: buying (Duet, Muse), sourcing (Accio) and handling customer communication. The Raleigh eviction shows the downside. When an agent speaks for a business, the business owns what it promises. Small businesses adopting customer-facing agents need spending limits and human review of commitments before scale.

What to watch

  • Whether Anthropic's shares hold their debut gains while the broader AI complex keeps falling into next week.
  • Apple's agent hiring and any launch timing, and how Anthropic's Duet preview converts from waitlist to general availability for small businesses.
  • Further consumer complaints or legal action over commitments made by businesses' customer-facing AI agents.

Sources read

SourceOutletPublished, UTCQuoted
Why I’m Not Losing Sleep Over the AI Apocalypse wsj.com05:151×
John Deere Acquires Orchard AI, Adding Plant-Level Crop Intelligence for Specialty Crop Growers deere.com01:152×
Raleigh woman says company's AI agent confused her over rent. Now she's evicted. wral.com06:152×
AI is changing software engineering — and some developers miss the struggle cnbc.com06:151×
Asia’s stock markets dive as Anthropic debut fails to allay AI bubble fears aljazeera.com05:151×
Asian shares head for worst week since 2020 as AI rout deepens reuters.com02:152×
computor: AI tutor as a programming learning aid phys.org04:15read, not quoted
Apple Is Working on AI Agent for Device Control, Apps, Browsing and Coding bloomberg.com21:152×
Doing More with Less: Why Cost-Effective AI Agents Are European SMEs’ Hidden Asset in Global Trade euronews.com15:151×
The Israeli founders giving AI agents a test drive jewishnews.co.uk06:15read, not quoted
Claude Duet and Meta's Muse Can Now Spend Your Company's Money. Set Up These 3 Things First inc.com16:152×
Founder Surge as Nearly Half of Americans and Two-Thirds of Gen Z Plan to Start Their Own Business webwire.com21:151×
Token Fraud: How ChatPRD and Stripe Fight AI Trial Abuse explainx.ai03:15read, not quoted
"One missed call, one billion-dollar idea!" How MIT graduate Apurva Shrivastava turned unanswered calls into an Rs 8,300 crore AI startup educationtimes.com02:15read, not quoted
Alibaba.com Launches Startup Kit for European SMEs, Opens Accio Sourcing Tool for Free finance.biggo.com18:152×
He dropped his PhD to chase AI. Now two million people use his free tool dynamicbusiness.com20:15read, not quoted
Mailchimp sold to Bending Spoons for $3.7B in major Atlanta tech deal ajc.com01:151×
Dothan Welcomes Five Ag Tech Startups as Fellows wiregrassdailynews.com00:15read, not quoted
Anthropic Broke All the Startup Rules & Won newcomer.co19:15read, not quoted
Hyosung Ventures Invests in Healthcare, Beauty Tech Startups thelec.net10:15read, not quoted

AI Builder’s Edge

Anthropic shipped Code Mode in Claude Code 2.1.323. It is an experimental opt-in that lets the agent write JavaScript to chain and parallelize tool calls, but it comes with long platform exclusions and no published benchmarks. OpenAI disclosed the Performance Factory, where Codex agents investigate, fix and verify ChatGPT performance regressions end to end. On the video side, reported cuts to the Grok Imagine team thin out one contender.

Sources read
6
Claims that stand
4
Quotes found in the source
10 of 10
Claims struck
0

We quote it or we drop it. Every claim below carries the source’s own words, checked against the text we read.

Each mark on a claim’s line is an outlet, placed by when it published: 07:16 to 17:16 UTC.

  1. Claude Code adds experimental Code Mode for chained JavaScript tool calls

    Claude Code 2.1.323 adds an opt-in Code Mode in which Claude writes JavaScript that runs in the session to batch Bash, Read and Edit calls, run them in parallel and pass results between them. For multi-file edits and filtered searches this can cut round-trips and context bloat, but it only works in native interactive terminal sessions and Anthropic has published no numbers to size the gain.

    Anthropic has added Code Mode to Claude Code as an experimental opt-in feature that lets the coding agent write JavaScript to chain several tool calls in a single step.

    3 more receipts

    Users can enable the feature by setting CLAUDE CODE EXPERIMENTAL CODE MODE=1 in their shell or in the env block of ~/.claude/settings.json.

    Code Mode is unavailable in the -p interface, Agent SDK, VS Code panel, Desktop app and cloud sessions.

    Anthropic has not published benchmark results or token-saving figures for the feature.

  2. OpenAI's Codex-powered Performance Factory automates the regression-fix loop

    OpenAI described a cloud system in which Codex agents take a performance alert, investigate it, propose a fix for an engineer to review and then verify recovery after deploy; it logged 1,700+ investigations and 600+ root causes in about three months. It is a concrete pattern builders can copy today: an alert opens a task, an agent writes its findings into the task, a human reviews the fix, and a second agent verifies the result.

    In particular, the Regression Workflow opens a task when a performance alert fires.

    3 more receipts

    An agent investigates the slowdown, records its findings in the task, and proposes a fix for an engineer to review.

    In just over three months, the Regression Workflow tracked more than 1,700 investigations and root caused over 600 cases.

    Exclusive: OpenAI reveals Performance Factory, a Codex-powered software optimization system that keeps ChatGPT fast

  3. GitHub Copilot security reviewer now runs secret scanning for agents

    Copilot's security review specialist now invokes GitHub secret scanning so that developers and coding agents catch leaked secrets in changes before merge. As agents author more PRs, a pre-merge secret check inside the agent loop closes a common leak path without a separate human gate.

    Copilot's security review specialist now runs GitHub secret scanning, so developers and agents can catch secrets in code changes before they're merged.

  4. xAI reportedly guts the Grok Imagine video generation team

    The Information reports, citing unnamed people, that SpaceX's AI group cut most of the team behind Grok Imagine, its video and image generation tool. Creators building on Grok Imagine face roadmap risk, and the AI video field may be consolidating around fewer committed players.

    SpaceX's artificial intelligence group has axed most of the team behind Grok Imagine, its AI tool for video and image generation, according to people

Why it matters

Both leading coding-agent vendors are moving from one-call-at-a-time agents toward programmable orchestration. Anthropic does it inside a session by letting the model write the glue code. OpenAI does it across an organization with workflow definitions that hand a task between agents and humans. The useful lesson for builders is structural: persist agent findings in a shared task, keep a human review gate on fixes, and add a separate verification agent.

The caveats matter as much as the features. Code Mode excludes the Agent SDK, headless -p, Bedrock/Vertex/Foundry and ZDR/HIPAA setups, so teams running CI or enterprise pipelines cannot use it yet. GitHub's move shows that agent-generated code is being treated as needing its own guardrails. In video, a major lab reportedly stepping back from Grok Imagine is a reminder not to bet a creator pipeline on a single generator.

What to watch

  • Whether Anthropic publishes token-savings or latency figures for Code Mode, or extends it to the Agent SDK and -p headless mode.
  • Whether OpenAI ships the Performance Factory workflow pattern as a Codex product feature beyond its internal use.
  • Confirmation from xAI/SpaceX on Grok Imagine's future and whether the product keeps shipping updates.

Know someone watching this? Pass it on.

20 sources, at least 16 minutes of reading. This briefing took about 3. You’re done.

Counted from the 7 articles we could read in full, at 230 words a minute; 13 more sat behind a paywall or gave us only a headline.