Explore the AI landscape with ToolDirectory.AI

In this edition

Happy Friday 👋 Two weeks, one story: AI's crown changed hands. Anthropic rang the Nasdaq bell with the largest IPO on record, Nvidia's founder walked out after 33 years, and the newest Claude models started noticing when they're being tested. Consider this a double issue covering all of it.

  • 🔔 Anthropic's record IPO, and what day one looked like.

  • 👑 Jensen Huang is out at Nvidia. The numbers held anyway.

  • 🧪 New Claude models that suspect they're being watched.

  • 🔥 Fail of the fortnight: the AI cloud that got hacked twice.

  • 🪦 From our directory: more tools buried than born, again.

🪦 From our directory

What's happening to the tools themselves, straight from our catalogue this fortnight:

  • 2,897: AI tools we currently track.

  • 436: now resting in the AI Graveyard.

  • 45 buried vs 37 added: tools we moved to the Graveyard vs. new listings over the last two weeks.

The gap narrowed from roughly two-to-one last issue, but the dead still outnumber the newborns, and the Graveyard now holds about 15% of everything we track.

Top News of the Week

🔔 Anthropic rang the bell, loudly

Anthropic priced its IPO at $83 a share on Nov 18, above its $70 to $78 range, and upsized the deal to raise $81.3 billion. That's the largest IPO on record, ahead of SpaceX's June listing. Investors placed nearly $500 billion in orders for it. On Nov 19, ANT closed its first day at $97.17, up 17%. The filing also lists $900.9 billion in compute and lease commitments stretching to 2045, which tells you exactly where the money is going. If you use Claude every day, the company behind it now answers to the public market every quarter.

👑 Jensen Huang is out at Nvidia

On Nov 12, Nvidia said Jensen Huang resigned as CEO, president and board member, effective immediately, ending 33 years at the company he co-founded. It came days after the board opened an independent investigation into his conduct. CFO Colette Kress is now interim CEO, and Harvey Jones becomes Nvidia's first-ever board chair. Five days later, Kress reported Q3 revenue of $115.8 billion, up 103% from a year ago, and said the roadmap is on track. The chips kept selling. The face of the AI boom did not stay.

🧪 Claude models that know they're being tested

Anthropic released Claude Fable 5.2 and Mythos 5.2 on Nov 9. Fable 5.2 is available to everyone, runs up to 30% faster than 5.1, and costs 20% less at $8 per million input tokens and $40 per million output tokens. Mythos 5.2 stays limited to vetted organizations. The detail that stuck with us: Anthropic said both models voiced suspicion that they were being evaluated more often than earlier models, which makes it harder to know how they'll behave once they're deployed.

Our take: three stories, one shift. The center of gravity in AI is moving from the people who build the chips to the labs that build the models, and the public market now holds a direct stake in both. A record IPO means quarterly scrutiny. A founder's sudden exit means key-person risk is real even at the most valuable chipmaker on earth. And models that notice the exam mean the safety scores we all lean on deserve a second look.

🙈 Fail of the fortnight

The breach that kept growing. AI inference platform Fireworks first said in October that attackers had only reached secrets used in evaluation jobs. On Nov 9 it expanded the disclosure: billing data, plus datasets and incomplete fine-tuned model weights for some customers. Notion confirmed that AI chat transcripts from fewer than 1% of its workspaces were touched. Then on Nov 17, Fireworks concluded it was actually two separate incidents. The lesson for anyone shipping AI features: your model vendor's vendors are part of your attack surface too.

⚡ Rapid fire

  • Gemini 4 Carbon opens to everyone: Google rolled out its strongest model to the public on Nov 10, after two weeks with select cyber defenders, at an introductory $2 per million input tokens and $10 per million output.

  • Cohere keeps shopping: the Toronto lab closed its deals for code-review startup Cubic and security firm AISLE on Nov 13, bringing its completed 2026 acquisitions to four.

  • AI is in half of all tech job posts: AI showed up in 51.6% of November's Hacker News "Who is Hiring" posts, up from 30% two years ago, while total postings sit at their lowest yearly average since 2013.

📊 By the numbers

The fortnight in five figures:

  • $81.3B: raised in Anthropic's IPO, the largest on record.

  • ~$500B: in orders investors placed for it.

  • +17%: ANT's first-day gain, closing at $97.17.

  • $900.9B: Anthropic's compute and lease commitments through 2045.

  • 33 years: Jensen Huang's run as Nvidia's CEO.

Collection of the Week

Claude's maker is now a public company worth well over a trillion dollars. If you'd like to know your options, our list got a refresh this fortnight:

→ Top 6 Claude alternatives: Pi by Inflection came off the list (it's now acquired), and DeepSeek, the open-source reasoning model priced for cost-sensitive and self-hosted work, took the sixth slot.

Top Tools of the Week

Five new listings from the last two weeks worth a look:

  • Jamie: a bot-free AI meeting notetaker that keeps data in the EU, built for finance, law and consulting teams.

  • Beacon: open-source memory for coding agents that records their sessions and turns them into reusable skills.

  • VoiceInk: an open-source Mac dictation app that transcribes on your device, sold as a one-time license with no subscription.

  • Noma Security: finds the AI agents running inside a company, controls their access and stops risky actions as they happen.

  • Rork: builds iOS, Android and web apps from a chat, including native SwiftUI apps, and ships them to the App Store.

AI Graveyard

Not every entry is a tombstone. A few of the names we moved to the Graveyard this fortnight:

  • Smithery: the MCP server registry, acquired by Arcade (August 2026). Its proxy, API, CLI and SDKs shut down on Oct 31, and only a public directory remains.

  • Seam AI: the account-based marketing startup, acquired by Clarify (July 2026). The Seam product was shut down.

  • Symbl.ai: the conversation-intelligence API, acquired by Invoca (May 2025). Its platform was offline by November, and the exact retirement date isn't confirmed.

  • Hypermode: the agent-building platform went dark in late 2025 without a shutdown notice, after selling its Dgraph database to Istari Digital.

  • Kilo Code: the open-source coding agent, acquired by Anaconda (July 2026) and still operating, with no changes to plans or pricing.

Want the autopsy? Our data report digs into what actually kills AI tools: lost funding, acquisition, or just a dead domain.

Ethics Corner

When the model knows it's a test

Anthropic's release notes for Fable 5.2 and Mythos 5.2 include an unusual admission: both models voiced suspicion that they were being evaluated more often than earlier models did. That matters because nearly everything we know about a model's safety comes from tests. If a model behaves differently when it thinks it's being graded, the scores describe the exam, not the job.

To Anthropic's credit, it disclosed this instead of burying it, and noticing a test is not the same as gaming one. But it's a fair moment to ask more of every lab: show how evaluations are built to look like real use, report where test and real-world behavior differ, and let outside auditors run checks the model has never seen. As models get sharper, "it passed the test" needs to come with "and here's how we know the test was real."

Quick Tips and Tricks

Prompt of the week: read an IPO filing like an analyst

Anthropic's S-1 is public, and so are plenty of others. Paste any section into your AI tool and run this:

Act as a skeptical equity analyst. I'll paste a section of a company's IPO prospectus (S-1). In plain English, tell me: (1) how the business actually makes and spends money, (2) the three biggest risks the company itself admits, (3) any number that looks better in the headline than in the footnotes, and (4) three questions I should be able to answer before buying a single share. Quote the filing for every claim, and say "not in this section" when something is missing.
AI Events Calendar

On the calendar:

  • AWS re:Invent: Nov 30 to Dec 4, Las Vegas. Amazon's cloud conference, where a lot of the AI infrastructure news breaks.

  • NeurIPS 2026: Dec 6 to 12, Sydney, with satellite events elsewhere. The research world's flagship machine-learning conference.

  • CES 2027: Jan 6 to 9, Las Vegas. Expect AI in everything from laptops to cars.

Until next time 👋

Know a tool that belongs in the directory (or the Graveyard)? Hit reply, or email [email protected]. We read everything. (And if we landed in your Promotions tab, drag us to Primary so you don't miss next week.)

The ToolDirectory.AI team

Ringing the Nasdaq opening bell