
My fellow AI explorers
This week, Anthropic became a public company, and the market gave it a party and a hangover on the same afternoon. The stock opened 34% above its offer price, gave back half the gain, and the rest of the AI sector fell anyway. And one day later, an Anthropic mathematician said Claude helped him prove the Riemann Hypothesis. Normal week.
This week, we'll deep dive into Anthropic's record IPO and the AI crash it couldn't stop
In today’s edition:
🏦 Anthropic's $107B debut: the numbers, the asterisks, and who cashed out
📉 The AI crash nobody wants to name: Nvidia, the Fed, and the market's six-day slide
🧾 30-Second AI Play: have ChatGPT draft your invoices straight from your inbox
Editor's Software Picks
Four tools we think are worth a look this week.
Base44
Describe the app you need, and AI builds it, with login, database, and hosting built in.
PixVerse
Turn a photo or a line of text into a short AI video, ready for Reels and TikTok.
CapCut
Edit short videos for TikTok, Reels, and YouTube with AI captions, voices, and templates.
NordVPN
Encrypts your connection and hides your location, at home or on public Wi-Fi.
Some of these are affiliate links. If you sign up, we may earn a commission at no extra cost to you.
30-Second AI Play
🧾 Let ChatGPT Draft Your Invoices Straight From Your Inbox
If you run a small business or freelance, invoicing is probably the most boring hour of your week. OpenAI just launched ChatGPT for Accounting, which reads your email, finds finished jobs, and drafts invoices in QuickBooks for you to approve. Here's how:
Install the Accounting plugin in the ChatGPT desktop app and connect your tool (QuickBooks, Xero, BILL, FreshBooks, NetSuite, and others are supported).
Pick one familiar task to start, like this week's batch of invoices.
Run the prompt: "Find jobs in my email that are ready to invoice, create draft invoices in QuickBooks for my review, and flag anything unclear."
Review before approving. Check amounts, account coding, and anything flagged. Nothing goes out until you sign off.
Save it as a reusable skill so next month's batch takes one click.
🔍 Why it's special:
It works in your existing apps and can operate the interface directly when there's no plugin.
One contractor reports saving 5 to 6 hours a week on subcontractor invoices.
On Business and Enterprise plans, your data isn't used for training by default.
💡 Pro tip: Once invoicing works, run bank reconciliation the same way. Have it match statements to your ledger and list only the differences. One franchisee reported reconciling cash in 94% less time.
Together with Terzo
💰 Big Companies Only Check Their Top 10 Deals. These AI Agents Check Every Supplier, Every Day.
A large company signs thousands of contracts and pays millions of invoice lines a year. No human team can read all of that, so most only review their top ten deals. The rest renews at last year’s price because nobody had time to check.
Terzo built the Nirvanai Platform to cover that blind spot. It pulls a company’s contracts, invoices, and spending into one place, then puts AI agents to work on it every day:
🔍 Savings Agents find better prices, discounts, and rebates the company isn’t getting
📅 Renewal Agents flag renewals before they close and build the negotiation strategy
🚰 Leakage Agents check every invoice against the contract and what was actually spent
Picture a supplier contract that says $40 a unit while the invoices quietly bill $46. A person would probably never spot it, but a Leakage Agent catches it automatically. Each finding comes with its dollar value, the math behind it, and the source documents. A daily briefing ranks what to act on first: renegotiate, consolidate, claim, or cut.
Terzo says its agents typically surface 2 to 5% of addressable spend, and Nirvanai has found $1B+ in savings opportunities for Fortune 500 companies across 14 industries.
🔮 Prediction: “We only review the top ten” is about to sound as dated as checking invoices by hand. If you lead finance or procurement, ask yourself which suppliers nobody at your company has looked at lately.
What could be hiding in your company’s contracts?
Want the version of this that actually runs in your business?
The 30-Second Play gives you the move. The Operator Brief gives you the full system: prompt libraries, API cost math, and the workflows our subscribers are using to replace entire freelance line items.
This week: how to get ahead of the game with ChatGPT’s Business Hub dashboard.
Meanwhile in AI…
🧮 One day after the IPO, Anthropic mathematician Levent Alpöge said he and Claude found a proof of the Riemann Hypothesis, a 167-year-old problem with a $1M prize attached. It’s been described as the Holy Grail of math problems, and Alpöge’s solution comes with a cool 17 million lines of Lean code. Because of that, no human has read it all yet, and mathematicians are still translating it.
🤖 Genesis AI raised a $500M Series A at a $3.5B valuation from Khosla, Premji, Eclipse, and Nvidia for its dexterous robot Eno, with customer deployments planned by the end of 2026. They call it “the world's most capable general-purpose robot that can learn to perform any kind of physical work.” Add on that human experience trains the brain, and you get a pretty capable humanoid collection of chips and wires.
💊 OpenAI launched MedChemBench, a drug-design benchmark that evaluates AI models on reasoning tasks drawn from medicinal chemistry research. GPT-6.2 Astra scored 84.5% but earned only about half the credit for picking the best compounds. OpenAI says that their results don’t actually establish whether models can discover new drugs or improve laboratory decisions, but can “help identify where models are advancing and where challenges remain.”
🌍 Trump threatened to ban US firms from building AI data centres overseas, gave little detail, and said some of his own team were hearing about it for the first time. On top of that, his push for more data centers in the US runs counter to Americans recently voting against data center initiatives. 64% of registered voters opposed an AI data center within 10 miles of their home, while just 13% of respondents said they would support one.
IPO
🏦 Anthropic Goes Public at $83 a Share, and the Market Blinked
Anthropic priced its IPO at $83 a share, above the $70 to $78 range it had marketed, and started trading on the Nasdaq Thursday under the ticker ANT. Demand was about six times the number of shares available, so the bankers raised the price and the share count.
Here's what stood out:
Record raise: About $81.3B from the IPO itself, plus a $26B private placement led by Nvidia and Google, for roughly $107B in total.
The valuation: About $1.5T counting every option and reserved employee share, or about $1.08T on shares actually outstanding.
The rollercoaster: It opened at $111.50, touched $115, then came back down and closed at $97.17, still up 17%.
Here's the uncomfortable part.
The "largest IPO ever" headline comes with an asterisk. Anthropic beat SpaceX's base $75B deal from June, but SpaceX's final offering with the over-allotment came to $86.25B. Anthropic only takes the crown outright if the underwriters exercise their option.
Look at who was selling, too. Each of Anthropic's seven co-founders, Dario Amodei included, sold nearly 3.4 million shares, worth about $282M each at the IPO price. That's not unusual for an IPO, but it says something when the founders of an "AI safety company" take nearly $2B off the table on day one.
Then there's the circular money. Nvidia and Google each put in roughly $10B through the private placement, and both make money when Anthropic spends on compute. Some of that "investor demand" is suppliers buying a stake in their own customer.
🔮 Prediction: ANT drifts back toward its $83 offer price before its first earnings report. Here's my reasoning. Pricing above the range left little room for an easy pop, so the people who got in at $111 are already underwater. Historically, hyped IPOs average only about 3.5% gains in their first 12 months, and the lock-up will release employee stock in stages, adding supply right as the novelty fades. But I don't think the business is the problem. Claude is likely the most important enterprise AI product right now. The problem is paying $1.5T for it in a market that is getting rid of AI exposure, not adding it. The long-term winners here are the people buying in the second half of 2027, not the people who bought at Thursday's open.
Did you buy ANT at the open, or are you waiting for the dip? Hit reply and tell me your entry price. I read every one.
Markets
📉 The AI Crash Nobody Wants to Call a Crash
Anthropic was supposed to save the AI trade. Instead, the S&P 500 fell 2% on Thursday, its sixth straight losing day and its longest slide since April 2024.
This didn't start Thursday. Here's how we got here:
Nov 12, the Huang shock: Jensen Huang resigned as Nvidia CEO amid a personal-conduct scandal, and Nvidia fell as much as 14% that day. CFO Colette Kress is now interim CEO.
Nov 17, the earnings shrug: Nvidia posted $115.8B in revenue, up 103%, and guided to $130B. The stock still notched its sixth straight loss on IPO day.
Nov 19, the debut hangover: Nvidia, Broadcom, Micron, Microsoft, and Oracle all slid, and neoclouds CoreWeave and Nebius took steep losses.
Then add the macro. Traders now price in a near-certain Fed rate hike in December. And after Democrats swept the midterms, including in states that oppose new data centers, investors fear regulation will slow the buildout.
Here's the uncomfortable part.
When record earnings can't move a stock, the market isn't pricing results anymore. It's pricing fear of whatever comes next. Strategists have pointed out that big waves of new stock, like 1999, 2007, and 2021, often show up late in the cycle. An $81B IPO also needs $81B of buying, and some of that money came from selling Nvidia and friends.
🔮 Prediction: This isn't 2000 yet, but the AI trade is splitting in two. One analyst described it as a bet on the winner versus a bet on the buildout, and the buildout bet is the one losing support. I expect the next three months to punish the debt-heavy middle of the stack, meaning neoclouds, data center developers, and anyone financing GPUs with borrowed money, while cash-rich labs and the app layer hold up. My reasoning: higher rates make debt-funded capex more expensive, political opposition delays power and permits, and Nvidia's leadership vacuum takes away the sector's best salesman. If the Fed hikes in December, expect a second leg down in AI infrastructure names in January. Every crash feels like "just a correction" until it doesn't.
Is this a correction or the start of the AI bubble popping? Reply with one word: "correction" or "pop." I'll share the results next week.
Advertise to 200k prosumers and tech professionals choosing what tools their companies build with.
Golden Nuggets
🏦 Anthropic raised a record ~$107B, but buyers at the open are already underwater, and the co-founders cashed out nearly $2B.
📉 Record Nvidia earnings couldn't stop a six-day slide, so the market is pricing fear now, not results.
🧮 Claude may have helped crack Riemann, but 17 million lines of code still need a human to understand them.
Would love to hear your thoughts! Send me your thoughts by replying to this email (yes, I read them all :)
Until our next AI rendezvous,
Anthony | Founder of Uncover AI
