There's a lot riding on Discord's IPO
Also: How rules lawyering saves costs for prediction markets
Welcome to Cautious Optimism, a newsletter on tech, business and power.
📈 Trending Up: US stocks … US job growth? … podcast monetization on Spotify … disease discovery … billionaires in California … LinkedIn Lunatics … AI voting … Chinese AI IPOs
📉 Trending Down: Paramount’s WBD dreams … Russia’s tanker fleet … the health of our children … billionaires in California … Amazon’s AI chops … Android’s open source bonafides …
Things That Matter
How the turns table: After Manus moved from China to Singapore and then agreed to sell itself to Meta, it looked like Chinese AI startups had found a path to raise international capital and reach a global audience.
Not so fast. The Chinese government is poking at the Meta-Manus deal, which means it could yet die.
xAI raises $20B more: As expected, xAI has raised a massive new funding round from Valor, Stepstone, Fidelity, Qatar Investment Authority, MGX, NVIDIA, and Cisco, among others. The company plans to use the fresh $20 billion to “accelerate [its] world-leading infrastructure buildout,” and make progress towards its ultimate goal of “understanding the universe.”
The deal shows investor demand for AI lab shares remains fevered.
The transaction ensures that the US AI race remains a stampede, with well-funded indies (OpenAI, Anthropic and xAI) and giants (Google, Microsoft, Amazon and Meta) all working to build better AI models than their rivals and partners.
It’s worth noting that xAI’s Grok has gone up in Similarweb’s 'Gen AI Website Worldwide Traffic Share’ from 2% three months ago to 3.4% in the most recent survey. Perplexity has dipped to 2% from 2.4% three months ago, while Google’s Gemini products have carved a pound of flesh from OpenAI.
xAI says it has Grok 5 in training, and will likely make news for its advancements over Grok 4.1, which scored very well on industry leaderboards. Sure, xAI can’t help but step on its own bits every few months, but it is building competitive models, too.
Claude Code is great: Following our coverage of the enthusiasm around Anthropic’s Claude Code yesterday, I got my credit card out, signed up for yet another AI subscription, and tried to get the thing working on my gaming rig. I had to download and install all sorts of software and set up my development environment, but after pushing through, I coaxed Claude Code into building me a neat webpage that tracks financial news, this blog, and more. Then I ran out of credits on my $20 plan, and was told to wait a few hours before I could continue.
Next up? Building myself a GTO trainer.
If Claude Code becomes 2x easier to set up, 2x as powerful, and 0.5x the cost in a year’s time, why won’t everyone 50 and under be able to roll their own personal digital suite?
Should I give OpenAI’s Codex a spin, too? Reply to this email if you think that’s a must.
Defining reality: Prediction markets are seeing such rapid adoption that leading player Polymarket is confident enough to begin monetizing its trading flow. But they remain a little brittle, despite the current regulatory favor and increased market awareness thanks to a blizzard of partnerships.
Kalshi users who bet on more than 10.5 San Francisco NFL wins this year are in a muddle after the prediction market allegedly didn’t pay out their winnings. Reporting indicates that Kalshi goofed up and settled the bets too early, and users who wagered correctly merely got their bets back. As you can imagine, they are not enthused.
Relatedly, Polymarket users who bet millions the US would invade Venezuela by the end of January thought that they won the wager. After all, the United States went in, killed dozens of people, extricated its premier, and now claims to run the country at the sole discretion of POTUS. The market in question states:
This market will resolve to "Yes" if the United States commences a military offensive intended to establish control over any portion of Venezuela between November 3, 2025, and January 31, 2026, 11:59 PM ET. Otherwise, this market will resolve to "No".
But Polymarket is splitting hairs. As The Guardian puts it, after the platform “clarified that the seizure of Maduro did not qualify for a winning bet, the odds of an invasion before the end of January crashed to below 5%.”
Here’s the “additional context” that Polymarket added to the market:
This market refers to U.S. military operations intended to establish control. President Trump’s statement that they will “run” Venezuela while referencing ongoing talks with the Venezuelan government does not alone qualify the snatch-and-extract mission to capture Maduro as an invasion.
Polymarket sponsors TWiST, the podcast I help host, and we therefore discuss its markets on the regular. We have stated ad nauseam that you must read the rules for markets you intend to enter and fully understand them. That said, it seems that we’re going to need much more robust wager resolution language to prevent similar situations reoccurring.
The conspiracy angle here is that Donald Trump Jr. is an advisor to Polymarket, and the Trump administration has argued that it didn’t invade Venezuela. So if Polymarket said that it was an invasion, well, you can imagine the messaging mess that would ensue. I struggle to give Trump Jr. credit for that level of operational fastidiousness.
Add NinjaOne to the IPO hopefuls list: IT operations and endpoint management startup NinjaOne passed the $500 million ARR threshold this week, CNBC reports. That figure is up “nearly 70%” from a year ago total (implying ~$294 million in January, 2024 ARR), and the startup, valued at $5 billion last year, expects to grow 60% to 70% this year. That would put it at around $800 million to $850 million ARR by this time next year.