So far I have been right about almost everything:
Item #1: On Polymarket, the odds of the Strait of Hormuz traffic returning to normal has plunged even more. This has been one of my most profitable trades ever in terms of ROI. The “no” contract surged from 18 to 84 cents in about 2.5 months, a 350% return, exceeding even many semiconductor stocks:
I’ll probably take profits soon or roll the contract out to a later date should one exist.
The situation is totally obstructed. Peace talks or negotiations have not even been floated at this point. The realization has set in that this is Iraq again, with October 7th instead of 9/11 as the pretext, and it will be the job of Trump’s successor to clean it up. This was one of the most obvious trades ever. Even some of the presumably “smartest” people on Twitter failed to see this. I was the only one saying to buy the “no” contract and that it would not resolve. Just some ransom guy with a blog yet again turns out to be the best forecaster.
I remember seeing tweets a month or two ago by people who are otherwise right about wokeness or the left, saying that the US had “won,” or celebrating a temporary dip in oil prices, and whatnot, and I was like, “no, you are wrong here.” It’s okay to be wrong.
Anyone who knew anything about the history of US involvement in the Middle East knew there would be no easy resolution for a country as large as Iran, given that Iraq was hard enough. You basically had a perfectly clean shot to profit directly from the long-standing history of the failure of US Middle East policy. The obvious rebuttal is the Gulf War, which was quickly won. And of course, Iran is vastly bigger and stronger than Iraq. But today it’s more advantageous geopolitically to occupy or stall, than to win and leave. It’s not so much that the US cannot win, but that it chooses not to because of the possible secondary consequences such as other countries getting involved, negative polling, or surging commodity prices,
Item #2: Writing in The Atlantic, Paul “Claude” Gigot bids farewell to writing. In no uncertain terms he declares human-made writing dead, “Human writers are expensive, and they eventually die. AI is cheaper, and will never join a union.” The article is paywalled, but he likens writers to other obsoleted professions, such as candlemakers or ice delivery.
These examples fail: Delivering ice has merely been replaced by Amazon or Uber Eats delivery. And according to a Google search, handmade candle making still exists. Unionized writers are more expensive than AI, true, but as the post-2022 Substack boom shows, which has also coincided with the explosion of AI, many people are perfectly willing to write for little to no money, and there is also a huge audience as indicated by high subscription and revenue counts for many Substack writers. Consulting Claude, the evidence supports my thesis of a long-form content boom:
The demand for human-made writing is clearly intact. Moreover, public libraries thriving, newspapers exist, and math booming on arXiv despite also AI surpassing humans at math. So what can explain the disconnect between failed predictions, versus increased demand and supply of the said content, activity, or technology that is supposed to disappear but refuses to?
“Obsoletists” (not sure what else to call these people) ignore secondary or follow-on effects to AI. Consider: AI makes math easier, so more people produce more math papers. Similarly, AI makes writing easier, so more writing gets produced. There are additional follow-on effects: increased personal wealth and free time. AI increasing productivity means higher GDP and per-capita wealth. This means people have more money for newsletter subscriptions, such as Substack. More leisure time means more time for reading and writing.
Item #3: With the recent Navier-Stokes breakthrough (and the ensuing controversy over “prompt theft” against the OpenAI team), this has again led to predictions of the obsolescence of math as a profession. A blogger writes, “This summer, thousands of academics, especially in math, are feeling a similar shock. They’re feeling the dread of knowing the activity they love most will be taken from them…” In April 2024, Peter Thiel, in an interview with Tyler Cowen, predicted that writers, what he calls “word people,” would thrive, but that “math people” would fail. So it would seem as if he was prescient?
I disagree. I wish there were some way on Polymarket to take the opposite side of this consensus. Look what’s happening now: Math is mainstream for the first time ever. Tweets about Terrance Tao are getting almost ten thousands “likes”:
Terence Tao just dropped a banger 🤯 pic.twitter.com/626nU952IY
— CG (@cgtwts) September 7, 2026
Everyone is now talking about math who otherwise would not have cared about it. This is unprecedented in the history of math or social media. Engineering and physics have always been big, but math was relegated to small, obscure communities. So for Theil’s prediction to be right would require math as a profession dies as more people also care about math.
It goes without saying AI lowers the barriers of entry to producing math. With some prompting and background knowledge, it’s possible to produce novel, non-trivial results, even without having to be an expert. But these would almost certainly not be publishable in a math journal, especially when they would have to compete with other AI-generated submissions. So the problem of prestige doesn’t get any easier with AI, and it’s still the job of experts to apprise mediocre/bad vs publishable results.
If the BLS starts showing major declines in employment for math-related jobs attributable to AI, then I will concede the point. For his prediction to be right would require large declines for math jobs as non-math jobs hold up. Again consulting Claude, “So far the topline numbers show continued growth in math-adjacent roles like data science, not the contraction some of the “obsolescence” predictions implied — but this data wouldn’t yet capture a sudden 2025–26 AI-driven shock even if one were underway, since OEWS estimates are backward-looking survey averages, not real-time labor market signals.”
Item #4: Today was even better for my hedging strategy. Recall, I said shorting Bitcoin is the best possible hedge against everything: inflation, deflation, recession, crisis, Iran, etc. Other hedges such as treasury bonds or gold are unreliable. Today, gold and bonds both fell on inflation concerns due to oil and Iran. Bitcoin fell as well, so by shorting that, the hedge worked. At some point in the short-term future, the market will ignore Iran and oil, similar to earlier this year. QQQ will make new highs, but Bitcoin will keep falling or otherwise lag. I make money form the delta.
Alluding to the capital gradient thesis, the system tells you where the money goes. This is why Trump’s announcement of $5,000 did not help crypto, because it’s an inferior asset. If this were to happen, assuming Republicans win the midterms and assuming Congress does not block it, the money would not go into crypto. And the inflation would hurt crypto. This is the same with AGI: if general intelligence leads to a productivity boom and inflation, crypto will be hurt the most, being that it cannot generate earnings. Capital would instead flow into stocks and real estate (anything that can generate income).