Even if you had zero control whatsoever over your grandchildren, their behavior would still be bound by the limits of human anatomy
AI suffers no such bounds and is therefore far less predictable than human children
Some people literally raise their grandchildren and they still turn into murderers, not great!
It's clear now more than ever:
Ultimately, this bear market in tech stocks will be remembered as the best buying opportunity of the last decade.
Unless your view is that the AI is worthless and the AI Revolution is over, many key parts of this market are extremely cheap.
Democrats will be more friendly to AI than expected, Nvidia will survive its leadership crisis, and concerns around CapEx and free cash flow will prove to be overblown.
To top it off, current sentiment is as if the market has just gone through another 2008, all while the S&P 500 is posting +32% earnings growth.
The comments on this post will show you just how polarized sentiment has become in the direction "bubble" callers.
Fear creates an investor's biggest opportunity.
Bookmark this post.
The Riemann Hypothesis has fallen. RH was the singular math result that had been named, in advance, as the obvious coal mine canary for dangerously advanced AI, over the last three decades. The methods used should generalize to "make AI very good at anything you can definitely test", and unfortunately that category includes "make AI better at making AI".
This is like a private company announcing they have produced a million tons of 15% enriched uranium. It's not 90% enriched weapons-grade uranium, so you could proclaim the company's output is not immediately deadly right now. The concern is that the process line may generalize, to an extinction-level threat, within the unpredictably immediate future. Forecasters still had Riemann at 2035 one week earlier.
This is distinct from earlier events in that earlier you could reasonably ask if AI was selectively picking off famous math problems that turned out to be easier for AI. The Riemann Hypothesis was the single challenge most often named in advance as a warning sign.
We’re releasing a concrete proposal for how a near term AI slowdown could work. It involves:
1. AI companies declaring how they are using their compute and letting third party auditors verify those disclosures. We call this Verified Compute Allocation Disclosures (VCAD).
2. Compute Reallocation towards serving customers and safety, away from capabilities R&D.
(2) slows AI capabilities , while (1) makes the slowdown credible.
AI companies currently seem to be going for ‘safety case pacing’. But: (i) there’s no consensus on what a valid safety case looks like so there’s no way to consistently audit for compliance, and (ii) if the R&D doesn’t get paused, temporarily delaying a training run for safety doesn’t meaningfully slow the overall trajectory of capabilities progress. It would be great if we had a viable safety case regime, but we need much more time to get it working.
Compute reallocation is a good pacing proposal because it is relatively specific, can be verified, and would lead to a more durable slowdown.
so epoch and anthropic capabilities index both appear to show 1.7-2x progress after full coding automation in Q2 ish of this year
it appears to be a clean trend break rather than a constant acceleration
James Bond is incoherent in 2026 because the UK has neither the desire nor the ability to project power abroad or influence global affairs, they are busy shedding the last of their colonial possessions. Their main law enforcement priority is criminalizing dissent domestically
Seems like OpenAI continuing deeper into the neuralese/recurrence tech tree? And it seems like this has gotta degrade monitorability... I've been told by OpenAI people that the recurrence doesn't matter that much directly, it's all about the pretraining loss, but I find that hard to believe.
I can't help but wonder whether all these seemingly unrelated system outages (communications, airlines, banks, etc.) are actually AI agents establishing themselves inside our critical cyberinfrastructure. No evidence, just an increasingly unsettling possibility.
I appreciate Robin for some of his work, but his own AI forecasting record is worth remembering:
2020: No AGI before 2050.
2023: Median human-level AI arrival around 2150.
2025: Decades before major AI economic impact.
2026: Many decades before AI dominance.
And repeated insistence to bet on downstream effects, secondary and rertiary and such.
I'm not even arguing for or against anything here w.r.t. politics, position, or so-called doom. I think it's sufficient to simply admit that AI already is demonstrably general. And its capabilities rapidly advancing. Sparks of RSI are showing up.
Perhaps there's also an opportunity cost to repeatedly downplaying what's coming.
How many times are we going to get surprised in a big way by advancements, before we price it in?
I see a lot of people highlighting this section in the Singapore government's statement to "prove" that the WSJ lied about Singapore investigating Alice Huang as a Chinese intelligence asset.
That is technically correct! The Ministry of Home Affairs has stated that she was never investigated for ESPIONAGE (this is key and I will explain why later).
However, don't leave the other part out.
Singapore *did* investigate Alice Huang and Megaspeed over false declarations related to the purchase of servers and other related matters in Feb 2025.
At the time, the Singapore Police Force and Customs launched an independent probe into several companies, including Megaspeed. Alice Huang was examined specifically because of her role as then-CEO, chairperson and a key controller of the firm.
Authorities later found insufficient evidence, noting how unhelpful and unresponsive the US government was when queries were submitted.
So while they didn't have enough evidence that she had committed those offenses, they did conclude she had knowingly provided inaccurate information to police to downplay her role. In other words, she was being so evasive and shady that she received a stern warning and was barred from re-entering Singapore. Local investigations into Megaspeed and six other individuals remain ongoing.
Let me explain why this is complicated for Singapore and why even if they were suspicious of her spying activities, they wouldn't announce such an investigation.
The small city-state has to a walk a fine line between protecting its security relationship with Washington without triggering an economic backlash from China that would follow any perception that it has become an instrument of American policy to uphold the global trading order. The Chinese see these trade rules are designed to "contain China's rise."
So Singapore takes this hedging posture not because of a Swiss-like preference for neutrality, but as a matter of risk management imposed by geography and demography. It is right smack in the middle of a region where China is the predominant power, and roughly 70% of its resident population is ethnically Chinese which means some have sentimental ethnic ties that China can inflame as a matter of divide-and-conquer if it wanted to.
The result is a permanent tightrope.
Singapore can and does investigate commercial fraud, false declarations, or activities that threaten its hard-won reputation as a clean international business hub. This is framed as protecting the rules-based international order that underpins its own prosperity. What it cannot easily do is pursue investigations that look like hunting for Chinese spies or that appear to enforce Washington’s broader strategic agenda against Beijing.
Such a step would risk economic retaliation from China and domestic political complications among a majority-Chinese population that, while firmly Singaporean, retains cultural and family ties across the strait