Should’ve left in the full footnote:

”The AI 2027 scenario is still roughly what we expect the future to look like: a mad scramble to superintelligence leading to either AI takeover or extreme concentration of power. So far reality is tracking closer to AI 2027 than even we expected. (2027 was our modal year at time of publication, not our median.) You can read more about our views on timelines here and here.”

So yeah you’re bang on the money. Funny how they went on a press tour hyping up Kokotajlo as a forecaster and claiming that dismissing AI 2027 as hype was a “grave mistake” to “yeah we did not expect for it to follow our predictions” and still being slower than what they predicted

Piggybacking off myself to say that the way they wrote the update on their 2025 review is weird. I’m assuming that they mean 2025 was 65% pace but 2026 crept up to 75%, but the way they wrote it makes me think that they actually meant to say “we originally thought 2025 was 65%, but new data shows it was actually 75%” which would be odd, I’m still assuming the former since it makes more sense to me (they also drop it with zero elaboration like, what metric increased with this new data since this was an overall % of pace? not to mention the real percentage was 58-66% based off their means and medians so it would actually look like 58-76% pace)

I mean no matter which way you spin it we’re still going slower than AI 2027, so hooray

(turned this into a second comment so they first one wouldn’t be a wall of text)

Oh damn, they originally said it was gonna be called AI 2030, so they pushed it back a bit

It’s not a rehash of AI 2027 with moved-back timelines like I thought it would be, instead its what they believe should happen in regards to AI development

Plan A is our positive vision for what should happen instead. In this scenario, humanity delays the development of superintelligence until 2040, makes all AI research public, allows dozens of companies globally to catch up to the frontier, and intentionally enters a regime of mutually assured compute destruction.

And in the first footnote:

So far reality is tracking closer to AI 2027 than even we expected. (2027 was our modal year at time of publication, not our median.)

According to them, preliminary analysis of the data as of July 2026 says that rate of progress is 75% of AI 2027. This says Daniel K only believes in a 25% chance of AGI by the end of 2027, their model also hasn’t changed so their medians are still shown as beyond 2027. Footnote 11 also flat-out says they have no idea how much things will keep progressing

I’m not gonna be able to go through the whole thing because busy today, but those are my observations based off a quick glance

so its Bad bad now

Unfortunately for everybody, it’s managed to outperform even the most cynical doomsday forecasts, to the degree that the US economy is now in even worse shape than it was right before an infamous downturn in the late 1920s. That’s according to the Telegraph‘s economics columnist Russ Mould, who notes that the overvaluation of US stocks has passed the level that brought the stock market to its knees to kick off the Great Depression.

The US Treasury has also admitted that the AI bubble poses systemic risks

get ready for it to get bloody

Harari is an open transhumanist from when I did some research into him when I found him on this interview so this seems in character

They also made the “Yudkowsky is frequently, confidently, egregiously wrong” post

They are writing another scenario called “AI 2030” (which we already discussed here) so I wouldn’t be surprised if they decide to ignore the elephant in the room

As expected. It’s grifting the whole way down, never about “safety” but always about money

Then hand me that hammer and let’s get to smashing those data centres

“We have to pick between the climate or AI”

No we don’t??? One problem directly affects the other. Stopping AI data centre build-out helps with stopping climate change.

I could’ve sworn I posted this here but apparently not: the first AI winter in the 1960s-70s which is pretty relevant to today

well great now I’m thinking about this

You’re telling me that the Autism Spectrum is a spectrum and not just the quirky thing people have?

The AI safety company when they actually have to make their AIs safe instead of just making vague suggestions

AGI’s ETA: Delayed (again) due to technical difficulties

came across this one when doomscrolling, pretty good article with healthy doses of skepticism. one of the three commenters on it appears to be a rationalist though (“empirical evidence” was a hard giveaway)

More on OpenAI’s financial situation

Appears so. Their doomposting shot them in the foot so they have to backpeddle

that the level of capability displayed there is widely available from other models (including OpenAI’s GPT-5.5)

please god let them block ChatGPT access like they did with Fable. it would be so fucking funny

On a related note Mythos has been released and as pointed out, the original “GPT-2 is too dangerous to be released” post was written by Dario Amodei, Daniela Amodei and Jack Clark, who are all Anthropic’s founders. Same ploy, different model

Is AI profitable yet? a website that tracks the spending and revenue of leading AI companies. The answer is so helpfully provided at the top of the page