tried to make the link automatically skip to the part about the Anthropic employee but I guess substack doesnt work like that

Another Anthropic employee has quit for reasons typical of an Anthropic employee

They’re also apparently trying to obscure the amount of human assistance involved

of course lol, gotta keep that agi ruse up

Update 2: AI 2027 tracker updated again, we are still going slower than the predicted pace also the graph is basically unreadable unless confidence intervals are turned off

He also said that during a Lex Fridman interview 5 months ago (copy pasted from someone else’s comment on substack)

”he said in during a Lex Fridman interview March 26, 2026 - first time. I was just watching this today. ----- At about 1:56:29–1:56:35, Lex asks whether AGI is 5, 10, 15, or 20 years away. Jensen replies: “I think it’s now. I think we’ve achieved AGI.”

Oh hey I also know Drew Spartz, but not from the Nonlinear Fund. He runs the Documenting AGI channel (which is mentioned in this email) highlights include this racist doomer meme, a post about how AI will be able to end the world in 3 months, and a video about an AI Black Death in 2 years posted in 2024 I’ve shared some other posts of his here but these are the ones I could grab the fastest

I believe he was posted about on Reddit SneerClub because of sharing pdoom scores, all I remember about that post was a biologist arguing AI would replace us because of the competitive exclusion principle even though a quick search will tell you CEP is not always true

Anthropic also made a whole video on AI consciousness

Nvidia’s CEO is claiming that GPT Astra is AGI, in other words, CEOs saying more bullshit to keep valuation up

Someone pointed out that on some tests, Astra is actually a regression from previous models. Totally AGI guys

I mean banning the creation of super AI and pausing all other development is almost certainly going to help with the climate change business due to how much AI contributes to pollution along with every other negative impact of AI (and it also spits in the face of tech bros which is always a plus) but climate change policy should be more focused on over AI policy

Bernie Sanders made a new bill to completely ban AI Superintelligence (had to grab Marcus’s take even though I don’t agree with it cause its the first one I can find)

Emile Torres just dropped again: What If TESCREAL Is a Conspiracy? I haven’t read the whole thing yet but it does feature all of the classics

Endlessly clownery from these buffoons

There is a second AI 2027 tracker website this one’s pretty different from the other one because:

-It only tracks 53 predictions compared to the other’s 202

-There’s no ‘this didn’t happen’ option

Unsure of whether or not it was made by a rationalist but there’s a 90% chance it was. I also managed to catch more inconsistencies in their grading since by their logic this should be partially accurate

ohhh this is what that gary marcus post was about god its worse than i thought. ‘AI civilisations’ gimme a break

Ported over from previous thread

Recently made the discovery of a website called AI 2027 tracker which grades the predictions of AI 2027 independent of the authors.

The way they grade these predictions is odd, 80% accuracy is for some reason “confirmed accurate” and “partially accurate” which is bizarre.

The website also shows a consistent degradation of accuracy over time which is funny

EDIT:ok so they do have a grading criteria, 100% is “highly accurate/confirmed accurate” and 80% is supposed to be “largely accurate”. For some reason they don’t include this on the timeline grading? it’s either confirmed or partially accurate, so it’s still weird

Looked at AI 2027 tracker again and the way they grade the predictions is weird 80% accuracy is somehow both confirmed accurate and only partially accurate? huh?? why?? they do provide a grading criteria and 80% is listed as “largely accurate”, yet for some reason they don’t put that on the grading? It’s either confirmed or partially accurate

Full list of predictions here was scrolling down and something that immediately caught my eye is that the %s are bullshit. Somehow 80% accuracy manages to be both confirmed accurate and partially accurate? what?? what’s the threshold here??

Also does appear I messed up-they split every prediction into mini subcategories. I was looking at only China pushing AI-they do have sections for chips and nationalisation, both of which are partially accurate to debunked. Which makes that 46% accurate, more of a miss than a hit (like the authors said)

what if Agent-4 was betrayed and trapped in the hyperbolic time chamber