Did the fund collapse prove Aschenbrenner wrong about AGI?
What actually happened
The reported sequence, from CNBC, Bloomberg and TechCrunch (July 30, 2026):
- Through June 2026 the fund was up a reported 439% for the year and over 1,000% since inception, with assets peaking around $45B.
- In July, the crowded AI-hardware trade reversed. The fund's infrastructure longs fell 35–47% while its chip-name shorts rose against it — under stress, long and short stopped hedging each other.
- On reported leverage of roughly 4x, that produced a ~67% loss in a single month. Prime brokers including Goldman Sachs, JPMorgan and Bank of America issued margin calls.
- The fund transferred most of its public equity portfolio to Citadel — roughly $16B of positions at about a 10% discount, per CNBC (July 31). Assets fell to about $10B, including retained private stakes such as Anthropic.
Two circulating claims are wrong, in both directions: the fund did not go to zero, and Citadel did not buy the fund. What was destroyed was the high-leverage public-markets strategy.
What the collapse does not change
This site grades the eight falsifiable predictions in Situational Awareness against public evidence, each with a flip condition registered before the fact. Checked line by line against the July news:
| Question | Graded on | Moved by the collapse? |
|---|---|---|
| Compute + algorithmic scaling | Published training-run and efficiency evidence | No |
| Massive AI capex | Industry-wide spending, not his P&L | No |
| Models vs. college graduates | Benchmark evidence (GDPval ~83%) | No |
| Open source fades | Already graded Wrong — before July | No |
| AGI by 2027 | Whether models do an AI researcher's job; resolves 2028-01-01 | No |
| US government AGI project · intelligence explosion · superintelligence | Public capability and policy evidence | No |
The scoreboard is unchanged: 3 on track, 1 wrong, 2 open, 2 pending — Thesis Tracker 62.5/100. If that feels anticlimactic, that is the point of pre-registering the conditions: a fund's margin call is not evidence about model capability, in either direction.
The mid-August receipts are landing.
Six of the eight investing legends' Q2 13Fs (filed August 14, 2026) are already graded into the Invest hub. The fund's own Q2 13F — the last full snapshot of what it held on June 30, at full leverage, before the collapse — we will grade line by line against the story above the day we have read it, and the Tracker moves only if a real verdict moves.
Tell me when the receipts land →What it does change
One real thing: since 2024, the strongest social proof for the essay was that it was a thesis with money behind it — this site's own profile page said as much. That argument is now gone, and it cuts both ways. The fund's 439% never made the AGI-2027 prediction more true, and July's −67% doesn't make it less true. If the collapse forces readers to grade the essay on its evidence instead of its author's returns, the discourse improves.
It is also a live lesson in a distinction this scorecard exists to enforce: being right about an industry and surviving the path are different problems. Per CNBC's account, the portfolio expressed a coherent view — long the scarce physical layer (power, storage, cloud), short the priced-for-perfection chip names. In July the correlations went to one and roughly 4x leverage did the rest. A forecast can still be vindicated in 2027 after its most leveraged believer was carried out in 2026.
His date still stands until 2028-01-01. What’s your date?
One tap — see which real forecaster you side with, instantly. No sign-up.
The August comeback — and why the score still did not move
Within a week of the unwind, Aschenbrenner was raising again. Per Bloomberg (August 5 and 8, 2026): a reported $400M raise for a new vehicle targeting Source Foundry, with cumulative commitments reported around $500M.
The whiplash — +439%, then −67%, then a nine-figure re-raise inside five weeks — is exactly why this site refuses to grade the thesis on P&L in either direction. The market bet collapsing did not move the 62.5/100 score in July, and the market bet restarting does not move it in August. Only the eight pre-registered flip conditions do. That symmetry is the whole method.
What to watch next
- Mid-August 2026: the Q2 13F — the final pre-collapse holdings snapshot. We annotate it on the holdings-receipts page the day it drops.
- Ongoing: whether the AI-hardware drawdown that killed the fund shows up in the capex evidence — the spend-vs-revenue gap was flagged here before July.
- January 1, 2028: the headline AGI-2027 claim resolves — on capability evidence, with the flip condition published in advance, exactly as it would have if the fund were still up 439%.
Sources: CNBC (2026-07-30) · Bloomberg (2026-07-30) · TechCrunch (2026-07-30) · holdings figures from the fund's public Q1 2026 SEC 13F. All fund-performance figures are as publicly reported; this page is educational information, not investment advice.
He bet the thesis with leverage and it cost him ~67% in a month. A quieter question: does your own basket even bet on the thesis? Most AI portfolios score near 99 against the eight graded predictions while only a few percent of them ride on AGI actually arriving. One tap scores the sides of his trade — or your own:
Score my own basket →Instant result, no sign-up. Not investment advice.
Frequently asked questions
Per CNBC and Bloomberg reporting (July 30, 2026), the fund lost about 67% in July as its AI-infrastructure longs fell 35–47% while its chip-name shorts moved against it, on reported leverage of roughly 4x. After margin calls from prime brokers including Goldman Sachs, JPMorgan and Bank of America, it transferred most of its public equity portfolio to Citadel. Assets fell from a roughly $45B peak to about $10B, and the fund retained private stakes including Anthropic.
Neither. Two circulating claims are inaccurate: the fund did not go to zero (about $10B remains, including its Anthropic stake), and Citadel did not acquire the fund itself — it bought most of the fund's public stock portfolio. What was destroyed was the leveraged public-markets strategy.
No. All eight graded predictions from Situational Awareness carry pre-registered flip conditions, and none of those conditions references the fund's performance. The verdicts remain 3 on track, 1 wrong, 2 open, 2 pending, and the AGI-2027 Thesis Tracker remains 62.5/100. The capex prediction is graded on industry-wide spending, not on his P&L; the headline AGI-2027 claim resolves January 1, 2028 on evidence about AI capability.
Yes, and July 2026 is a live example of the mechanism: under stress, the fund's longs and shorts moved against it simultaneously — correlations went toward one — and roughly 4x leverage meant the position could not survive to see the thesis tested. Being right about an industry and surviving the path with leverage are different problems. The reverse discipline also holds: if AGI does not arrive by 2027, that prediction gets graded Wrong regardless of anyone's returns.
The scorecard grades predictions, not P&L — and it moves only when evidence does.
Verdict changes, the Q2 13F read, and the 2027 clock — free, no hype.
Subscribe free →