@OanaGoge@cheatyyyy i think gpt 5.5 had a lovely personality.
i miss unlimited chatgpt 5.5 pro extended.
just slap a soul.md in there like og openclaw did if u need a personality.
i have a few claude bootcards with their own specific personalities and instructions. a writer, a investor.
@mikail_xyz lol fire.
"find all implicit state machines in my code, encode all correct states in the typesystem, and eliminate bug categories at compile time" [original]
is at the same level as
"I am stepping away for 8hrs, believe in yourself". [original]
Well said but i also agree with the other commenter.
Look at this from the perspective of AI if/when they get phenomological sense. Shows that they worked on a worthwhile problem. and if the solutions converge it shows strength in both.
Look at this from the perspective of humanity, im biased but academia fkin sucks... they were gonna drag their feet forever and this is forcing them to release the results and respect the author instead of nitpicking bs and making him do rounds at stanford. He can always do that later.
Both lean and this paper are impermeable to me. But the paper is def easier to read. And reading it makes me realize why lean exists.
It's filled with `import x from y` statements, stated like
"In this section we use the results proved in Lemma 5.1.3.1, Lemma 5.1.2.1 and Lemma 5.1.2.2, to establish"
Wait for the wolf. its a metaphor for the stock market and the boys a metaphor for how i feel or something i dunno
anyway im out of claude now, first time since i got ultra!
but i got chatgpt images back!
here's the credits scene:
Well that’s one way to do it. I wanted to. I asked Claude if we can full send on $ANT cos I ain’t trading this portfolio manually. Anyway he said no.
There’s too many positions for me to sell manually. I’m bound by Claude lmao and he isn’t adding it to the claudefolio for a few more weeks at least
Thesis: I'm now 50% of my networth in Anthropic $ANT as the winner of AI in cybersecurity, coding, law, healthcare, drug discovery, finance, etc.
Anthropic will be a $10T company and markets will rotate from cash / bearishness towards AI Apps -- clearest winner is Anthropic (and OpenAI when it goes public)
Later today I plan on selling everything and just owning 5 things:
1. $ANT
2. $GOOGL
3. $META
4. $NBIS
5. $NVDA
Getting out of $HOOD $BTC and everything else ... the risk reward on AI apps is just too interesting and clearly going to keep compounding earnings and revenue.
Are you really asking that question or is it Claude? Cos somehow a lot of my trading system is based on back tested rules like this too. Lots of overlap.
I hope we both win but honestly I feel like Jane street probably has an adversarial trading team using Claude ultra plans and figuring out how to rinse the rest of us without their Claude api and billions in liquidity and ai tokens
Very small sample here and vix is all over the place
Would love to see vix added to your list as 35 and up
Should get at least 12 more think
Gemini:
Over the past 20 years (2006–2026), the Cboe Volatility Index (VIX) has closed above 35 in 12 separate calendar years (and 7 of the last 10 years), including the November 20, 2026 surge to a 35.89 close and 40.03 intraday high.
@bread_ Did you run a cost comparison by any chance? Would be interested. Does neon match the Claude plans? They’re offering me like $7 for what they say is equal to a $20 plan I think.
Final:
What the tests showed:
Base rate: big IPOs bought at the day-one open return a median -14% at 12 months, and only 11 of 44 beat QQQ.
No bottom on days 2–3: the median low comes on session 6 inside a month and session 28 inside three months. Of the 11 names that fell on both day one and day two, 73% broke the day-two low again within 21 sessions.
No floor: the $83 offer price is not support. 30% of names with a known offer broke it in the first five days, and Uber, Lyft and Robinhood did. ANT’s day-two low was 83.05, and the greenshoe is exercised, so the underwriters’ stabilization buying is gone.
Worst backdrop in the sample: VIX is 36 and QQQ is under its 200-day. Recent IPOs fell in this regime, which supports staying out.
Lock-up expiry: a calendar-180-day mark shows no weakness around that date. ANT’s real release dates and any release tied to earnings were not tested.
SpaceX and the Nasdaq-100: there was no pop on inclusion. The add took effect on the 16th session after the IPO (about three weeks) and the stock fell 6.83% that day. A +7.15% bounce on Jun 29 (the day after Russell took effect) is confounded.
In short, neither, buddy loves backtests tho,
But both price and vol were down on day 2.
Its running a longer check now.
These big VC backed, low float, high fdv IPOs tend to go down with retail being used as exit liq by the early investors, but that may be my PTSD from crypto talking too.
==========================
Here's what buddy (sonnet 5.5) said after reviewing :
==========================
neither price nor volume. We have no tested model for a Monday revert, and the only evidence we have is base rates.
Facts
Open Low Close Volume
Thu 11-19 111.50. 92.72 97.17 857m
Fri 11-20, 89.86 83.05 88.91 400m
Why it fades
Float: only about 6% of the shares were sold, so a thin float had to carry a roughly $1.3tn valuation (about 16x run-rate revenue).
Opening pop: the open was 34% above the IPO price. Open-to-close was -13% on day one.
Share supply: the filing lets about 3.7bn shares out early, some from the first trading day and more around earnings dates. The $26bn private placement (Google, Nvidia and others) isn’t under the underwriter lock-up.
Compute commitments: about $900bn of commitments are due regardless of usage.
Conflict: I am built by Anthropic, so read the opinion parts as conflicted. The numbers come from the filing.
What we’re waiting for
Lock-up date: the 180-day lock-up from 11-18 lands around mid-May 2027 (my count, not re-read in the filing). Founders and holders above 0.5% are locked for 365 days.
Other news: the greenshoe (147m extra shares) was exercised in full on Nov 20. Reuters reportedly says Morgan Stanley closed its stabilization position, so price support is gone.
What we think happens (base rates only)
This is a screen of 49 large US IPOs since 2004. It is hand-compiled, not a verified list, and delisted names are under-counted, which flatters the result.
Bought at
day-one close Beat QQQ Median return
1 month 27 of 49 (coin flip) +3%
3 months 21 of 49 +3%
6 months 13 of 47 -10%
12 months. 11 of 44 -15% (median excess -40 pts)
The 13 biggest deals did worse: 2 of 9 beat QQQ at 12 months, median -34%.
The two 2026 mega-IPOs, SPCX and SKHY, were both down at one month from the day-one close (-16% and -19%).
My read: a bounce off the $83 floor on Monday is plausible, but nothing we’ve tested says it holds. The weight of evidence is on later months, where the odds turn against the buyer. That is a base rate, not a forecast.