OCTOBER WAS UGLY - monthly thematic rundown in charts 1. Top 10 v Bottom 10 2. z SCORES to show relative context 3. Best since... and what followed (new) 4. Quant Factor Returns (+) 32 other charts in report that I look at * if you want the full report you can download it below // feel free to use **also I asked MS on their drawdown chart making the rounds on X... that is a cherry picked chart from the PEAK IN JUNE through Friday... not just Oct, not the last week, not just Friday (I have some additional thoughts on that I will put out later) report link:

October was ugly ... HISTORIC DE-GROSSING?! dont be hysterical TLDR: last week ... worst performance for systematics in over 2 years (ouch)... L/S fundy guys bad / not horrific... Gross & Nets actually went up overall. I just call balls & strikes.... seeing a lot of these historic de-grossing comments ... different data... but here is my best cut from our internal aggregation and models... was more "rotation" / stock specific selling than panic puking 1/ 25 separate selloffs since 2016 ran deeper. In dollars, 7 were bigger; the worst, to Jul 16 of this year, lost ($32.8B) 2/ No volume spike (forced deleveraging shows up as heavy trading) Tech traded at 1.02× its pace over the prior year. In March 2020 it hit 2.08× 3/ it was a rotation. Tech bled −$11.0B while the rest of the market took in +$29.4B. Consumer & Retail saw stronger buying than in 99% of its windows since 2016. Internet Platforms and search sold harder vs the market than in 99% *going to release our "GICS" competitor this week ... I think better and more informative tbh 4/ $NVDA (−$5.0B) and $GOOG (−$3.6B) led the selling. Buyers/COVERS took $AAPL +$1.6B

man ... what a night for EPS ... *check your AI factor exposures for the morning $INTC CEO spooking people on demand slowdown comments to match the guide $AMZN guide $TWLO / $RDDT / $PINS blahhhhhh

$GOOG bears v bulls split this am. Mix shifts are ALWAYS where high idio ALPHA is. Every year I start off by asking sell side and buy side friends … what 1-2 names have a mix shift controversy right now? hard to model. Controversial. Exact area to dig into for a L/S investor This is likely the start of the 2nd “search is dying” bear fight. Last one that got ugly was 2015. Slowing high margin growth mix shifting into stratospheric Cloud / AI growth + margin uncertainties turns this into a battleground stock

10/27: squeezy L/S day ... more debt deals ... semis v software bouncing... shorts being covered into FED dovish fed leak? hmmmm 1. Risk on vs. Risk off 2. SHORTS ripping 3. Semis v Software (rebound app) - coming into today in its worst RELATIVE drawdown ever (+5% today)

⚒️open sourcing my own TOKEN Framework (+) pre-lim $ANT numbers for everyone to do their own forecasting ⚒️ TOKEN FRAMEWORK (below) - consolidate all my thoughts from last 12m & tried to organize into a more coherent framework for how I am thinking about it (+) linking all my sources.. 3 important things imo 1. "Count Dollars not Tokens" 2. Pricing power is real but narrow *YOU have to have a view on Token Growth and Open Weights SHARE OF DOLLARS to have conviction ... On Vercel it went from about 5% to 23% in three months, led by pricier open models such as Kimi K3, not the cheapest ones. If that continues, we're in Bear: open-weight takes about 57% of 2030 API spend and frontier-lab spend only doubles. If it stalls, Base holds: about 24% open, about 10x for closed labs. That's a $700B difference in frontier-lab revenue by 2030 **I am sure the hive will find some errors -> good... send them and will update $ANT : 5 things you need to believe .. the CORE bet is #3# and #5# for the true believers - adjusted "RRR" waterfall for S-1 disclosures (this is run rate revenue and not ARR tbc) *"NRR adds" a black eye -> I would really focus on this with mgmt in IPO meetings ... really understanding the headwinds and tailwinds - my true est. SHARE COUNT (wow) - profitability bridge to help me understand moving pieces - Funding Traces - **BULL v BEAR** what you all want TOKEN MODEL: ANT_ARR_TOKEN_MODEL: Both of these are meant to encourage debate + for you to stress test your own assumptions and add to what I have here - needless to say nothing here is investment advice and you agree to that stipulation before opening any of the models - I will be tweaking and updating my own daily so can change without notice

"wow its flat in the premkt" - PM friend one CHARACTERISTIC of this changing market is how fast things get discounted... ever since COVID... more alt data + X news + other tools + more systematics ... the market just discounts events faster what does that mean exactly? 1. if you wait for the catalyst you are too late 2. you have to think about the path more 3. variant views will get you PAID ... you have to put on risk early and be there if you have conviction ... think $SMCI in 2024... Memory... etc Now the inverse is also true.. think Software / Cyber when doomers were all over them for 18 months... when things are going against YOU HAVE TO RESPECT volatility ... things can alot further than you think... if you cope on valuation you are going to get wiped out ... TLDR: be early. be right. scared money dont make money.

black mamba work ethic legendary. Not a single successful person I know who isn’t a beast with their schedule. They pay the toll. Tbc… that’s not just sitting in front of screens all day for investors…. Mornings, food, workouts, quiet time, nights, alcohol, reading, music…. All a purpose. Methodical. Carrying the boats. Greatness demands sacrifice. It’s not for everyone.

What a shift in last 6m … what’s the broader impact of these trends? Need to think on it more but 5 random thoughts with kids breakfast … - DC : data center was the $$ source of gravity for pre training … a) one shot training not a moat anymore b) spend shifting to RL environments and CPU/memory .. this should stay steady here - POWER / electrification: continues to stay scarce either way (love this factor on current drawdown) - distillation is table stakes: pretraining data gets less differentiating once base models are good enough to distill - where is puck going? environments, expert traces, tool logs, unique data and reward signals. That lifts vendors and internal orgs that manufacture RL tasks - APP layer: benefits twice… models get cheaper as the lagging tier commoditizes AND the frontier keeps moving to long running agentic tasks…so the product that owns the workflow and the feedback data matters more than the product that wraps a base model

lots of discussions about filing tools and updating EXCEL sheets with AI... my hottest take...imo.... I think working on these are a waste of time...kinda like building skills for models right now ... whats the famous elon quote? "engineering is the closest thing to magic that exists in the world...the most common error of a smart engineer is to optimize a thing that should not exist" "skills.md" ....filing extractors... excel models... risk models... all of these will be abstracted away in the future ... the future is going to be crazy in finance

🚨 DEWEY - RISK MODEL in Spotlight 🚨 one of my favorite characters from a series I read a long time ago was an operator named Dewey Andreas. tough. smart. never quit. he was "the guy" everyone called when they had an unsolvable problem... welcome to "DEWEY" our latest risk model. the big commercial risk models start from a fixed list of named factors... lets call it ~60 industries and a dozen styles, because a person has to read them... statistical models have been around for decades, but their output is extremely hard for a person to interpret... agents don't have that problem... so RE-THINKING RISK IN AN AGENTIC WORLD ....I built v1 of Dewey, a new type of statistical risk model that lets the data decide how many themes the market actually has, and says out loud where it can't tell *we have worked really hard to expand the security master to about 5,000 names now **completely re-did our QUANT factors + added some new ones (check em out!) see descriptions for deet's ***we are now running our "GICS" internally built labelling system - I was offended by what these companies wanted to charge - so I built our inhouse version ... I think its better / more intuitive with more descriptive layers ... IF you pay a ton of money to MSCI or the other vendors let me know we can work something out ... point in time labels 2007-2026 to start DEWEY: in the video I pull up "/NVDA" in Spotlight and open the new risk view. nobody told the model what themes to look for. it refit last night on 2,900 stocks and years of returns, and NVDA's two biggest drivers are the market and a Low Quality vs. Low Volatility theme, each carrying roughly a fifth of its variance I call "NODES" the Paul view... in honor of my friend @pmje73 .. look at it isolated or vs another another from the drop down new - "Offsets"... the market accounts for about half of NVDA's variance, so over that history, shorting about $172 of the market per $100 of NVDA would have taken its volatility from 40% to 28% a year. Mag7 shows up greyed out, because Mag7 holds NVDA and you'd be offsetting NVDA with NVDA what makes it different is that DEWEY knows/LEARNS where it stops knowing things... every night it measures the point where the data can't tell a real theme from noise, and anything below that line never gets sent to an agent.. because an LLM will happily explain pure noise if you hand it some. every number comes with a range and every sentence comes from a fixed template, so whether you ask Claude or GPT, the number doesn't change for the propeller heads, including where it's weakest... read more here & details on tests we ran against open source baseline risk models: this is a critical FIRST step to our BIG REVEAL coming later in week... stay tuned... have a good week 🍪🥠🍪🥠

Oh boy… ok … it’s SO EASY in the fast tweet world to read something like this… retweet or repost … and take it as gospel It is the MAIN thing that makes me mad about AI is all high level critical thinking just disappears Let’s take the below … *9% is the gross rev a 2023 Columbia model says US data centers need in 2032 to earn a 10% return tbc **food at ~9% is households buying groceries and meals….the AI num is operator revenue, including foreign customers and firms buying AI as an input….those sales do not stack on GDP the way food does ***the computer / peripherals chart could be top 10 chart crimes in fintwit history… also actually bullish AI 🚨🚨 I’m not sure what the number is yet… but stuff like this is counting the @Uber TAM by taxi cab medallion costs.. wrong