🚨 *FED PROJECTS 2 MORE 2027 RATE HIKES *FED 2027 MEDIAN FED FUNDS 4.1% (+0.5) *FED 2027 MEDIAN CORE PCE 2.5% (UNCHANGED) *FED 2027 MEDIAN PCE 2.3% (UNCHANGED) *FED 2027 MEDIAN UNEMPLOYMENT 4.1% (-0.2) *FED 2027 MEDIAN GDP 2.4% (+0.1)

Last wk, S&P/Nas/Mag7 +0.4%/+0.1%/-0.8%. Cooler inflation (CPI, PPI) & economic data (consumer sentiment, retail sales) but +5% oil steepened the yield curve but lowered odds of a rate hike. Looking forward, I continue to believe the impact of Agentic AI with the advent of OpenClaw on January 30th has at least a year to run: 1) Token production has gone up roughly ~7.5x from the end of January more than offsetting the nearly 50% token cost reduction seen since open-weight model usage started to take off in May. 2) Combined annualized run-rate revenues for OpenAI and Anthropic which ended last year at $29B seems to be around $100B currently with Anthropic getting profitable in Q2. 3) Capex from the Big6 hyperscalers accelerated from 84% y/y/ in CQ1 to 92% in CQ2 with forecasts for nearly 100% in Q3. But this is being supported by cloud revenue growth at the 3 Big Public cloud vendors of $AMZN $MSFT $GOOGL accelerating from 23% y/y in Q1:25 to 35% in Q1:26 to 43% in Q2:26. Arguable more important is public cloud operating margins expanded from 34% to 37% and 39% during those time periods. 4) The $500B financing deal backstopped by up to $125B from $NVDA adds even more lower cost money to fund AI capex spend for the non-hyperscaler players. Nvidia gained 0.5% last week. 5) The liquidation of Situational Awareness and retail accounts during July cleared out some of the frothiness in the AI related names In terms of negatives: 1) The cost of money (yields on government bonds) remain near the highest levels for the 30 yr tenor at 5.3% since 2007. 2) Given large scale offensive US military actions are seemingly off the tablein favor of financial sanctions, probably driven by current election polls, I now believe Iran is likely to hold the Strait of Hormuz hostage until past the US mid-terms. This would be akin to them releasing the US hostages in 1981 (they were held for 444 days) just hours after President Reagan was sworn in replacing Carter. There were severe financial sanctions then also. 3) Since 1990, which happens to be the Gulf War, from the end of July through November 9th, which covers the reaction to all mid-term results, the performance is worse than non mid-term years. For mid-term years the median S&P500 gain from 7/31-11/9 is 0.9% with gains 56% of the time but the median peak loss from 7/31 is 6.2% (intra-period median peak loss of 9.9%.) For non mid-term years the median gain is 2.7% from 7/31-11/9 with gains 59% of the time and the median peak loss from 7/31 is 3.5% (intra-period median peak loss of 5.2%.) This year with the momentum seen by the Socialists which are not big business friendly, I see more risk than normal. 4) The easy money on the AI technical rebound from oversold levels on 7/29 due to the forced sale by Situation Awareness is probably over. There were negative stock reactions to headline beat and raise earnings on both revs & EPS for AI infrastructure winners $CSCO (-8% for the week but still up +45% YTD), $AMAT (-6%/+97%) and $COHR (-14%/+77%). While negatives can always be found, their biggest crime was arguably their recent bounce from 7/29-8/7 of 8%, 24% and 71% respectively and their market beating YTD gains. In summary, I remain bullish. Even from the end of July through November 9th during mid-term years since 1990, the S&P has an additional median gain of 4.2% to its peak before giving some of that back closer to the election. Given some of the negatives, especially the reaction to solid earnings data, I would add some hedges back on further market gains and get more selective. Consumer discretionary hedges should also make sense if oil is higher for longer. I believe value should continue to accrue to the infrastructure layer which includes 1) the public cloud vendors such as Amazon, Microsoft, Google and 2) the semiconductor companies. $INTC, my favorite semi company, still gained 0.8% last week despite: 1) a $20B equity offering which causes ~5% dilution and 2) being up 178% YTD. This clears the funding overhang. All the best in the week ahead.

Last wk, oil +9% & ylds +11-26 bps across 2/30 curve w/ S&P/Nas/R2K -0.3%/-0.7%/-2.4%. This wk, I am watching reaction to 1) oil/rates, 2) calls to slow down AI development & 3) Fed on 9/16. I remain on the cautious side till US mid-terms on 11/3. This weekend, the CEO of Anthropic called for a slowing of frontier model development over safety concerns. This follows comments along similar lines by the CEO of OpenAI to employees last week if other companies were willing to do the same thing. The fundamental issues I have with this is 1) foreign adversaries would welcome the US slowing down AI development, 2) I view this as an attempt to slow down open-weight model development which would help the market dominance of OpenAI and Anthropic which are currently in the lead and 3) I do not see other companies agreeing to anything that slows down progress catching up to these two market leaders. Having said that, I could see 3rd party evaluators to limit liability risk going forward and some sort of executive order from the White House. But I hope the longer-term result of these actions is broadly distributed personal AI capabilities for all individuals versus having it become concentrated in the hands of a few companies. Along this vein of AI competition, after releasing their paid API of Muse Spark 1.3 two weeks ago with open-weight versions coming later, $Meta launched their personal AI agent Muse last week with the stock gaining 5%. With 3.6 billion daily active users, a hit product could yield large results. Meta is increasingly showing other ways they can monetize their AI capex spend. This should help the stock to re-rate from a 17x CY27 PE to a multiple closer to peers trading in the low 20s. Meta Connect on September 23–24 is another potential catalyst given their leading frontier model Watermelon should be coming at the latest by October. On the front of broadly distributed AI capabilities, $AAPL stock gained 4% last week on their new product launch. The foldable Duo will provide a personalized AI agent in your pocket with a 50% larger screen than a Pro Max. I continue to see a big upgrade cycle next year. The change from a 4” screen to 5.5” screen with the iPhone 6 drove revenue growth from 7% in FY14 to 28% in FY15. The Android ecosystem has had a foldable Samsung phone since 2019. As for the Fed on Wednesday, I believe Warsh will raise by 25 bps and echo his hawkish statements from Jackson Hole on August 28th that “Price stability is not self-executing… 65 months of sustained, elevated inflation sits squarely with the Central Bank.” The ECB statement last week when they hiked might provide some hints: “For inflation excluding energy and food, the baseline foresees 2.5% in 2026, 2.6% in 2027 and 2.3% in 2028. Compared with June, the baseline projection for inflation in 2026 is unchanged, while it has been revised up for 2027 and 2028… The outlook remains highly uncertain, with risks to the upside for inflation and to the downside for economic growth.” In summary, my caution between now and the US mid-terms on 11/3 remains for reasons I have fleshed out in prior posts including: 1. Don’t fight the Fed: The market historically under-performs during a hiking cycle with the bond market discounting 2 raises by year-end and 3.5 raises by mid-June of 2027. 2. Seasonal headwinds: September is down -0.5% on average and up only 48% of the time since 1957. 3. Historical volatility: S&P drawdowns of 10% between 7/31 and 11/9 have occurred in the lead-up to mid-terms since 1990. 4. Regulatory friction: There is bipartisan pushback against datacenter expansion that could hurt the AI buildout in the near-term. 5. Geopolitical risk: Despite US efforts to de-escalate, I believe Iran drags out hostilities at least through the 11/3 US mid-terms, keeping oil prices elevated. 6. Macroeconomic pressure: Long-term government bond yields are hitting multi-decade highs for several countries, slowing down growth and providing a reasonable alternative to stocks. I believe in not fighting the Fed, the bond market or seasonality. I like the odds stacked in my favor which should improve at least seasonally following the mid-terms.

During Tesla Cybercab launch week in Austin, I took a total of 12 paid Cybercab rides and kept every receipt. TOTAL: • 12 rides • 29.7 miles traveled • $108.34 total spent • $3.65 average cost per mile • 2 hours and 19 minutes riding around Austin in a Cybercab • Compared to Uber and Waymo, every Cybercab ride was still ~30-50% cheaper EACH OF MY CYBERCAB RIDES: 1/ Sept 3 — 7:23 PM 3.9 miles · 16 min · $8.70 ≈ $2.23/mile 2/ Sept 3 — 7:53 PM 0.5 miles · 6 min · $4.65 ≈ $9.30/mile 3/ Sept 4 — 6:52 AM 2.5 miles · 12 min · $6.99 ≈ $2.80/mile 4/ Sept. 4 — 7:13 AM 2.4 miles · 11 min · $6.84 ≈ $2.85/mile 5/ Sept. 4 — 7:39 AM 1.8 miles · 11 min · $6.16 ≈ $3.42/mile 6/ Sept. 4 — 8:03 AM 1.4 miles · 8 min · $5.69 ≈ $4.06/mile 7/ Sept. 4 — 8:20 AM 3.3 miles · 15 min · $7.90 ≈ $2.39/mile 8/ Sept. 4 — 10:58 AM 4.0 miles · 19 min · $8.78 ≈ $2.20/mile 9/ Sept. 4 — 11:58 AM 2.6 miles · 8 min · $7.08 ≈ $2.72/mile 10/ Sept. 4 — 2:17 PM 2.8 miles · 11 min · $13.71 ≈ $4.90/mile 11/ Sept. 5 — 12:57 PM 2.4 miles · 10 min · $16.18 ≈ $6.74/mile 12/ Sept. 5 — 1:36 PM 2.1 miles · 12 min · $15.66 ≈ $7.46/mile One of the biggest things I learned from actually using Cybercab is that you cannot judge the economics from just one ride. Pricing is dynamic. Some of my longer rides came in cheaper around $2.20–$2.40 per mile. Some of the shorter or higher-demand rides were much more expensive. Sometimes, the Cybercab was even more expensive than the Model Y (Unsupervised) bc the demand was so high. Some of the rides also became more expensive as Cybercabs became bookable on September 4 to the general public. However, it’s hard for me to look at these receipts and not wonder how Uber and Lyft can compete if Tesla can scale Cybercab the way it wants to. Uber and Lyft were built around connecting riders with human drivers. That means every ride has to support a much bigger cost stack... like 1/ A human has to be paid. 2/ That person needs a car. 3/ That car has to be financed or purchased. 4/ It needs insurance. 5/ Maintenance. 6/ Fuel or charging. 7/ And then Uber or Lyft still needs its cut. Cybercab attacks the biggest cost in that entire system which is the driver... there is no driver waiting to be paid at the end of the ride. And Tesla is supplying the software. Tesla can build the car. Build the battery. Build the autonomy system. Operate the network. Control the app. Control the charging. And the team designed the Cybercab from Day 1 around being a Robotaxi instead of taking a normal consumer car and trying to make the economics work. This is a big deal. Bc if Tesla can manufacture these at huge scale, keep them on the road for most of the day, and keep bringing the cost per mile down, then Uber and Lyft are going to be competing with the company that built the vehicle, the driver, and the network ALL AT ONCE. And this was during launch week... with just ~45 Cybercabs, before this vehicle is anywhere close to mature scale. If Tesla can push that cost down further as the fleet grows to thousands and millions, while Uber and Lyft still have to pay a human being to sit behind the wheel… I really don't see how these companies can survive... I think the days of Uber and Lyft are limited.

August @NewYorkFed Services PMI up +0.5 vs. +8.7 prior … business climate -25.7 vs. -24.0 prior; employment +2.4 vs. +4.3 prior; wages +30.7 vs. +30.7 prior; prices paid 70.1 vs.66.7 prior

The 4 BACKTESTING TECHNIQUES behind WINNING Strategies: i've spent the last 2 years running backtests on everything from mean reversion setups to volatility arbitrage to prediction market signals some strategies survived and tbh most of them died and the difference was never the strategy itself, it was how i tested it a backtest is not proof your strategy works, it's a stress test to see how easily it breaks here are the 4 techniques i've actually run, what worked, what broke AND what i still use --------------- technique 1: standard in-sample / out-of-sample split verdict: broken by default, NEVER TRUST THIS the setup is easy, take 5 years of data and train on the first 4, test on the last 1 the problem is subtle - every time you tweak the strategy and re-run, you're peeking at the test data and after 30 iterations your "out-of-sample" is FULLY contaminated the first strategy i ever backtested was a simple pairs trade between two energy stocks that showed a Sharpe of 2.1 on the standard split, so i deployed $2,000 of my own money and lost 40% of it in 3 months going back later i realized i'd re-run that backtest 47 times during tuning, the test data was never really untouched use this only for a quick first look, NEVER as the final validation --------------- technique 2: walk-forward validation verdict: the real workhorse, this is what i actually use instead of splitting once, you slide a window through the data train on 2018-2020, test on 2021 train on 2019-2021, test on 2022 keep sliding each test window is data the model has never seen and you get 5 or 6 test periods instead of JUST ONE what this catches: > strategies that only worked in one regime (the pattern shows up immediately) > parameters that shift wildly when retuned (unstable strategy, red flag) > strategies that survive across every window (this is real edge) at our fund we killed a stat arb strategy that showed Sharpe 2.4 on a standard split, but walk-forward revealed it worked beautifully in 2019-2020 and completely died in 2021-2022, the regime had shifted underneath us and it saved us months of losses but this is slower and more painful than a standard split and it's also the reason institutional backtests match live P&L :) --------------- technique 3: purged k-fold cross-validation verdict: fixes a hidden bug in walk-forward financial data has memory, today's price is not independent of yesterday's when your training window ends on december 31 and your test window starts january 1, information leaks across that boundary and your Sharpe looks better than it should purged k-fold fixes this, Marcos Lopez de Prado covers it in Advances in Financial Machine Learning the idea is simple: > split data into folds like standard cross-validation > when a fold is used for testing, remove the adjacent observations that overlap in time > this eliminates the leakage a QUANT friend of mine who runs an ML-based factor model showed me his numbers before and after adding purging, Sharpe dropped from 1.9 to 1.4 on the same strategy with the same data and the extra 0.5 was pure leakage he didn't know he had use this when you're training ML models on financial data, the leakage in tree-based models is brutal without it --------------- technique 4: monte carlo trade shuffling verdict: the reality check that saves capital EVERY SINGLE TIME your backtest shows one sequence of trades, Monte Carlo randomizes the order and runs it thousands of times why this matters: > your backtest might have gotten lucky with sequencing, what if the drawdown happened in month 2 instead of month 10 > the max drawdown you observed is one path, Monte Carlo shows the full range > the 5th percentile drawdown is often 2 to 3 times worse than what you saw few months ago (during the hype of 15-min BTC markets) i built a systematic prediction market strategy that showed 12% max drawdown across 18 months of backtest, but before deploying i ran Monte Carlo with 10,000 shuffled sequences and the 5th percentile scenario showed a 34% drawdown ofc i didn't deploy at full size, i sized it at 25% of what i originally planned and three months in the strategy hit a 22% drawdown, but the smaller size meant i could hold through it and the strategy recovered to finish the year up 31% Monte Carlo is why i stayed in that trade instead of blowing up --------------- what i actually use in production NOW: > walk-forward validation as the primary test > purged k-fold when the strategy uses ML models > Monte Carlo shuffling on strategies that survive both, before any real capital > standard in-sample/out-of-sample only for the very first pass (rarely tho) if your backtest is designed to make you feel good, it's designed to LOSE you money for real.

September @NewYorkFed Services PMI slowed to -8.7 vs. +0.5 prior … business climate -35.9 vs. -25.7 prior; employment -4.9 vs. +2.4 prior;  wages +30.3 vs. +30.7 prior;  prices paid 69.0 vs.70.1 prior

HERE'S WHAT THE 🇺🇸 STOCK MARKET LOOKED LIKE TODAY Up: - Bloom Energy $BE +10.8% - Lumentum $LITE +5.7% - Applied Materials $AMAT +5.2% - Corning $GLW +4.7% - Marvell $MRVL +4.5% - Oracle $ORCL +4.0% - KLA $KLAC +3.9% - Meta $META +3.2% Down: - SLB $SLB -3.1% - United Rentals $URI -2.8% - Texas Pacific Land $TPL -2.8% - Medtronic $MDT -2.7% - Apple $AAPL -2.7% - Illinois Tool Works $ITW -2.4% - T-Mobile $TMUS -2.1% - Walmart $WMT -1.8% Meta recovered about two-thirds of Monday's -4.8%. Apple and Qualcomm fell for a second straight day. Utilities went from lower on Monday to the best sector in the market today.

Considering the speed and magnitude of the move higher in interest rates the returns for US Treasuries this year really aren't that bad: 2 year +0.2% 3 year -1.0% 5 year -2.9% 7 year -4.1% 10 year -5.2% 20 year -7.1% 30 year -8.1% These are really bad days in the stock market