steve kim

@JaneMcNally_ Jones is about Donato's age? I remember seeing him play in the 80's.

Wild turnover for ECAC coaches recently: 2026- Donato 2026- Whittet (Brown) 2025- Alain (Yale) 2025- Schafer (Cornell) 2025- Jones (Clarkson) 2025- Smith (RPI) 2024- Fogarty (Princeton) 2023- Vaughan (Colgate) 2022- Bennett (Union) 2020- Gaudet (Dartmouth) 2019- Morris (SLU)

@citrini #3 is now #1

Analyst #3 is now safe and sound back in the free world 🫡.

Strait of Hormuz: A CitriniResearch Field Trip The Field Report from Analyst #3 is live. citriniresearch.com/p/strait-of-ho…

@garrytan I've tried and it is not very stable. Claude Code can build routines that are more consistent.

I'm telling most everyone I know that they should build a personal OpenClaw

@chamath What about Trump running the country to the ground and now got us stuck in a war that he can't get out.

California is going bankrupt before our eyes.

@BillAckman you supported the lying pedophile nut. good luck.

By far the best piece I have read on U.S. strategic options in Iran.

@JaneMcNally_ @chnews I feel the prior teams were more ready to block shots, which takes a new level of sacrifice. That really helped Ian Shane.

Cornell was 17-5 on Feb. 7. After that? 5-6-1. "I thought we might have peaked maybe a month too early.” 22 wins in Casey Jones’ first year is a pretty good feat. But brush aside the positives and there is even more to build upon. For @chnews: collegehockeynews.com/news/2026/03/2…

@danielsethlewis @BuffaloBills How many Bills fans are in Nashville besides you?

Wife: "Something bit your Bills Mafia sign out back -- a dog or raccoon prob” Me: Wtf?!? Puka Nacua-- Noooo!! @BuffaloBills

@JaneMcNally_ Do you know if Devlin’a playing? Team seems to play better with him on the ice.

Puck drop 6 ET. Only on ESPN+.

BRACKETOLOGY: Final bracket is as straightforward as they come Albany Region • Michigan (1) • Minnesota Duluth • Cornell • Bentley (AHA autobid) To read full article go to uscho.com/2026/03/22/bra…

@findatasets What are the adjustments?

Pull historical price data in 1 API call. Daily OHLC, volume, and adjustments. 14,000+ stocks. 20+ years.

@TailThatWagsDog Out of sample?

Much respect to these women👍👍👍 Yes, your president is a big liar and a fraud who abuses his position to insult and bully his fellow citizens. Still, I believe you must have shown a great deal of heroism in making this decision. Thank you for that👍

@JaredKubin ChatGPT pro is very good.

@elonmusk When?

SpaceX will build a system that allows anyone to travel to Moon. This will so insanely cool 🚀💫🤩

@AyusoValue DCF is easy to build. Trick is in getting the parameters and assumptions right, and how can one trust an LLM black box to do that?

I’m honestly blown away that Claude integrated into Excel can build a full DCF model in 10 minutes just by answering 4 questions. A lot of tasks are becoming pure commodities, if you don’t keep up, you’re done.

@VuikoBerkut @FundamentEdge Things like compustat. AI agents can get this sort of stuff from filings. Options data is needed by a niche group and most don’t care.

@SteveKim1234 @FundamentEdge This. Public company data is free - what “proprietary” / licensed data are people blabbering about? Some hyper-niche options pricing feed?

What people miss about the financial terminal oligopoly is the true moat of these businesses is an aggregation-disaggregation moat. These businesses are essentially group purchasing organizations for financial data. Bloomberg, for example, spends an estimated $3bn annually on acquiring, producing & curating data then slices that cost across 325k terminal users at ~$30k/seat. No individual buy-side firm could replicate that content stack on their own...not even close. In my opinion, AI actually strengthens this moat, not the other way around, as AI allows an explosion in data driven decision making broadly. AI needs data. The raw data still has to come from somewhere...exchange feeds, proprietary collection, licensed content, expert networks. Can some of this be scraped by AI? Sure, but not most. On a more practical basis, S&P and FactSet have been liberal in licensing their data feeds to LLMs, and Claude Finance connects this via a (brittle) MCP server, whereas Bloomberg has decided not to API their data to labs/co-pilots. What if S&P and FDS management wake up and reverse that decision? Where does that leave the finance co-pilot space? Cooked. As cool as Claude is for certain use cases, with this massive data disadvantage, it can't begin to touch the data aggregation moat of the terminals.

@Jason @matiii @elevenlabs I compared it to GoogleLM, and Google was more natural sounding.

@matiii @elevenlabs 11L is dope

Today, @elevenlabs is announcing a $500M Series D at an $11B valuation, led by Sequoia, with a16z quadrupling down and ICONIQ tripling down. It reflects the trust of customers and partners building at the frontier alongside us - and gives us momentum to ship even faster.

@GfI_Himmelreich @P123Finance Are strategies in the book at fixed dollar allocated, or rebalanced regularly?

@SteveKim1234 @P123Finance 20% hedge via a short Russell 2000 position

90% of Quants (including Institutions!) underestimate the power of Cross Factor Mean Reversion, you need a Portfolio of Strategies (in @P123Finance "Strategy Book") to master markets: The Power of Simple, Structural Diversification It’s tempting to chase complex signals and perfect timing. But some of the most robust portfolios come from a simpler idea: Diversify structurally — and rebalance with discipline. (Think: cross-factor mean reversion at the portfolio level.) Our “One Portfolio of Free P123 Strategies – Hedged” is a great example. It is live OOS since Oct 2024. The underlying strategies have OOS live histories ranging from decades to ~3 years How it works (no factor timing, no tactical shifts): We combine several standalone, rules-based engines — e.g. Built for Stability, Expecting Dividend Growth, Higher Beta + Earnings — and add two structural diversifiers: - 25% Gold ETF - 20% hedge via a short Russell 2000 position Why this matters: The strategies are chosen for low covariance — they behave differently across regimes (stability vs. high beta, equities vs. gold, long large-cap vs. short small-cap). They don’t all win and lose at the same time. The only “timing” rule: Set static strategic weights once → then rebalance back to target weights periodically. That simple rule systematically forces: ✅ trim what ran hot ✅ add to what lagged → “buy low / sell high” across the whole book. The result: Any single strategy might look average. But the portfolio structure can deliver a better risk-return profile: smoother ride, smaller drawdowns, and better compounding. The lesson: it’s not about predicting markets — it’s about building a resilient structure.