Building something in an industry where the incumbents try to prevent progress through artificial rent seeking is interesting. I don't see it working out well long term for a lot of companies in the AI age.
We've heard about it recently in clean room AI coded implementations of major software products, but it exists in data too.
People who have been in technology 20 to 25 years, used multiple languages, and know how to design systems have such an advantage at this point that it’s unbelievable. The problem is they’re looking for a job and more than that. They’re looking for the same job that they used to have. I feel for them but perspectives have to change.
McDonald’s slide: traffic worries outweigh real-estate links
MCD fell 22.3% in the year to Sept. 28, 2026, while SPY rose 14.5%. Slowing U.S. sales and negative guest counts are the clearest concerns; higher rates may pressure valuation and budgets, but real estate data show no matching collapse.
https://t.co/nS4gbVAsJy
@assaf_elovic This isn’t an either or. Think of search and embeddings as getting you into the right place to start and then Jev or other models like that can make decisions or give you relevancy at a specific point.
Search definitely still earns its place in an AI context limited world. Many problems are reducible to give a model just the piece it needs to make a decision from a big document of text. This pattern works well with Jev too. Search, get content, Jev.
NQ test: $100,000 fell to $31,975
That’s a 68% loss, with a 68.8% max drawdown and 0.45 profit factor. Whole-contract sizing stopped trades after August 25, 2021; this does not validate the expected $3.77M result.
I would love to know the discrepancy, if it’s in our data or not.
https://t.co/kRXVySPNe8
NVIDIA’s $211 price exceeds owner-earnings value
TTM operating cash flow was $134.36B and capital spending $7.35B through July 31, 2026. After D&A and stock compensation, owner earnings were $123.43B; at 11% discount and 5% perpetual growth, value was $89/share versus $210.89 market price.
https://t.co/QXEueoLUiD
After a year of building, we soft launched Secmast this week.
Secmast is built around a simple idea: research markets and test strategies without spending most of your time finding, cleaning, joining, and wiring together data.
We have 20+ years of historical data across stocks, options, futures, forex, and crypto, plus SEC filings, economic data, rates, Treasuries, and are building a historical securities master.
On top of that is a research agent that works directly with the data. It can search filings, query large datasets, write and execute scripts, generate charts, and test strategies against historical data.
A few early unknown users have already spent 5+ hours in a single day testing strategies and research ideas, which was a lot more usage than I expected this early.
If you work with markets or financial data, come give it a try.
https://t.co/hsAv7Yrrsc
DM me if you want some extra credits to stress test it.
15 years ago, if you did the responsible thing & bought $10,000 of 'risk-free' long-term Treasuries, you'd have $11,072 today.
If instead you spent $9,995 on drugs & $5 on bitcoin, you'd have $42,778 today.
Volatility surfaces for Nvidia on earnings day and a couple days after. It is interesting seeing how everything is uniformly high on earning day.
A new widget for https://t.co/OKTkCgK9UX. Just ask for a volatility surface for any ticker and day going back 12+ years.
The $NVDA circular financing is a counter attack and brilliant strategy.
- Hyperscalers are all working to create their own chips.
- 40-60% of data center cost are GPUs. Nvidia can provide Ai data center computation cheaper than absolutely everyone else.
- If the businesses succeed, Nvidia has growing investments. If the businesses fail, Nvidia becomes the provider.
- Models and demand are the other parts of the stack. Nvidia will continue to support open models.
- Demand will continue to rise. Companies are turning away from OpenAI/Anthropic towards open models due to cost.
- Companies will go to Hugging Face, select the model for the job, and have it auto deployed to the cloud. Millions (billions) of agents will be deployed that way.
- Nvidia likely has internal numbers showing they can make $$ renting their chips versus selling them outright. The circular financing is a land grab.