We built a flight simulator for public equities. V1 is here. Press play.
We can show you or you can fly yourself; both option doors here: https://t.co/qOT0FNjRT3
The point is not that ours is prettier. It is structural. A language model can describe a causal structure in prose; what it cannot do is BE one.
So scaling it gets you a sharper paragraph, the left side, never the right. The right side is a different kind of object: a causal structure you can interrogate. Ask 'why' and you get a mechanism, not a rationalization, and every number it returns is the image of evidence rather than an assertion. That is the gap a bigger model does not cleanly close.
https://t.co/sy8iELnn4y
Same company. Same investment. Two different machines.
A word model writes a plausible paragraph. A world model builds the probabilistic structure underneath it.
You can tell which is which without reading the labels. This view is not released yet. We are dogfooding it now.
In the meantime, our V1 is live.
Category Error: Every AI tool built for investing was designed for complicated problems.
Complicated means deterministic. There is a right answer:
• Code compiles or it doesn't.
• A bridge holds or it doesn't.
You can decompose the system, solve each piece, and reassemble.
Investing is not complicated. It is complex.
Complex means non-stationary:
• The patterns that worked in 2020 are actively harmful in 2025.
• Variables are causally entangled.
• The act of observing changes the system.
There is no right answer, only probability-weighted scenarios with varying degrees of conviction.
This is not a capability gap. It is a category error.
No amount of scale fixes a category error. A larger language model trained on more data is still a pattern matcher operating in a domain where patterns break. It will produce fluent, plausible, confidently wrong output, and you will not know it is wrong until the position moves against you.
The architecture has to change. Not the model size. Not the prompt. The architecture.
That is what neurosymbolic means in practice: an engine where every output traces through a causal graph, where probability is computed rather than hallucinated, and where you can click through to the evidence chain and see exactly why the system believes what it believes.
Retrieval gives you speed. Reasoning gives you conviction. They are not the same thing.
Both are required.
Btw, it's not just investing that's complex... it's the entire economy.
#neurosymbolic #investing
The trial puts the same surface on your own names. An LLM generates a memo. Primordia computes one. The conviction is yours. Try to break it. It's an early look... DM or comment below for access.
Overloaded, but we'll reply to all, TIA.
If you make fundamental equity investments, pay someone to make them for you, or allocate to those who do, we are opening Primordia to you for 30 days. Full access. No gates, no feature walls, no sales call.
We built a flight simulator for public equities. V1 is here. Press play.
We can show you or you can fly yourself; both option doors here: https://t.co/qOT0FNjRT3
Ask our in-app analysis copilot what would need to be true for the thesis to flip, and it traces the chain to the specific assumptions that would need to change. Every claim links to its evidence. That is what the V1 trailer below shows.
"The polished PowerPoint presentations. The terabytes of information on Bloomberg.... The bacchanal dance to worship the god of information. It was all hot air. The financial crisis touched down and upended" it all. (Dobelli, 178)
@jgreenhall I reference a bit on my pers. site, but forget that; best on-ramp is strategically skim Mythen's book (https://t.co/Vyjz7bBimY). I hear echos of Beck in much of the x-risk discourse (thanks to your work, DS, others), incl. his 1986 text, Risk Society. Ur cmt re: experts prompted.
@OrlandoBravoTB Survival vs. Vanity. "Strategies that allow you to survive are not the same thing as the ability to impress colleagues." (Ed Thorp in AQR Interview, Jan. 2018)