Kevin Riley

Building an interface between computation and matter.

[Kevin Riley @kmrNYC]

@DrEliDavid

[Kevin Riley @kmrNYC]

@politico Politicians can't "pass the bill to find out what's in the bill" anymore -- LLMs can immediately identify all the pernicious non-sense they embed in hundreds of pages that they used to be able to hide. They always need total narrative control, hence AI is "dangerous" for *them*.

[Kevin Riley @kmrNYC]

@GovPritzker These people can't "pass the bill to find out what's in the bill" anymore -- LLMs can immediately identify all the pernicious non-sense they embed in hundreds of pages that they used to be able to hide. They always need total narrative control, hence AI is "dangerous" for *them*.

[Kevin Riley @kmrNYC]

It also seems that AI has fully stopped congress and the lobbyists who actually write the bills. They can't "pass the bill to find out what's in the bill" anymore -- LLMs can immediately identify all the pernicious non-sense they embed in hundreds of pages that they used to be able to hide. They always need total narrative control, hence AI is "dangerous" for *them*.

[Kevin Riley @kmrNYC]

Volatility harvesting (Shannon’s Demon) requires highly volatile uncorrelated assets which rarely exist in liquid markets — it moves money out of winners into losers and will fail in a trending market. Cover’s actual algorithm tends to move money into the last known winner, a follow the leader approach.

[Kevin Riley @kmrNYC]

I’m looking to connect with people who have deep practical experience at the intersection of electrochemistry, chemical process engineering, and separations.

We have a proposal for a programmable chemical-process platform designed to control aqueous chemical state and selectively move matter through a sequence of useful states, thus allowing difficult, variable, or waste-derived feedstocks to be converted into purified resources, chemicals, materials, or biologically manufactured products.

We are looking for a partner or team to run a focused proof-of-concept experiment that will validate or rule out the core physics and, if it validates, support subsequent process development. Prior work in electrochemical process development and separation and purification is relevant to aspects of our proposed design.

At this stage, I am particularly interested in people who have actually designed, built, and operated experimental process systems. Experience with electrochemical separations, electrodialysis/bipolar membranes, aqueous speciation, membrane processes, hydrometallurgy, selective recovery, or coupled reaction/separation systems are especially relevant.

The immediate objective is deliberately narrow: build the minimum experimental system necessary to determine whether the proposed architecture works as intended, understand its limits, and establish the engineering basis for scaling it if the results warrant doing so.

Please reach out if this is for you.

[Kevin Riley @kmrNYC]

@zarathustra5150 “The caveats I’d keep clustered rather than salted through the above, because each is load-bearing.” — Claude analyzing a research paper. Is this nonsense the new watermark?

[Kevin Riley @kmrNYC]

@SenSanders So instead of @JeffBezos spending his equity, @SenSanders gets to spend the equity. What problem does this solve?

[Kevin Riley @kmrNYC]

@theobjectivist @naval Their objective is to seize assets. Engaging in a discussion about “how that will work” is a fool’s errand. Capitalism is not failing; socialism is failing. And that is why they are desperate to take more.

[Kevin Riley @kmrNYC]

“The caveats I’d keep clustered rather than salted through the above, because each is load-bearing.” — Claude analyzing a research paper. Is this nonsense the new watermark? Is this why it’s becoming impossible to parse anything Claude writes?

[Kevin Riley @kmrNYC]

@mcuban @SaraForTexLege There can be no price discovery when no one knows the prices. This is not a radical policy; just publish the prices.

[Kevin Riley @kmrNYC]

I asked Fable to analyze my existing project files -- nothing new -- and got kicked back to Opus. @AnthropicAI

[Kevin Riley @kmrNYC]

@DeepLearn007 @elonmusk I agree this is inevitable. But it also may require re-thinking markets, money, and society beyond the idea of UBI — the transition could be difficult for many people and the second order effects of UBI in the economic current structure may not do what people think.

[Kevin Riley @kmrNYC]

@investingidiocy Non-stationarity aside, an LLM may lie to the user about backtest results since it has seen the past market regimes in its training data and will use that in its reasoning. These things are already known to have terrible out of sample performance.

[Kevin Riley @kmrNYC]

@AlphaEncoded @quantbeckman Prediction target is next day percent return.

[Kevin Riley @kmrNYC]

Indeed. As a tool, AI can 10x (or more) your ability to iterate. Too many people are viewing this is static terms. AI may be more like the Industrial Revolution; wealth grew while agriculture became a relatively small component of GDP — and there were ideas back then that humans would be replaced by machines.

[Kevin Riley @kmrNYC]

@WSJ Self-dealing much?

[Kevin Riley @kmrNYC]

@quantbeckman Indeed. Building from the bottom up reveals what patterns are actually in the data and the statistics of those patterns.