Jay Weeldreyer
I find the actual problem inside complex systems.
I work on consequential systems that are already moving: businesses, markets, organizations, and AI-assisted research.
The visible problem is often not the controlling one. The product is too broad. The process is organized around the wrong constraint. A technical result is being asked to support a decision it did not establish. I work with the people who know each part and turn the answer into something they can test, build, operate, or refuse.
Research, product, and operating work.
Two domains, one function
Research & AI systems
LLMs can generate plausible work faster than anyone can verify the claims attached to it. My current research asks how to keep generated artifacts, claims, and authorized actions within what was actually checked.
See the current research→Companies & commercial systems
A broad vision is not yet a product. I work with founders, technical teams, and operators to identify the real customer problem, operating constraint, or source of advantage and build around it.
See selected operating work→Current research
A fuller account sits behind the link.
When the check passes but the claim does not
A Lean proof can compile even when the broader claim attached to it exceeds what the proof established. Claim Fidelity is a public prototype that makes that difference explicit.
It separates artifact validity from claim support and provides warning and refusal paths when the evidence does not license the requested claim. The next step is an empirical test of what happens under incomplete oversight and selection pressure.
Authorship boundary.I originated every research line and topic pursued in Claim Fidelity. I formed the questions and hypotheses; designed the experiments, architecture, controls, rival baselines, falsifiers, tests, and claim boundaries; read and interrogated the outputs; introduced counterexamples; and continuously redirected the work. Structured teams of LLMs produced the code and substantial draft analysis. I retain responsibility for the research, its adjudication, and what the artifacts license me to claim.
How I work
How I work.
I absorb the system, ask the questions that expose what has been inherited or assumed, and find the constraint that is actually carrying the outcome.
The result may be a product definition, operating process, experiment, decision frame, or control system. The form changes with the problem.
Then it has to meet reality. If the result does not survive contact with customers, operators, evidence, or a stronger baseline, it changes.
Selected operating work
Three evidence-bounded cases.
Translated data science and automated underwriting into products, customer commitments, and operating structures.
- Situation
- Data science and automated underwriting capability had to become something customers and operators could actually use.
- Structure changed
- Data science and automated underwriting were recast as product definitions, customer commitments, and operating structures.
- Jay's work
- Jay worked across product definition, commercial commitments, and the operating structures required to support them.
- Observable result
- The technical work reached product, customer-commitment, and operating-structure form.
- Claim boundary
- This public summary establishes the translation role and the forms produced; it does not state performance metrics or commercial outcomes.
Turned a broad “Fidelity of Crypto” mandate into trade-operations infrastructure, then helped shape crypto and NFT launch-infrastructure and market work.
- Situation
- A broad “Fidelity of Crypto” mandate was too abstract to build or operate directly.
- Structure changed
- The broad mandate became a defined trade-operations stack spanning normalized venue data, backtesting, order management, and routing.
- Jay's work
- Jay shaped the mandate into crypto trade-operations infrastructure and later helped shape crypto and NFT launch-infrastructure and market work at Polyient.
- Observable result
- The mandate became a defined system spanning normalized venue data, backtesting, order management, and routing, with later launch-infrastructure and market work at Polyient.
- Claim boundary
- This summary describes the structure and Jay's role; it does not claim trading performance, customer adoption, or investment outcomes.
Converted evidence into product and ecommerce systems and tested assumptions across a wider manufacturing and distribution system.
- Situation
- Evidence about products and demand had to be translated into product and ecommerce systems inside a wider manufacturing and distribution environment.
- Structure changed
- Evidence was bound to product and ecommerce decisions and tested against manufacturing and distribution constraints.
- Jay's work
- Jay converted evidence into product and ecommerce systems and tested assumptions across the wider system.
- Observable result
- The work produced product and ecommerce systems that could be tested against the surrounding operation.
- Claim boundary
- Jay's role in the wider manufacturing and distribution system was as an observer testing assumptions; this entry does not claim operating control of that whole system or publish commercial performance metrics.
Featured writing
An editorial pick, not the most recent.
Essay2026-06-279 min
The site is not only a personal website. It is also an experiment in making delegated LLM work operational through public, inspectable structure.