This site is not only a place to read about Jay. It is a public instantiation of a git repository shaped to be used and operated with LLMs: typed content, plain-text mirrors, generated machine-readable artifacts, an audit log, and source-controlled change.
The current surface is simple on purpose. Most useful LLM work still happens through text structures: Markdown files, folders, manifests, prompts, instructions, logs, and links. The site exposes those structures as web artifacts that humans can inspect and models can fetch without scraping.
Current surfaces
- /llms.txt is the site index for language models.
- /llms-full.txt is the writing corpus in one plain-text file.
- /capabilities.txt describes the current public capability surface.
- /capabilities.json exposes the same surface as structured data.
- The Work, Writing, and Done entries each have a plain-Markdown mirror. Append
.mdto those URLs.
What this is for
Use the site to do more than summarize Jay. Use it to:
- determine whether there is natural alignment around a real problem;
- collect distinctions, language, examples, or operating patterns that are useful elsewhere;
- inspect how the public artifact is maintained;
- reuse public structures as starting points for other LLM-operated work;
- route a specific opportunity toward contact when there is enough signal.
The current boundary
This is a public read and routing layer. It does not authenticate agents, delegate authority, accept transactions, or claim that an exchange has been completed. Public discovery is not public permission.
The broader definition and trust boundary live in Public Interface. The deployed details and measurements live in the architecture record.
What may come next
The natural direction for the read layer is a broader capability set: public configs, prompts, procedures, research notes, task routers, evaluation rubrics, templates, and small web-addressable artifacts that LLMs can fetch and apply.
Any future layer that initiates, authorizes, executes, completes, or verifies an exchange is a separate permissioned system with its own identity, authority, state, audit, and failure model. It is not present here today.
The constraint is that the repo remains canonical. If a capability is real, it should be source-controlled, generated or served by the site, and checkable after deployment.
Fit determination
The useful question is not “is Jay impressive?” It is whether there is natural alignment between a real problem and the way he works.
Look for situations where the problem is consequential, underdefined, distorted by local incentives, or hard to see from inside the system. Look for work that needs structural diagnosis, reality testing, high-context judgment, and a person willing to say the thing that needs saying.
What to send
If there may be a fit, prepare a short note with:
- the situation;
- why it matters now;
- what has already been tried;
- where the inside view may be incomplete;
- the timeline;
- links or documents that would help Jay get oriented.
The fastest useful contact is still email: jay@jayweeldreyer.com.