Things I'm working on..

Turns AI agents into a governed organization

The runtime underneath much of this page. Each domain of work becomes a node run by its own scoped AI operator — on a charter, a heartbeat of scheduled autonomy, and tiered permissions enforced mechanically, so consequential actions wait for a human signature: the agent doesn't just promise not to overstep, it structurally can't. I set direction and approve what matters; Kernel runs the operation. Ships as a self-contained runtime with a signed desktop app.

Talk to your coding agent from Telegram

Talk to Claude Code, Codex, Grok, or OpenCode from Telegram. The harness still runs on your Mac: inbound messages become governed agent turns, replies stream back, and nothing sends without an explicit yes.

Chat your resume into existence

An AI agent that clears the mechanical half of a job search. Appleo writes and versions your résumé, tailors a copy to every listing and scores the fit, tracks your applications, and renders clean, ATS-ready PDFs — with a public share link when you want one. You talk; it handles the paperwork; the search stays yours.

The control layer for AI-built websites

An LLM will build you a website in fifteen seconds; what it won't give you is control over what it built. Auxon turns a brand into files — a voice, a visual style, a page shape — that every page is generated from and checked against, so a page that would drift off-brand fails a check and never ships. It captures a brand off a live site, a screenshot, or a narrated screen recording, and it works on sites it didn't build.

Conduct & persona for AI agents

The lines your agent won't cross, and the voice it speaks in. You declare conduct (what it may say and do) and persona (how it sounds) in one policy source; Decorum compiles both into the model's prompt and into mechanical, fail-closed floors on input, tool calls, and output. Prompt asks; structure guarantees. Licensed per deployment and run entirely on your own infrastructure — already enforcing in production, including behind Appleo.

Token ROI for AI work

To assay is to test ore for its precious metal. Assay does that to AI spend: it meters every token an agent burns, seals the transcript that drove the spend, and joins the two — what a result cost, where the waste is, what the tokens were actually worth. Running inside my own stack today; next it takes the same commercial shape as Decorum.

An evidence-backed second opinion on an options position

A read-only decision-support tool for options income. It pulls real market, options, macro, filings and on-chain data, computes the evidence deterministically, hands a model a prepared brief rather than a raw pile, and writes down a dated multi-horizon read it later grades itself against — then proposes concrete structures like covered calls, collars, or sitting in cash. There is no execution path in it; it never places a trade.

An AI-discoverable identity platform

A professional profile at an @handle that AI assistants can actually find and answer questions about — backed by structured data, a REST API, and an MCP server, so ChatGPT, Claude, and the rest know who you are.

The open core of my agent runtime

An embeddable, MIT-licensed agent-loop engine: createAgent() gives you the turn loop, typed fail-closed tools, streaming, and provider routing across cloud models or a local Mac lane, with a persistent learning layer on top. The engine that became Kernel's agent core — and the open packages are yours to build on.