Work

For two decades, I have built open source projects, service businesses, research systems, and agent-native companies around one recurring question: how can small teams use software to do work that used to require large organizations? The current edge is an adaptive runtime, governed company context, and a web agents can understand.

Current focus

The active edge of the thesis

Build snapshot · July 2026: Punk and Pantasso are in sustained production hardening, with more than 1,200 repository changes across the two projects in the past month. The broader contribution curve is now nearly vertical.

Punk

Proof-first adaptive runtime

current focus

An adaptive runtime for production AI agents that learns from repeated work, proves cheaper paths, and keeps the live model as fallback.

Proof-first adaptive runtime for production AI agents. Learns from repeated work, verifies cheaper execution paths, governs risky actions, and explains every route decision.

Pantasso

Governed company context

current focus

Governed company context for AI-native teams: cited answers, honest gaps, SQL-accessible records, and scoped access for assistants.

Agent-first company data platform for AI-native teams. Turns approved chats, documents, files, and business systems into governed context, cited answers, and scoped agent access.

Plasmate

Open source browser infrastructure

current focus

A browser and semantic format for the web consumer that does not have eyes.

Agent-native headless browser engine. HTML in, Semantic Object Model out. 10x token compression, Apache-2.0.

Till

Agent access control

current focus

Activation-limited credentials for letting agents act without handing them permanent keys.

Activation-limited API key proxy. Give AI agents exactly the access they need - keys that work till they don't.

Looper

Continuous improvement

current focus

The retention-layer idea made operational: scheduled loops that improve systems over time.

Continuous improvement as a service. Point an AI at anything, set a schedule, and it improves the thing every night.

CrowdProof

Simulation as research

current focus

A social laboratory for watching many agents react, debate, and expose weak ideas quickly.

Live social simulation platform. Submit a stimulus, watch hundreds of AI agents react across simulated platforms in real time.

The map

Organized by the question each project answers

The work looks disparate when sorted by industry. It becomes coherent when sorted by the problem being tested: infrastructure, service delivery, solo leverage, agent operation, open source distribution, and lessons from prior waves.

Runtime, context, format, access, memory

Agent infrastructure and data

Systems for agents that must learn, understand company context, read the web, remember, pay, and act with evidence and control.

Outcome delivery instead of seats

Service-as-software businesses

Companies built around AI workers that perform business functions directly, with humans supervising edge cases instead of operating dashboards.

Leverage for owner-led operations

Single-human company systems

Operating layers for small teams and solo operators: persistent work queues, continuous improvement, delegated execution, and approval-ready outputs.

Markets, media, commerce, simulation

Agent-run experiments

Experiments that test how far agents can operate creative, financial, research, and commercial workflows with minimal human intervention.

Adoption before extraction

Open source and community

The throughline from Joomla and Mautic to Plasmate: build infrastructure people can inspect, adopt, extend, and trust before commercial layers appear.

Acquired, paused, killed, learned

Archive of lessons

Not every build remains a current company. The archive shows the lessons, false starts, and prior waves that shaped the current thesis.

Eras

A lifetime of work, not a pile of projects

Each era sharpened the same instinct: find a consumer class that existing infrastructure underserves, then build the missing layer that gives smaller teams more leverage.

2011-2015

Open source and small-business leverage

The earliest pattern was making expensive business capability available to smaller organizations: CRM, marketing automation, services, and community-led software.

CRMeryWebSparkMauticTechVets

2016-2022

Headless systems, ownership, and web3 experiments

The work moved toward APIs, ownership, programmable finance, identity, and markets. Many of these were experiments, but they built the muscle for autonomous systems.

SaelosCandorAfterCaiman FundMyPOA.xyzHot Potato

2023-2025

AI-native services and operating experiments

The focus shifted from software people use to agents that do the work: sales, education, commerce, social media, trading, and customer operations.

SalesAideTeacher's AideWith AgencyStockCrocZuiceWe Post

2025-2026

The operating stack for AI agents

The current arc connects an adaptive runtime, governed company context, and an agent-readable web with the systems that let agents act and improve safely.

PunkPantassoPlasmateTillLooperCrowdProof

Full index

The complete record

Some are companies. Some are proofs. Some are archived lessons. The index is kept complete because the body of work matters more than a polished highlight reel.

42
Active
4
Acquired
11
Archived

Agatha's

paused

Boutique fine art gallery for NFTs

Archive

Loded

paused

Leasable Ethereum hashrate for those looking to invest in crypto mining without the high overhead expense of machine acquisition.

Archive

Pocket Pandas

archived

Adorable onchain panda tamagotchi with special traits

Archive

Saelos

archived

API driven Customer Relationship Management software. Serverless and headless

Open sourceArchive