BootstrappedAI

AI work crew platform where named agents operate inside scoped workrooms, producing receipts, approvals, tasks, calls, and business artifacts with visible governance.

A headless Multi-Agent harness and framework as a self evolving community platform connecting creators with top-tier education infrastructure. This is a deep research in-house funded project to discover how agent collaboration can support and empower the next generation of AI supported Students, Agents & Educators.

Trust-first autonomous business operator that validates ideas, runs approved work, verifies outcomes, and only charges for proven execution. Thesis is that: Given an Agentic Harness that supports long workflows over multidimensional node-based tree shaped org-wide data context-aware with time-base workflow-branch event-correlation awareness (ex: medical report came via email does provide inference of legal-case stage/status alongside work tree - case is accepted on med-legal eval phase awaiting further tests [inference of such self-triggering future event expected-by and temporal retriggering as future scheduled follow-ups] ) is able to transverse (as humans do) parallel-bifurcated/odd-shaped/async-operating node-based work trees of human orgs - requires a dedicated harness that is (context layered/persona projected [ case is med-legal evaluated while accounting invoiced in same temporal juncture - thereby workflow awareness triggers parallel multi-headed-eval (ex:as accounting persona or legal analysis persona) runs using the (persona projection) @ array defined by org-wide work tree at the specific work-tree temporal event juncture] ) allowing a " sensed" event (ex: work-tree position awareness -> receipt of identified document arrived -> becomes stage defining -> future event scheduled triggers to be evaluated or executed on future heartbeats ) de-facto simulating case/future events temporal/procedural awareness along the org-wide work tree in real-time. Such multilayered/persona projected harness should have "tools" for the self-contained-persistent-org-aware agent-persona-projected harnessed LLM loop to ("look into the case past along work tree"- "precog into the case future along the work tree"). A self-contained-persistent-org-aware harness projection might need projecting/mutating to transverse dissimilar containers where the projection substrate proves uneven (ex: Claude file-folder-based hierarchy navigation effecting progressive disclosure of SOP+Scripts might need to be projected in an equivalent-yet-compatible but dissimilarly-shaped progressive disclosure able harness version representation; that is database-stored for a more deterministic eval/execution harness that runs on ex: a Vercel Serverless function) all while able to project kanban-like (inferred from work tree branches temporal projection that is advanced/maintained/curated by the above-mentioned event-work-tree positional/temporal correlation) views for the human-harnessed co-operator. This thesis builds upon previous wcpdr research. All while and most importantly the harness is able to provide agents such work-tree-navigation/temporal-awareness over dissimilar channels (Slack/SMS/Teams/Voice) thereby allowing the centralized memories to be attached to the node-based work-tree for progressive-as-needed context-sensitive progressive disclosure. Such agent harness persona projection becomes both contextually/temporal aware while able to navigate purpose-driven environments (transverse from computer-retrieval-file-based-substrate to communicate with channel-bound (ex: Slack/SMS/Voice) human operators) - all-while maintaining operational/temporal/org-wide contextual memory

Glass-box team agent platform for Slack, Teams, and web work, built around predictable cost, approvals, audit trails, and never-pay-for-failure trust.

A self-funded startup studio, media hub, and automated sponsorship CRM.

Agentic Harness for California Workers Compensation Case Analysis & Valuation. The research thesis behind the project is that every long workflow - requires a dedicated harness simulating in real-time the target best outcomes via complex evaluation/simulation of those desired complex outcomes, having that simulation infrastructure (in-place) that provides ongoing feedback for multi agents teams, enables the implementation of that ( self-eval ) reinforced self-correction ( cyber-proprioception of the data/intelligence system ) over time, thereby using such specialized + layered + self-assembling / error-sensing harness that enforces monitoring by managing agents, allows those directing agents to actively & globally course correct operations from lessons learned arising from the aggregated feedback loop of the commonly shaped case load. Due to the high complexity / long span ( years ) of this workflow, it has proven to be a worthy research target.

Agent Driven Adversarial 3D Simulatrum Mafia Game - Video simulation & Game Theory Research - Active exploration of self-contained multi-agent games - Agents actions and communications via an internal harness A2A bus that provides in-game orchestration of outcomes via MCP servers.

Agentic Engineered Strudel REPL Editor - Agentic Strudel music composition, rendering and orchestration research using Web-MCP as orchestrating layer. Ongoing research.

Autonomous developer teams powered by MCP, executing complex codebase migrations securely.

AI-powered appointment reminder, confirmation, and recovery SaaS for businesses on legacy or disconnected systems. Works from CSV, email parsing, and calendar sync with no integration projects required.

A spatial, agent-native IDE that renders codebases as living forests — orchestrating multiple AI agents across multiple projects with real-time health, scope, and conflict visibility.

Aggregated Research and Forward Engineering on AI Agents & Tools integration for MSP Security and Operations Automation. Dogfood.

Instant, AI-powered course generation for independent educators and thought leaders.