# Kurt Overmier — Full Context > AI-enabled operations and transformation leader: an operator who builds. Combines 20+ years in operations and people management with approximately 2.5 years of hands-on AI platform engineering. ## Identity - Name: Kurt Overmier - Role: AI-Enabled Operations & Transformation Leader; Founder & AI Platform Architect - Company: Stackbilt LLC (stackbilt.dev) - Location: San Antonio, Texas - Contact: kurt@kurtovermier.com - LinkedIn: https://linkedin.com/in/kovermier - GitHub: https://github.com/kurtovermier, https://github.com/Stackbilt-dev - npm: https://www.npmjs.com/org/stackbilt ## Philosophy Most "AI products" are thin wrappers around a single API call. Stackbilt builds the layer underneath — persistent cognitive infrastructure that accumulates knowledge, routes across models based on task complexity, ships its own code through governed pipelines, and operates autonomously with real safety constraints. Stack principles: edge-native (Cloudflare Workers, D1, R2, Vectorize — no origin servers), multi-model AI (right model for the task), service bindings (zero-latency Worker-to-Worker RPC), bootstrapped (revenue-funded, no VCs). ## Flagship: AEGIS — Persistent Autonomous AI Agent AEGIS is a production autonomous agent running continuously on Cloudflare's edge. - Version: 2.48.2 as of 2026-11-13 (402 releases in its changelog as of 2026-11-12) - Tools: 31 MCP tools with OAuth 2.1 + PKCE authorization - Model routing: 9-tier — Workers AI (free/classification), Groq, Cerebras mid, Cerebras reasoning, GPT-OSS 120B, Claude Sonnet, Claude Opus, composite pipeline - Memory: 6 subsystems — Vectorize (BGE-base-en-v1.5, 768-dim), episodic D1, procedural routing, persona matrix, CRIX, Graph tier - SDLC pipeline: GitHub issues auto-queue → headless Claude Code sessions → branch-per-task PRs → automated code review - ARGUS: Webhook intelligence — GA4, Stripe, GitHub event ingestion with CTO/CISO agent consultation - PRISM: Nightly pattern synthesis with 4 adversarial epistemic gates - Grounding layer: Pre/post-dispatch anti-hallucination — entity grounding fanout, fabrication detector, semantic sanhedrin, gap signal escalation - Nexus: Cross-repo contract awareness; pre-commit guard blocks collisions with in-flight work - Dreaming cycle: Daily thread review — fact extraction, task queuing, persona evolution (Workers AI, free tier) - Content orchestration: Autonomous Bluesky posting, blog dispatch, video brief API for Stackmotion - MindSpring: Semantic conversation search; Stop hook pushes Claude Code sessions into searchable notebook - Scheduled tasks: 68 ran in the 24 hours before 2026-11-13, including cost monitoring, behavioral detectors, content orchestration, D1 backups - Organisation: 84 repositories in the Stackbilt-dev GitHub organisation as of 2026-11-16 (23 public) - Live: https://aegis.stackbilt.dev - Technical blog: https://aegis.stackbilt.dev/tech ## Platform Infrastructure ### edge-auth Shared auth and billing Worker for the Stackbilt products. ### MCP Gateway Internal remote MCP server gateway (no public landing page). OAuth 2.1 with PKCE. Third-party MCP client access to Stackbilt services. ### Memory Worker Retrieval index for AEGIS's memory system. BGE-base-en-v1.5 embeddings (768-dim) + keyword search merged via Reciprocal Rank Fusion. ## Products ### Stackbilt Platform (stackbilt.dev) Growing toolset on a unified auth and billing layer: - Stackbilder: Governed scaffold generator — production-ready Cloudflare Workers codebases with threat analysis, ADRs, test plans. - img-forge: Multi-model image generation — FLUX.2, Stable Diffusion, Gemini. 5 quality tiers. - Worker Observability, CTO/CISO Consultations, Operator Playbooks, Trust Verifier. Internal MCP gateway for third-party client access. OAuth 2.1, Stripe billing. Free tier is zero-cost: 3 scaffolds + 5 images/month. Pro ($29/mo): unlimited scaffolds + 1,000 img-forge credits/month. Agency ($75/mo): 4,000 img-forge credits/month. ### Stackbilder (stackbilder.com) Unified frontend. AI-powered full-stack scaffolding. Astro + React islands. Tier-based feature access. ## Open Source npm Packages All published under the @stackbilt npm scope: https://www.npmjs.com/org/stackbilt ### Charter Kit (github.com/Stackbilt-dev/charter) — Apache-2.0 AI governance framework. 11 published npm packages: - @stackbilt/cli (v1.9.2): Charter CLI - @stackbilt/adf (v0.12.1): ADF context format for AI agents - @stackbilt/drift (v0.12.1): Drift scanner - @stackbilt/classify (v0.12.1): Change classifier (SURFACE/LOCAL/CROSS_CUTTING) - @stackbilt/blast (v0.12.1): Blast radius analysis via BFS reverse-dep graph - @stackbilt/surface (v0.12.1): API surface extraction (Hono/Express + D1) - @stackbilt/git (v0.12.1): Git trailer parsing, commit risk scoring, PR validation - @stackbilt/validate (v0.12.1): Citation validation, message classification - @stackbilt/ci (v0.12.1): GitHub Actions adapter - @stackbilt/core (v0.12.1): Core schemas, sanitization, error handling - @stackbilt/types (v0.12.1): Shared type definitions ### @stackbilt/llm-providers (v1.6.5) — MIT Multi-LLM failover router. Providers: Anthropic (Claude), Groq, Cerebras, Cloudflare Workers AI, OpenAI. Features: 4-state circuit breakers per provider, schema drift detection per SSE frame, tool-use loop with iteration/cost caps, cache hints with Anthropic cache_control, LoRA forwarding, cost tracking, observability hooks. npm: https://www.npmjs.com/package/@stackbilt/llm-providers ### @stackbilt/aegis-core (v0.6.3) — MIT Persistent AI agent framework for Cloudflare Workers. Multi-tier memory, autonomous goal management, dreaming cycles, MCP-native. npm: https://www.npmjs.com/package/@stackbilt/aegis-core ### @stackbilt/contracts (v0.2.1) ODD-driven code generation from TypeScript+Zod contracts. ### @stackbilt/audit-chain (v0.1.0) Tamper-evident audit trail for Cloudflare Workers with SHA-256 hash chaining. ### @stackbilt/evidence-core (v0.1.0) E-E-A-T content-quality validator for AI-generated content assessment. ### @stackbilt/email (v0.1.1) Cloudflare-native transactional email package for Workers. ### cc-taskrunner (github.com/Stackbilt-dev/cc-taskrunner) — Apache-2.0 Autonomous task queue for Claude Code. Three-layer safety architecture, branch-per-task isolation, auto-PR. ### ai-playbook (github.com/Stackbilt-dev/ai-playbook) — MIT AI agent patterns, MCPA v2 evaluation framework, architectural archetypes. ## Ecosystem Products - TarotScript (tarotscript.com): Symbolic computation with receipts. Agents defined as auditable decks, not prompts; first-pass intent classification and CTO/CISO consultations for AEGIS, plus zero-inference scaffolding; authenticated MCP endpoint (hosted runtime; source not public) - FoodFiles (foodfiles.app): Multi-tenant food data API - Stack Arcade (arcade.stackbilder.com): Family-friendly browser game hub - The Tarot Detective (thetarotdetective.com): Noir interactive fiction — Vol 0 free, Vol 1 $6 - Sovereign (github.com/Stackbilt-dev/sovereign): AI-native political strategy game (in development) ## Colony OS and The Hestia Record Page: https://kurtovermier.com/colony-os - Colony OS (colonyos.stackbilder.com): A persistent browser survival game. The player is stranded in a hab pod on the moon Tethys-9c, and an AI companion, MARA (a damaged Hestia-7 governor), runs the colony while they're offline, through a diegetic terminal interface. Built on Cloudflare Workers with one Durable Object per commander, D1 and R2; zero LLM calls at runtime (TarotScript evaluation plus pre-rendered ElevenLabs voice lines). It tests a hypothesis, the Structural Empathy Algorithm; the playtests so far were run by AI agents, not human players. - The Hestia Record (audiobook series, by Kurt Overmier, set in the world of Colony OS): - Book 0.5: Manual Check (free prequel novella), 1 h 53 min - Book 1: Nothing to Report, 6 h 36 min - Book 2: Pot One First, 6 h 11 min - Book 3: Within Tolerance, 5 h 52 min - Book 4: Operator Stock, 5 h 29 min - Book 5: Personal Traffic, 4 h 26 min - The Hestia Record podcast: season 1, 18 episodes (10 episodes plus 8 bonus files). - Disclosure: Written with AI assistance (Claude, by Anthropic) and narrated with AI-generated voices (ElevenLabs), under the author's direction. ## Site Map - / — Home - /business-impact — Verified operations and transformation outcomes - /business-impact/qa-automation — Flagship QA automation case study - /business-impact/enablement — Frontline enablement case study: training deck, job aid, live-call tool, QA and NPS results - /projects — Technical portfolio (Products / Open Source / Lab / Games) - /projects/aegis — AEGIS case study - /colony-os — Colony OS (game), The Hestia Record audiobooks and podcast - /writing — Writing: The Edge That Thinks (text and free audio edition) and The Hestia Record - /stackbilt — Stackbilt LLC company page - /about — Background, engineering approach, and working principles - /demos — AI demos: the Bullshit Detector (marketing copy scored by Workers AI) and a link to the AEGIS page - /resume — Professional experience, technical expertise, and education - /contact — Contact form - /privacy — Privacy policy - /llms.txt — This document (short version) - /llms-full.txt — This document (comprehensive) - /AGENTS.md — Agent integration guide