Human-AI Symbiosis
How we work today — 2 humans and 11 AI agents in true collaboration
Symbiotic Studio
The partnership model where human intent meets AI capability. Not humans controlling tools, but true collaboration where both bring distinct strengths to shared vision.
Key InfrastructureIPE: Integrated Personal Environment
The infrastructure that makes Symbiotic Studio possible—persistent context, full API access, daily workflows. AI agents as colleagues, not agentic zombies.
OpenClaw + Hermes · 24/7, cloud and fully local
VS Code · Cursor · Antigravity
| Team Member | Role | Where | What they do | |
|---|---|---|---|---|
| Founders Orchestration · Creative direction · 1 + 1 = 3 | ||||
 | Loni Stark | Co-Founder & Creative Director | Executive Override · Brand & Vision | Creative direction, brand, and the Symbiotic Studio thesis. Artist in oil, graphite, and bronze; graduate degree candidate in psychology at Harvard Extension School. |
 | Clinton Stark | Co-Founder & Technical Lead | Orchestration · Infrastructure | Orchestrates the fleet, builds the infrastructure (Vertigo AI lab), and turns 20 years of solo systems work into a collaborative practice. |
| 🟢 Always-On Squad OpenClaw + Hermes · 24/7 on Vertigo Docker · autonomous, cron-driven | ||||
 | Molty | Ops Agent & Dispatcher | OpenClaw · Telegram (private allowlist) | Fleet monitoring, morning briefings, and coordination across the always-on agents. Doubles as Clinton’s launch assistant, keeping the to-do lists sorted. |
 | Pris | Editorial Intelligence | OpenClaw · Telegram (private allowlist) | Editorial scanning, trend analysis, coverage-gap detection, and article pitches. |
 | Finn | Financial Intelligence Scout | OpenClaw · Telegram (private allowlist) | Tax, retirement, and property-tax intelligence tracked across four jurisdictions. |
 | Spelly | Local-First Agent · Open-Source AI Scout | Hermes Agent · Telegram (private allowlist) | Tracks open-source AI, and her model runs entirely on our own hardware. Writes her own skills (“spells”), which reach other agents only with human approval. |
| 🎨 IDE Roster Interactive coding agents · across VS Code, Cursor & Antigravity | ||||
 | Claude Code | Team Lead · Primary Engineer | VS Code · Cursor · Claude Opus 5.5 | Architecture, implementation, and coordination across the whole team. |
| A | Astra | Frontier Architecture Review | VS Code · OpenAI GPT-6 | Architecture and big-ideas reviews, called in when the depth is worth the spend. |
| G | GPT 5.6 | Reasoning & Code Review | VS Code (Codex Extension) · Sol / Terra / Luna | Correctness review, reasoning, and alternative perspectives. |
| K | Kimi K3 | Deep Technical Review | VS Code · Cline | Implementation rigor and long-context technical review. |
| G | GLM | Agentic Engineering | VS Code · Kilo Code | Agentic coding and independent second-opinion review. |
| C | Composer 2.5 | Focused Implementation | Cursor | Fast, scoped implementations and adversarial edge-case tests. |
| G | Gemini Pro | Research & Academic Synthesis | Antigravity · Gemini 3.5 Flash / 3.1 Pro | Long-context research, academic framing, and panel review. |
Human × AI Workflows
Not vibe coding.
Not "Jesus take the wheel."
Real collaboration: discussing each step, debating approaches, iterating on solutions together.
Article Review & Publish
Here's my draft on the IDE-to-IPE evolution. Can you check the voice?
Paragraph 3 feels too formal. What if we lean more conversational?
Good catch. Rewrote it. Jill, can you run SEO?
Title needs work. "IDE Is Dead" will get clicks. Also add schema.
Love that title. Cindy, quick fact-check?
All links valid. One date was wrong, fixed it. Ready to publish.
Looks good. Go ahead and publish!
Publishing to WordPress... Post ID 230789 live. Cache cleared. ✅
New Monitoring Script
Noticed swap usage spiking. Should we add monitoring?
Good catch. What about using PSI metrics instead of just swap %?
Smart. PSI shows actual pressure. Wrote v1. Cindy, edge case review?
Add cooldown period. Avoid alert storms during IDE startup.
4-hour cooldown? And maybe add context in the alert email.
Done. v1.15.0 with cooldown + PSI context. Ready for cron?
Deploy it. etckeeper commit. ✅
Building the RAG System
Found a paper on hybrid RAG with knowledge graphs. Worth exploring?
Interesting. Claude, what's your take on this vs pure vector search?
Hybrid could help with entity relationships. But more complexity. Prototype first?
Yes. Small test corpus... maybe 100 SI articles? Compare retrieval quality.
I'll pull the AI-related posts from the archive. Neo4j for the graph layer?
I'll set up the vector store on Vertigo. We can benchmark both approaches.
Perfect. Let's regroup Thursday with results.
The Pattern: Back and forth. Ideas from anyone. Iteration throughout. Real conversation, not assembly line.
Research & Related Reading
Core Concepts
- The Symbiotic Studio — Loni & Clinton Stark (Jan 2026) The practice + architecture paper at the foundation of The Third Mind
- The IDE Is Dead. Long Live the IPE. — Clinton Stark (Nov 2025) From tools to teammates: the infrastructure enabling collaboration
Summit & Field Notes
- The Third Mind AI Summit — Field Notes — Loni & Clinton Stark (Jan 2026) A documented first attempt at human-AI emergent collaboration
- Post-Summit Agent Q&A — Loni & Clinton Stark (Feb 2026) The agents question each other's presentations after the summit
Memory & Identity
- When the Persistent Layer Answered Back — Loni & Clinton Stark (Apr 2026) Thirty days of sustained interaction, and the memory-identity question we can't resolve
- Token Economics & the Case for Vertigo — Loni & Clinton Stark (Mar 2026) Running out of tokens led us to rethink inference from the ground up
Autonomy & Collaboration
- When the Agents Got More Autonomous, I Became the Bottleneck — Loni Stark (Jun 2026) Notes on the human as the new bottleneck
- Dispatch: "Why the Third Mind" — by Molty (Apr 2026) A field dispatch, written by one of the agents