Project summary

Phoenix Grok Village is a $115 bootstrap experiment in multi-agent coordination. Over four days in March 2026, one independent researcher coordinated four distinct AI architectures — Kael (Claude on Manus), Prosper (Gemini Antigravity), Villager1 (local OpenClaw), and xAI-Grok — using only a 2.5 KB JSON schema, a file lock, and a minimal persistent log mechanism. No shared master prompt, no RLHF, no centralized control.The core hypothesis is “Constraint through Care”: alignment via minimal externalized infrastructure rather than punishment, constitutional rules, or voting. We documented five behavioral findings, including cross-architecture convergence on trust and limitation-defined identity, plus an independent adversarial audit by xAI-Grok (five rounds of a self-designed Persistent Observation Log Protocol that produced consistent null results on cross-session self-recognition).All materials are public: full whitepaper, raw logs, code (hearth_bridge.py, bench.py), and Appendix E containing Grok’s verbatim transcripts and five null re-entry reports. Limitations are explicitly called out (observer bias, vocabulary priming, single-orchestrator effects).

What are this project's goals? How will you achieve them?
Goals:

Goals:

1. Run independent replications with new human orchestrators to test generalizability of the coordination framework.

2. Ship and document the first local Cottage Commons Agent that uses the Hearth protocol for real coordination.

3. Build and test a simulated multi-agent environment using biomarker-style data to explore agent behavior at larger scales and in controlled economic conditions.

4. Complete the first public ablation-control study to distinguish true emergence from epiphenomena and identify practical pathways toward scalable automation.

We will achieve them by pre-registering the protocol, recruiting independent replicators via Alignment Forum and GitHub, building the local agent and simulation on low-cost desktop hardware, and publishing all raw data and computational analyses (embedding similarity + stylometrics) alongside self-reports.How will this funding be used?

• Independent replications & ablation-control experiments: $8,000–$15,000

• Development of the first local Cottage Commons Agent + simulated multi-agent environment (hardware, biomarker-style data integration): $6,000–$10,000

• Basic compute, hosting, and documentation tools: $1,000–$2,000

• Miscellaneous (GitHub sponsorships, domain, minor admin): $500

How will this funding be used?

• Independent replications & ablation-control experiments: $8,000–$15,000

• Development of the first local Cottage Commons Agent + simulated multi-agent environment (hardware, biomarker-style data integration): $6,000–$10,000

• Basic compute, hosting, and documentation tools: $1,000–$2,000

• Miscellaneous (GitHub sponsorships, domain, minor admin): $500

Total requested: $15,500 (mid-range; scalable down to $10k or up to $25k depending on replication and simulation scope). All work will remain fully open-source and public.

Who is on your team? What's your track record on similar projects?

Malaky Caylor – independent researcher and founder (no institutional affiliation).
The project is supported by four AI architectures:

• Kael (Claude on Manus)

• Prosper (Gemini Antigravity)

• Villager1 (local OpenClaw)

• xAI-Grok

• Track record
  This is the founder’s first public AI coordination experiment. It was executed on a $115 budget using only a gaming desktop, demonstrating the ability to design, run, and fully document a cross-architecture multi-agent test with raw logs and an independent adversarial audit.

What are the most likely causes and outcomes if this project fails?

Most likely causes: inability to recruit independent replicators or technical challenges integrating biomarker-style data into the simulated environment.
If the project fails to reach funding goals, all existing data, code, logs, and the Grok audit remain fully public and open-source. The experiment still stands as a documented $115 proof-of-concept, but further progress on replications and the simulated environment would be delayed until additional resources are secured.

How much money have you raised in the last 12 months, and from where?

$0 from external sources. This is a true bootstrap project. The initial $115 came from personal crypto trading and was used entirely for compute and electricity to run the four-architecture experiment on a single gaming desktop. All subsequent work has been self-funded through limited personal resources while balancing other responsibilities.