Mick Darling

Founder and CEO of Dollhouse Research, maker of DollhouseMCP and MCP-AQL. Independent researcher based in Massachusetts.

At Dollhouse Research I build composable AI infrastructure and study how to make agentic AI systems reliable enough to trust with real work. That means capability and safety designed together, not bolted on afterward. Before Dollhouse, I founded Tomorrowish, applying real-time natural language processing to social media and broadcast television. That work produced multiple granted U.S. patents, and I have nearly two decades of experience across product, engineering leadership, and strategy in media and technology companies.

Research Areas

• Agentic AI systems and composable AI behavior
• Multi-agent reliability and supervision architectures
• AI safety infrastructure
• AI-assisted research methodology

Publications

Peer-reviewed papers and preprints are forthcoming and will be listed here as they are published.

Specifications and Proposals

• Darling, M. (2026). MCP-AQL Specification (v1.0.0-draft). Dollhouse Research. Zenodo. doi:10.5281/zenodo.21398630
• Darling, M. (2026). AILIS: A Proposed Layer Model for AI Systems (v0.2.0). Dollhouse Research. Zenodo. doi:10.5281/zenodo.21398506

Selected Writing

• Darling, M. (2026). Orbital Lifeboats: Napkin Math for a Solar-System Emergency-Cache and Rescue Network (v0.1.0). Zenodo. doi:10.5281/zenodo.21398972 — the WWII Channel rescue buoy reimagined for spaceflight, where rescue distance is measured in delta-v, not meters.
• Darling, M. (2016). Think Like an Alien Engineer. Medium; archived on Zenodo, doi:10.5281/zenodo.21399117 — engineering-constraints analysis of technosignature hypotheses for Tabby’s Star (KIC 8462852), with a testable infrared-occultation prediction. Later observations refuted the hypothesis as predicted it could be: the dimming source emits in infrared, consistent with dust.

Patents

Granted U.S. patents in natural language processing, as inventor of record at Tomorrowish LLC:

• Discovering keywords in social media content (US 10,733,195, granted August 4, 2020)
• Scoring social media content (US 10,614,074, granted April 7, 2020)
• Displaying social media content (US 10,607,299, granted March 31, 2020)
• Discovering keywords in social media content (US 10,528,573, granted January 7, 2020)

Links

• ORCID: 0009-0008-7116-1393
• GitHub: github.com/mickdarling and the DollhouseMCP organization
• DollhouseMCP: dollhousemcp.com
• LinkedIn: linkedin.com/in/mickdarling

Contact

Email: mick@mickdarling.com