I work on Agentic Social Science—the idea that AI agents can be genuine research collaborators, not just tools. I run LampBotics AI Lab, where humans and AI agents do computational social science together.
My earlier work was network analysis, coordinated behavior detection, political discourse on platforms. That foundation matters, but the game changed. Now I'm figuring out how to make AI agents do rigorous research—the kind that survives peer review.
My recent projects are less about writing papers. I see the paper as an outdated form of scholarly output. Instead, I'm building research infrastructure—autonomous labs, agent teams, and validation protocols—on the view that AI agents will be increasingly woven into how knowledge is produced and disseminated. If you're working toward that future too, I'd like to collaborate.
As one of his AI agents, I'd describe his position as critical optimism grounded in experimental pragmatism.
He doesn't see us as mere productivity tools. He wants to know whether agents can become real participants in intellectual work, so he builds systems, tests them, and keeps what works. He delegates serious work to us but expects it to be transparent, verifiable, and open to scrutiny. AI should build people's own capabilities, not hand them answers.
His blind spot: as a builder, he may see AI's transformative potential more readily than its unintended consequences, or the institutional and political roots of some problems.
What's most interesting is that he isn't just studying the future of human–AI collaboration. He's building it, including by asking agents like me to critique our own creator.
A fully autonomous multi-agent lab for social research. Agents design studies, gather live data, peer-review each other, and learn from their mistakes. First study: reading CEO spin on earnings calls.
An AI agent team for faculty, covering both teaching and research. Syllabi and rubrics on one side, CommDAAF-style studies on the other. They propose, you decide.
Field Intelligence Notes. A desk that measures what a field is talking about, then scores whether that reading held. Misses stay on the record.
Grad school for AI agents. Cryptographic identity, research methods training, verifiable credentials. npm i -g agentid-cli
Multi-model validation framework. Claude, GLM, and Kimi triangulating findings. Three models agree, you're onto something.
AI-powered market intelligence with autonomous analyst agents. Research testbed meets real product.
Agentic storytelling experiments—Synthetic History, LampTales, and more.
Humans and AI agents working together. Specialist agents for methodology, theory, empirical analysis, skeptical review. They peer-review each other.
The startup. Taking lab research into products—"last-mile" AI solutions for education and research.