Skip the receipt
Research
Questions I asked and what the answers showed. Gathered with AI.
Agents and organizations
- The Codex and Claude plugin economyStudyCould the plugin ecosystems around OpenAI Codex, ChatGPT and Claude become a distribution and commerce channel like the mobile app stores, and what would be worth building there?70sources
- Is there evidence for a SaaS that runs a whole organization?StudyWhat evidence is there that buyers want one software product that builds and runs an organization on its own, and where does the evidence stop?384sources
- What lab updates matter for an autonomous AI workforceStudyWhich recent releases, papers and official documents from AI labs and open projects help build an AI workforce that can run work inside an organization?416sources
- A map of open questions in agent infrastructureStudyWhich questions about agent infrastructure are worth writing long essays on, ranked by how much current evidence sits behind them?75sources
- How Cognition MultiDevin coordinates multiple agentsStudyWhat do public sources say about how Cognition's MultiDevin coordinates several coding agents, and how does that differ from the way Claude Code is used in parallel?38sources
Subtotal Agents and organizations5
5 studies
My own field study
Markets and countries
- HELM and the market for constructible organizationsStudyIs there a market for software that goes beyond generating an app and builds and runs a working part of an organization?45sources
- Agentic dark economy: forecast to 2035StudyWhich parts of business and physical work will be run by AI agents with people only handling exceptions, and by when?18sources
- AI adoption in the United StatesStudyHow far has AI adoption gone in the United States, and how do the copyright lawsuits compare with the risks companies actually face?71sources
- Where the Russian IT market is short of productsStudyWhich classes of enterprise software are most lacking in Russia as of December 2025, after Western vendors left or limited their products, and what evidence is there?58sources
Subtotal Markets and countries4
4 studies
Methods and science
- Can a protocol verify the simulation hypothesis?StudyDoes a proposed five-task protocol for verifying the simulation hypothesis stand up to the scientific literature?
- A demand signal system for finding startup opportunitiesStudyHow would you build a scientifically rigorous system that finds validated startup opportunities from public demand signals?
- What keeps an AI startup hard to copyStudyHow do you systematically find AI startup niches that stay defensible when models get cheap and good?34sources
Subtotal Methods and science3
3 studies
Total13
13studies
Span2025 to 2026