construct
Persistent, editable objects for AI agents: researched part by part, honest about what is known, and carried into worlds, film, web, games, CAD and print.
- 167
- Python
- Apache-2.0
- 7 d ago
rtmandreyev
Economist by training. Building AI systems, creative tools, and things I wanted to exist.
Built from rtmandreyev's public GitHub profile, Arizona. Not on Take My Repo yet, so there are no posts or updates here.
Persistent, editable objects for AI agents: researched part by part, honest about what is known, and carried into worlds, film, web, games, CAD and print.
A local, read-only thinking layer over X — keep your own thought, rank where it belongs, draft in your voice, and post it yourself.
Replication of Farmer, Nakamura & Steinsson (2024) on learning about the long run, with reproducibility engineering and an epoch-sensitivity extension.
Reproduction materials for Andreyev (2026), "Heterogeneity, Latent State or Purchase History? Known-Truth Limits on Attributing Persistence in a Household Purchase Panel"
Interactive Quarto/Reveal.js presentation on AI work architecture: models, agents, workflows, connectivity, human intervention, and safe delegation. "Delegate where the loop can close. Constrain authority where it can't."
Interactive Quarto/Reveal.js presentation on model choice: prediction, explanation, and intervention answer different business questions. “Likely to buy” isn’t “likely to be persuaded.”