Skip to content

Thinking with Agents (UVM Agentic AI Bootcamp)

  • Author/Source: Erkmen G. Aslim & Emily Beam (Department of Economics, University of Vermont)
  • Original: thinkingwithagents.github.io — landing page for a two-session agentic-AI bootcamp delivered at UVM in April 2026, with companion slide PDFs, prep guide, and an applications gallery of reusable skills

Key Ideas

  • A two-session agentic-AI bootcamp for economists, hosted by the UVM Department of Economics. Session 1 (April 22, "Getting Started with AI Tools") is the conceptual on-ramp; Session 2 (April 27, "Research Workflows, Teaching & Applications") is the applied demo of reusable skills on a live paper.
  • No installation required to attend — the prep guide (Getting Set Up) handles install for those who want to follow along, and an Outside Resources page collects further reading.
  • Session 1 covers the core agentic concepts: chat AI vs. agentic tools, the context window, standing instructions, planning, voice files, skills, and live demos (website redesign + custom skill creation). Detailed in agentic-bootcamp-1-aslim-beam.
  • Session 2 is the applied half, organized around two streams: Emily on research workflows (auditing a replication package with /code-review, paper auditing with /econ-audit and /review-paper, running Lee bounds live), and Erkmen on teaching & applications (lecture-builder skill via /academic-beamer-deck, idea stress-testing with /research-brainstorm, web scraping with /find-data). Detailed in agentic-bootcamp-2-aslim-beam.
  • "What to try next" homework is concrete and low-stakes: try a data-exploration → analysis-plan workflow on your own data, draft an assignment, run /econ-audit on a paper before submission, or use Claude Code to revise a draft.

  • Skills are distributed as an installable bundle — npx skills add thinkingwithagents/skills pulls the full economist pack; individual skills can be added with --skill <name>. This makes the bootcamp's playbooks reusable across institutions, not just UVM.

  • Site update (June 2026 re-clip): from bootcamp page to standing resource hub. The later version of the landing page is reorganized around three entry points instead of the two sessions: Getting Started (terminal, Git, setup, in "layers," with a $0 path for "one week and no budget"; the claim is that most economists can be doing useful work in an afternoon), Skills & Workflows (installable research and teaching skills plus 30-minute workflow recipes), and Teaching Modules (drop-in lesson plans and exercises for your own courses and workshops). A "Where should I start?" router sends readers by persona: never used a terminal → Layer 1; uses ChatGPT but wants real tools → Layer 3; already on Claude Code → Skills & Workflows; wants to teach → Teaching Modules. A companion Substack now carries posts on hidden researcher decisions and pre-registration under agentic AI, a "Workflow Wednesday" series (starting with transcription apps), and reflections on iterating with agents in teaching prep.

Summary

This is the landing page for Thinking with Agents, a two-session agentic-AI bootcamp run by Erkmen G. Aslim and Emily Beam at the University of Vermont's Department of Economics in April 2026. The page is a thin index — its real content lives in the two slide decks (linked) and a separate applications gallery where each /-named skill (e.g., /code-review, /econ-audit, /research-brainstorm, /find-data, /academic-beamer-deck) gets its own page with the SKILL.md spec and a usage walkthrough.

The pedagogical structure is "concepts first, applications second." Session 1 takes attendees from chat-AI to agentic tools, introduces the context window as the load-bearing concept, walks through planning files, and lands on safety/permissions. Session 2 then demos those concepts on a real paper-in-progress (a Bangladesh remote-education RCT) — code audit, adversarial paper review, and a live "do Lee bounds" exercise from referee comment to results — followed by a teaching-side build of a lecture-builder skill via /skill-creator.

The site is positioned as a course-in-a-box: anyone can install the skill bundle, run through the prep guide, and replicate the bootcamp at their own institution. It is the closest thing in the wiki to a turn-key agentic-AI-for-economists curriculum. By mid-2026 the site had outgrown the bootcamp: it is now a persona-routed onboarding hub with a layered setup path, a skills/recipes gallery, reusable teaching modules, and an ongoing Substack. Beam (associate professor; development economics and RCTs) and Aslim (assistant professor; health economics and applied micro) present it as a standing resource rather than a record of one event.

Relevance to Economics Research

Most "AI for economists" resources in the wiki are individual posts or talks. Thinking with Agents is unusual because it ships executable infrastructure — installable skills with versioned SKILL.md files — alongside the slides. For a faculty member running an internal AI workshop, this is a near-complete starter kit: two 90-minute sessions of slides, a prep guide, an outside-resources reading list, and a one-line install for the skills demoed. The bootcamp also models a realistic adoption arc (concepts → research workflows → teaching tools), making it useful both as a curriculum to copy and as a pattern for what to teach in what order.