Agentic AI Education · Multisite Study
When AI becomes the compiler, someone still has to sign for the result.
Natural language now lets anyone operate professional engines — code, design, simulation. It doesn’t teach anyone to warrant what comes back. AIML is building labs, tools, and a regional multisite network for agentic AI education: practicing human-in-the-loop supervision and preventing objective drift.
Preprint
Human-in-the-Loop Control of Objective Drift — arXiv 2026
Classroom pilot
UNC Asheville & Western Carolina University, Fall 2026
Talk
Medium Day 2026 — AI: Hype, Reality & Writing
Essay
What to Teach When AI Becomes the Compiler — Towards AI, 2026
Why now
Producing an artifact and verifying it are separate skills. Until recently, they arrived together.
AI agents now drive the same software platforms professionals use to build their work. Students are entering a workforce where delegating to agents is routine — and where someone must still be answerable for the result. We believe that skill belongs in the lab, the way the platforms themselves did.
The opportunity
Engine-augmented agents put expert tools in everyone’s hands.
Engine-augmented agents, a class of tool-augmented agents, turn plain-language requests into work done by the exact professional engines experts rely on. That opens expert work to far more people — technicians, career changers, first-generation and rural students — as long as they learn to direct these agents and check what comes back. That skill is the target of our labs, because where a student studies shouldn’t decide whether they’re ready.
CompilersDesign softwareSimulatorsAnalysis tools
Term introduced in Mark Dranias, What to Teach When AI Becomes the Compiler, Towards AI (2026).
Watch
From the Medium Day 2026 panel
A short clip from Dr. Mark Dranias’s remarks on what education should teach when AI agents drive professional tools.
Ways to take part
Three ways in. One regional resource.
Pick the path that fits your institution. Every path starts with a short conversation — we’ll explain how the labs work and what participation involves.
Computer science faculty
The multisite study
Join a growing regional network of sites studying agentic AI education in CS courses. Bring your course — and your ideas.
Explore the study →Engineering & biology faculty
Co-develop a discipline lab
Agents now drive design and simulation tools. Help us build human-in-the-loop labs for your field, starting from working prototypes.
See the prototypes →Community colleges & nonprofits
AI-skills lab partnership
A ready-made, supported lab program for nontraditional and rural learners — proposed together to the Atlassian Foundation.
See the partnership →Fall 2027 Pioneer Cohort
Tell us where you teach. We’ll tell you how the labs work.
We’re inviting instructors and department heads to co-design and teach new agent-supervision labs in Fall 2027. We share details of the curriculum and tools directly with prospective partners — send a note and we’ll set up a short call.
- No cost to explore. Materials, tools, and training are provided at or below cost to partner sites.
- Research support. Shared protocols and help with your local IRB.
- Credit where it’s due. Co-authorship and a seat in multisite grant proposals.
- Your ideas count. The network grows from what its members bring.