University Portfolio

University of Kansas · School of Architecture & School of Engineering

Review of mini-agents

← All assignments
Summary

This assignment reviews custom-built mini-agents such as Custom GPTs (or comparable setups like Google Gems): how they are scoped, prompted, and evaluated in practice. The coursework asks for hands-on use of two public GPTs (one from the featured gallery, one aimed at the AEC industry), share links to those conversations, and written commentary on what worked and what did not.

Below, the three deliverables are summarized, a purpose-of-assessment exchange documents a custom mini-agent concept for basement conditions in a historic building (adaptive reuse, codes, life safety, and preservation), and the paragraph submitted with the assignment reflects on what it feels like to rely on an agent day to day.

Three deliverables (course outline)

  1. Use and review a Featured GPT. Browse public Custom GPTs at chatgpt.com/gpts (or a similar non-OpenAI setup). Pick one that interests you, use it for its intended purpose, and work through enough prompts to get a real result. Deliverable: link to the conversation (10 pts).
  2. Use and review an AEC-focused GPT. Find a GPT meant to help AEC professionals in any capacity. Use it as intended, including follow-up prompts if needed. Deliverable: link to the conversation (10 pts).
  3. Commentary on both GPTs. Brief notes on what worked, what could improve, and whether a Custom GPT is the right tool for each task. Deliverable: comments in the course portal, a short document upload, or similar (10 pts).

The featured GPT does not have to be AEC-specific; the point is to see how different domains structure agents and to borrow ideas.

Purpose of the assessment: custom mini-agent overview

The screenshot captures a shared-style ChatGPT thread titled Purpose of the Assessment. The user side frames a systematic approach to collecting data, documenting existing conditions, and identifying problems and solutions for a basement in a historic building, including adaptive reuse, building codes, life-safety requirements, and historic preservation. The assistant responds with a structured overview: the problem the agent solves, primary users (architects, preservation specialists, inspectors, owners, students, and related roles), the knowledge the agent should hold (codes, egress, accessibility, moisture and structure, preservation guidance, community context), and expected outcomes (organized findings, problem framing, and reuse strategies that stay compliant).

Screenshot: ChatGPT conversation on Purpose of the Assessment for a historic basement mini-agent, showing the user prompt and assistant overview
Assessment prompt and assistant summary for the historic basement / adaptive reuse mini-agent concept.

Submitted reflection

Links & files

Continue