Iteration 2 · Mixed Engineering · Spring 2025

C.3.1 Institutional context

Iteration 2 ran from January 13, 2025 through April 30, 2025 as GEEN 3830-001 Special Topics in the College of Engineering and Applied Science (SY-2). The course was three credits, held in classroom DLC170, with an audience of approximately twenty-five undergraduates of mixed engineering backgrounds (computer science, mechanical, electrical, civil, and aerospace engineering, among others). The institutional move from ATLAS in Iteration 1 to the College of Engineering in Iteration 2 widened the student audience from the creativity-and-design CTD cohort to a broader engineering population.

This was the consolidation phase of my pioneer practice. I had spent a fifteen-week semester running Iteration 1 and had the contemporaneous reflective journal (WU-1 series) and the curricular artifacts (CV-1, DK-1.FG, DK-1.W01, FP-1) to draw on. Iteration 2 was the opportunity to test the four-theme architecture and the curriculum sequence against a new student population, with substantial tool turnover already required by the year of generative-AI change between Spring 2024 and Spring 2025.

C.3.2 Week 1 · January 13, 2025 · Prompt engineering opener (same as Iteration 1)

Week 1 of Iteration 2 opened with prompt engineering. The opening deck (DK-2.JAN13) preserved the structural commitment of Iteration 1: prompt engineering as the entry point, the Midjourney Self-Portrait Assignment as the Week 1 or Week 2 image-generation exercise, and named student outputs as the visible trace of the dialogue-with-informants criterion.

The Midjourney Self-Portrait Assignment migrated forward to Week 1 itself in Iteration 2, slightly earlier than in Iteration 1 (where it was a Week 2 assignment). The Iteration 2 Midjourney deck (DK-2.JAN15) carries the assignment's framing for the engineering cohort.

The stability of the prompt-engineering opener across both iterations is one of the cross-iteration stable elements documented in §C.6.5.

C.3.3 Week 3 · DeepSeek added within weeks of public release

DeepSeek released its R1 model publicly in late January 2025. By Week 3 of Iteration 2 (early February 2025), I had incorporated DeepSeek into the curriculum. The Canvas export (CV-2) records the addition, and the syllabus (SY-2) lists DeepSeek among the tools students would engage. My dissertation chair Tom Yeh delivered a guest lecture on DeepSeek on January 28, 2025 (visible in CV-2 and on the Iteration 2 guest-lecturer spreadsheet AS-2.GUESTS).

DeepSeek's incorporation within weeks of its public release is one of the documented instances of immediate tool adoption that the cross-iteration analysis (§C.6.1) treats as evidence of how pioneer practice handles fast tool turnover.

Iteration 2 also added Claude 3.7 and Grok 3 as named tools in the Week 8 readings (per CV-2 module structure), bringing the major proprietary frontier-model line-up explicitly into the curriculum.

C.3.4 Weeks 7 to 9 · AI Agents as a full new module

Iteration 2 introduced a full module on AI Agents in Weeks 7 through 9, with the explicit assignment "Create your own AI Agent" as a hands-on component (per CV-2 module structure). The module did not exist in Iteration 1. The addition reflects how the field had matured between Spring 2024 and Spring 2025: AI Agents had become a workable pedagogical topic with concrete tool support and an addressable assignment scope, neither of which had been clearly possible a year earlier.

The AI Agents module is the single largest curricular addition between Iterations 1 and 2 and is one of the structural changes documented in §C.6.2.

C.3.5 Guest speakers as curated network in Iteration 2

The guest-speaker roster in Iteration 2 differed substantially from Iteration 1's, and the difference is itself part of the iteration's pedagogical contribution. Rather than treating guests as visiting one-offs, I curated the iteration's roster deliberately to track the changing technology landscape (DeepSeek and AI Agents were not topics anyone could have credibly taught in Spring 2024), to maintain CU-internal continuity (Nolan Brady recurred from Iteration 1 with a different topic), and to bring industry practice into the classroom (Justin Shacklette on GenAI for Software). The administrative spreadsheet AS-2.GUESTS records the iteration's guests; the four central appearances were:

The curation work is the contribution at the network level, not just at the per-lecture level. Students did not encounter generative AI as an abstract category; they encountered it as a set of named practitioners and named topics actively assembled around the iteration's curriculum. The contrast with Iteration 1's roster — almost complete external-guest turnover, with CU-internal speakers stable but rotated to different topics — is documented in §C.6.3 as evidence of how the network is actively curated rather than passively maintained.

C.3.6 Student teach-out presentations

A pedagogical move new to Iteration 2 was the student teach-out: a class session in which a student took the front of the room and taught the class on a topic of their choosing. Six student teach-out presentations are preserved in the archive:

• SP-2.DEEPFAKE (Daniel Debretsion): DeepfakeAI as an Ethics-and-Industry topic
• SP-2.SINTRA: Sintra.ai as an Industry-and-Education topic, in Alt Style format
• SP-2.HALEY (Haley Phillips, two files)
• SP-2.DAKOTA (Dakota A)
• SP-2.ROBOTICS: The Integration of AI and Robotics as an Industry topic

The teach-out distribution across topics shows engineering-oriented choices weighted toward Industry-theme applications, with the DeepfakeAI presentation pulling Ethics-theme attention through the Industry surface. The teach-out artifacts are part of the dialogue-with-informants-beyond-self evidence base.

The administrative spreadsheets AS-2.TEACHOUT-DATES and AS-2.TEACHOUT-INDEX record the iteration's teach-out schedule and slide locations. The final-presentations schedule AS-2.FINAL-SCHED records the iteration's end-of-semester capstone presentations.

C.3.7 Structural reshuffles from Iteration 1

Iteration 2 was not Iteration 1 repeated. Several module-placement and curricular moves changed between iterations:

The reshuffles are documented in §C.6.2 as the module-placement-evolution pattern.

C.3.8 What I learned across Iteration 2

Iteration 2 had less structured contemporaneous reflective journaling than Iteration 1. The Keep Up Newsletter (KN-EP1, KN-EP2, KN-EP3, April through May 2025) and the Keep Up Podcast (KP-EP2, KP-EP3, May 2025) carry my public-facing reflective writing across the Iteration 2 period, and analytic autoethnography permits this retrospective-public frame when it is named (§B.6.3). Three observations drawn from the iteration's artifacts and from these reflective channels are worth marking here.

First, the four-theme architecture survived the audience shift from CTD to mixed engineering. The Iteration 2 syllabus (SY-2) explicitly schedules the themes (Weeks 1-4 Education, Weeks 5-11 Industry, Weeks 12-15 Accessibility, with Ethics cross-cutting), and the iteration's curriculum delivered against the schedule. The architecture is not dependent on the creativity-and-design student type that Iteration 1's CTD cohort represented; it holds against an engineering-student type as well.

Second, the tool turnover pattern became visible. Adding DeepSeek within weeks of its January 2025 release (CV-2 records the Week 3 addition), bringing AI Agents in as a new module, adding Claude 3.7 and Grok 3, NotebookLM, Be My AI, Sora, HeyGen, ElevenLabs, and Custom GPTs with Wolfram API constituted a substantial tool refresh between iterations. The architecture absorbed the refresh without re-design of the themes.

Third, the cross-iteration guest-speaker turnover pattern became visible. External guests from Iteration 1 (Anthony Pinter, Diane Sieber, Matt Zago, Nikolaus Klassen, Suibi Weng, Daniel Ritchie) did not return as live guests in Iteration 2; CU-internal guests (Nolan Brady, joined now by Tom Yeh, Bobby Hodgkinson, and Justin Shacklette) carried the iteration. The pattern is documented in §C.6.3 and reflects how the instructor's actively-curated guest network adapts to the iteration's topic structure.

Iteration 2 completes the semester-length consolidation phase. The next two iterations (§C.4, §C.5) compressed the consolidated curriculum into five-day online workshops, opening the distillation phase.