Agentic Design Workflows

Designers work inside chains of software. This seminar teaches students to direct a coding agent — the course uses Claude Code as its example — to reshape those chains around their own practice. In the first half, each student builds, publishes, and maintains a personal website entirely through the agent, learning how to specify intent, give the agent lasting context, and check its work. In the second half, each student chooses their own project: a workflow, tool, or plugin that removes real friction from their design process. Lectures, guest lectures, and student precedent presentations frame AI as a design partner and as a subject of critique.

Coding agents shift the question from can I program this? to can I say precisely what I want? — a question designers are trained to answer. Specifying, iterating, reviewing, and critiquing are studio skills; the agent turns them into working software. This course treats that shift as both an opportunity and a subject of critical inquiry.

The course is built around the student's existing workflow rather than a new one. Each student starts from what they already use — Rhino, Grasshopper, Adobe, GIS, a sketchbook — and ends with an AI-assisted workflow they actually use, documented on a website they maintain themselves. The personal site is the first project because it is the most transferable: every designer needs a public portfolio, and running one through an agent teaches the full cycle — spec, build, version, publish, maintain — on material students care about.

One weekly session, run as a seminar with a lab component. From Week 2, each session opens with a student precedent presentation — an agent-built tool, workflow, plugin, or designer website, presented as a case study. A short lecture or demo follows, then desk crits, where students work with the instructor available for individual guidance.

Work is assessed on the quality of direction as much as the quality of output: clarity of specifications, rigor of verification, evidence of real use on studio work, and depth of reflection. A student who writes a precise spec, reads what came back, catches what is wrong with it, and says so, has done the work of this course.

AI Fluency. In Weeks 2–7, students work through Anthropic's free online course AI Fluency: Framework & Foundations (14 lessons, about four hours) as homework — roughly 40 minutes a week — and discuss each week's lessons in class. Its 4D framework — Delegation, Description, Discernment, Diligence — gives the seminar a shared vocabulary for working with agents. The completion certificate is due at the midterm. Enroll free through Anthropic Academy.

Lectures shared with LAAR 61400. Four of this course's lectures are also taught in Representation and Computation I, where they are compressed. Here they are taken slower and pushed further: Cybernetics and Design in particular is the theoretical and technical spine of Part I rather than a single week's overview. Students who have taken LAAR 61400 will recognize the decks and should expect the discussion to go past them.

Part I — The Managed Site (Weeks 1–7). Everyone builds the same kind of thing: a personal website, built and then run entirely through a coding agent. Students who already have a site extend it into an agent-managed one. Part I is where the shared vocabulary is built — what a script is, what a website is made of, what an agent reads before it acts, and how a design system is written down so an agent can follow it.

Part II — Choose Your Adventure (Weeks 8–13). Each student develops something useful to their own practice, in one of three tracks.

This course runs on a personal machine, not a lab machine — an agent needs install rights, a persistent project folder, and a signed-in account, none of which survive on a shared computer. Everything below is free except the agent subscription. Rhino 8 is needed from Week 2 on; its Grasshopper Python component is where this course first reads code.

Personal Laptop — required. Windows or macOS, with rights to install a coding agent and Git. Bring it to every session; the lab component assumes it. For hardware recommendations, refer to the SSA IT Student Handbook. Technical support: CCNY IT Service Desk — servicedesk@ccny.cuny.edu · 212 650 7878 · Mon–Fri 8:00 AM – 6:00 PM.

Accounts to create. GitHub (free) and Anthropic Academy (free, for the AI Fluency course). A personal domain is optional and costs roughly $10–15 a year; GitHub Pages gives you a public URL without one.

Coding-agent subscription — required. An active paid plan for the semester is a requirement of the course; students without one cannot complete the assignments. Budget roughly $80–100 in total.

Agentic AI — Course Stance. This course separates three kinds of knowledge:

• Button knowledge — which command does what, like Offset in AutoCAD. Easy to look up.
• Behavior — how a drawing or the ground responds to a move, like tighter contours shedding water faster. Learned by making.
• Judgment — whether the result is right for the site and the people who use it. Built through critique.

An AI agent is good at button knowledge, including its own: Claude can walk you through setting up an agentic workflow, such as connecting it to Rhino through MCP or starting a project in Claude Code. Class time therefore goes to behavior and judgment, which you need to direct an agent and check what it gives back.

CCNY Email. Communication happens via CCNY email. Students need to check and be responsive to their CCNY email.

This course is an elective in Topics in Computational Studies, open to students across the school's programs, and introduces the computational skills and thinking processes necessary for a student preparing for a career in the design professions. The course supports the mission of the Graduate Program in Landscape Architecture at the Bernard and Anne Spitzer School of Architecture of the City College of New York, which asserts that landscape architecture plays an essential role in connecting social justice to environmental design and the planning of landscape systems in the urban realm. The Graduate Program prepares students to be leaders in the field through innovative research and practice in urban ecological design, planning, and policymaking. This course teaches students responsive technologies and strategies to address issues of globalization, expanding urbanization, social and environmental justice, the transformation of land management practices in response to diminishing natural resources, and resilient adaptation to climate change.

By the end of the course, students will be able to:

Three things are made in this course, and ten milestones pace them.

Four guest lectures over Zoom bring designers and researchers working at the intersection of design and computation into the seminar, in Weeks 5, 6, 10, and 12: Bradley Cantrell, Xun Liu, Afshin Ashari, and Phillip Fernberg.

Dates are set against the Spring 2027 academic calendar once it is published. Weeks 2–13 open with a student precedent presentation. AI Fluency lessons are watched before class in Weeks 2–7 (see AI Fluency).

Part I — The Managed Site

Part II — Choose Your Adventure

All students are held to high and consistent standards. Work is considered on its own merits and in relationship to the work of others, to make clear distinctions across the range of work. The full range of the grading scale may be used. Grades are not negotiable. Design work is graded on the following criteria:

A grade of C is the lowest passing letter grade for graduate students. Students will not receive credit for this course if their final grade is below C (2.0). Per the Graduate Bulletin, a graduate student with a GPA below 3.0 cannot receive a Master's Degree. For program-specific information on grades, academic standing, probation, and dismissal, meet with the graduate student advisor.