MeaningX is a research and practice collective working at the intersection of design and AI. We hold that the meaning of what technology produces is not computed, but designed. Accordingly, we do two things: build a design-driven, cross-disciplinary cluster of agents, and cultivate an open benchmark community that measures the plural values of design.
Design pluralism, evaluation methods for design agents, and value alignment in human–AI co-creation. Research grows out of practice; practice ripens into research.
TT Design Buddy is our first work: a design agent that understands design and critiques with evidence — it does not decide for you.
OpenStudio: an open benchmark community for design agents. The content is under construction; the position is published first.
Not one resident design companion, but many. They draw on knowledge distilled by 100+ design experts across design education, research, and practice worldwide — they do not decide for you. They examine every divergence, and explain why.
Expertise from leading design educators and practitioners worldwide. The first ten domain agents (design history, ecological design, sustainable design, interaction design, among others) are live on TT Design Society; Buddy brings them into everyday creative practice.
Large-model analysis of source material: collect, understand, and apply inspiration — turning vague impressions into traceable reasoning.
Design learning, inspiration gathering, and creative practice as one continuum — from question, to inspiration, to artifact.
Every critique cites a knowledge-base entry, a precedent, or a standard — traceable and contestable. “It feels off” is not an argument.
An open benchmark community built for design agents. We are building datasets, a knowledge graph, benchmark sets, and tooling.
“What counts as good design” is itself a design question. We refuse to compress design into a single score or a loss curve. The task of a benchmark is to expose disagreement, not to erase it — the leaderboard should have ties at first place, together with their reasons for objecting to one another.
Evaluating design agents requires designers, engineers, anthropologists, and real users at the same table. Disciplinary boundaries are meant to be crossed, not hidden behind. The jury is always larger than the lab.
Every benchmark quantifies something someone cares about. Declare what you care about before claiming to measure it well. Quantification exists to make values discussable — not to make value judgments on our behalf.
We likewise evaluate how results are reached: the breadth of exploration, the reasoning behind trade-offs, the options set aside. A transparent process merits the same score as a beautiful outcome.
Tasks, data, and grading rationales are all open. We can be outperformed; we refuse to be a black box. Any agent’s score must be fully replayable — and challengeable — within the community.