About PLANETAI
PLANETAI is a federated AI causal-inference instrument for distributed production at bioregional scale. Four pilots: Barcelona, Boston, Santiago de Chile, Bali. Built inside the Fab City network as the third generation of a measurement program that began in 2014.
What we have built
A working prototype observatory at planetai.fab.city that couples Earth foundation models — Aurora, GraphCast, Google Flood Hub — to community sensor networks (Smart Citizen, OpenAQ, iNaturalist) and the global fab-lab registry (fablabs.io). A canonical paper grounded in a decade of Fab City research and the Full Stack Metrics Framework [Vivanco 2024], operationalising Diez's Fab City Full Stack chapter (Springer 2024). A network-curated data discovery list at awesome-fabcity-data organising 32 seed sources across four pillars (Environmental, Social, Economic, Governance) and five scales (Planet, Bioregion, Region, City, Community).
Two pre-registered hypotheses anchor the science: H0-T on material throughput (with a 15% null threshold) and H0-A on action latency versus a policy baseline. We will publish the null result if that is what we measure.
Lineage
Fab City as a global initiative launches in 2014, framed around a 2054 self-sufficiency pledge. The Fab City Whitepaper (Diez 2016) introduces the PITO / DIDO vocabulary still used in the PLANETAI methodology layer. The first Fab City Index (Utopies 2018) and the second (Boeing 2024) ceiling at 37/100 for the best Western city — the public-data-only ceiling — naming the gap PLANETAI is built to close. The Fab City Full Stack (Diez, Niaros, Ferro 2024, Springer) and the Full Stack Metrics Framework [Vivanco 2024] formalise the four-pillar × five-scale measurement substrate.
The action-latency substrate is operationalised through three completed Fab City Challenges (Bali 2022, Bhutan 2023, Mexico 2024) — 187 cumulative participants, 13 host locations, 39+ prototypes — with an independent ADB ex-ante BCA on Bhutan returning a $5–12 benefit-cost ratio (Thomas 2023). PKC (Koo, Aliana & Corley 2023) is the individual-scale sovereignty-by-architecture prior that PLANETAI extends upward through five tiers.
What's next
The forward surface lives on the tracks page. Seven open work-tracks today: refining the four pilots; building a hyperlocal dashboard prototype anchored in Bali; activating the existing fab-lab network from registry to actively contributing nodes; aligning metrics with Hamburg (Helmut-Schmidt-Universität / Moritz et al. 2024) and MIT CBA; FAB26 actions in Cambridge in July 2026; explicit funding scenarios for the program's continuation under different funding outcomes; the parallel funding tree.
How this site works
Every piece of state is a YAML file in this repo. The decision log is canonical: every non-trivial decision carries prior state, change, reasoning, what it supersedes, who needs to ack, and an attribution block that names human and AI contributions distinctly. The review queue surfaces work in flight. The tracks page surfaces what's open for partner contribution with concrete one-hour, one-day, one-week entry points.
AI sub-agent attribution is not optional — PLANETAI is built with rotating AI sub-agents under Tomas Diez's direction, and partners deserve to see clearly which work was AI-drafted, which was human-authored, and where an AI's read differs from a human reviewer's read.
Who is doing this
Tomas Diez (Bali, UTC+8) leads the project as PI. Fab City Foundation is the programmatic lead. Pilot partners include IAAC + Fab Lab BCN + BSC + Barcelona City Council (Barcelona); MIT CBA + Fab Foundation (Boston); UC Chile + Núcleo Milenio FAIR + CENIA (Santiago); IT Del + CAST Foundation + Meaningful Design Group / Fab Lab Bali (Bali). Methodology, observatory and metrics work draws on continued collaboration with Tomas Vivanco, Lauri, Lucas Marangoni, Guillem Camprodon, Ben Koo, and a rotating set of researchers, fab-lab operators and city open-data leads across the network.
Last updated 2026-05-05 · v0.1