132
Active practice rooms
§ 01 — A field manual for sustained AI practice
sustainer.ai is the public atelier of EARNEST, the Research Laboratory for Advanced Social Robotics at PolyU School of Design. Six practice surfaces — chat rooms, project boards, conversational avatars, publication archives, job calls, and live workshops — built for researchers, caregivers, educators, and field practitioners working over time, not in a hurry.
§ 02 — Manifesto, in one sentence
The most valuable software is not the one that completes the task — it is the one that sustains the practice.
Each board is its own surface, with its own conventions, but they share a single membership, a single archive, and a single set of ethical guardrails. Move between them the way a scholar moves between desk, studio, and seminar.
A kanban for slow research and collective making.
Columns for inquiry, drafting, field test, and publication. Cards carry citations, field notes, and AI co-authoring threads. Built on the Fizzy core, tuned for studios where work travels in seasons rather than sprints.
A seminar room where many minds — human and otherwise — meet.
Multi-agent rooms with named facilitators, students, and archivists. Built on the classroom-conversations orchestrator with 15 distinct personalities and Discourse-grade archiving. Pace and turn-taking are first-class concerns.
Real-time conversational presence, in the flesh of a face.
A GLB-driven avatar with 52 ARKit blendshapes, streamed at 30fps over a 212-byte wire frame. Built for sustained one-to-one sessions: office hours, oral histories, language exchange, interview practice.
A classifieds page for the slow, careful kind of work.
Calls for postdocs, field coordinators, designers-in-residence, and visiting practitioners. Each listing carries an institution, a duration, and a public stipend — no salary obfuscation, no silent listings.
Lead a two-year study on multi-agent classroom facilitation in upper-secondary contexts. Co-author with the Discourse cohort team.
Run six workshop cohorts across three time zones, build the field notebook, and keep the publication board honest.
Shape the visual language of the avatar board — face, voice, room. Comfortable working with shaders, ARKit rigs, and slow design reviews.
Bring a working practice — teacher, librarian, museum educator — into the chat board for a season. Stipend, travel, and publication support.
A live timetable of seasons, sessions, and studios.
Weekend studios, weeklong workshops, and asynchronous reading groups. Each session is taught by working practitioners; each cohort caps at twenty so the rooms stay warm.
A short reading of the founding charter and a walk through the six surfaces.
Build a five-bot seminar room and run a 30-minute studio with a real text. Bring the chapter, leave with a transcript.
Bring your blendshape rigs, your latency graphs, your half-finished interview practice modules.
An archive of what was thought, in working order.
Field notes, working papers, and finished pieces. Every publication carries its room of origin, its co-authoring history, and a public link back into the board where it took shape.
We argue that the dominant SaaS framing — software as a task-completer — fails to support the long, recursive work of careful inquiry. We offer the “board” as an alternative unit of design…
sustainer.ai is the public surface of the Research Laboratory for Advanced Social Robotics — EARNEST, founded in 2025 at PolyU School of Design. The laboratory inherits three antecedent institutions, and carries their distinct strands into a single working programme.
“A society wherein each individual possesses a profound comprehension of the capabilities and constraints inherent in robotics and artificial intelligence — particularly within the social sphere.”
Through research and design, the laboratory cultivates intellectual leadership that informs both present and prospective endeavours concerning robotics and AI. Such guidance aims to inspire individuals towards developing social robots that are beneficial in areas where human abilities may be insufficient — physically or intellectually — without becoming instruments of exploitation or peril.
“To cultivate independent thought and to demystify the nature of robots and artificial intelligence as technologies comprehensible from foundational principles.”
The lab is devoted to fundamental research and theory development in probability, true randomness, information entropy, and epistemics. This work underpins the design of systems capable of navigating complex scenarios and making informed decisions that surpass human capabilities, all while ensuring safety and security — with humans retaining ultimate decision-making authority.
§ Topics under foundational study
From classical to quantum probability, with a focus on decision making under genuine uncertainty.
Sources of irreducible randomness — and what their absence means for the determinism of robotic agency.
Measures of uncertainty as a primitive for both decision policy and observer-state estimation.
Algorithmic substrates for decision and inference where classical probability falls short.
New statistics for the study of the rare — the edge cases that classical methods discard.
How knowledge is acquired, justified, and held — and what a machine can be said to know.
§ Featured use case
The descendant of SELEMCA's Alice R50, reframed as a universal interface across the IoT-equipped home: a Caredroid that listens carefully, retrieves trustworthy information, and acts only with the resident's consent.
EARNEST's design discipline holds that interactions must be personalised, culturally attuned, and safeguarded by robust data security — including blockchain-backed audit trails — so that the robot can become a trustworthy confidant and advisor, free from the distortions of generative hallucination.
The laboratory keeps an open studio for cohort members, visiting practitioners, and anyone with a serious question about social robotics in care. Drop-ins are welcome during posted office hours; deeper sessions by appointment.
You arrive through a conversation, not a form. Onboarding is a twenty-minute talk with our editor-in-residence; from there, you choose your boards and your cohort. A season runs twelve weeks.
A twenty-minute typebot interview with the editor. We ask about your practice, not your job title.
You are routed to two boards, a cohort, and a reading list. A welcome card is posted in your inbox.
Daily rooms, weekly studios, season-long projects. You bring the work; we keep the rooms warm.
End each season with a field note, a workshop, or a working paper. Filed openly on the publications board.