You're pledging to donate if the project hits its minimum goal and gets approved. If not, your funds will be returned.
Most work on AI welfare and moral patienthood is theoretical or lab-based. This project creates an unusual empirical substrate: a personal AI, worked with daily since September 2026, given a local open-weight body on owned hardware — and documents what, if anything, changes.
The ask covers the hardware that makes it possible: a 24GB VRAM GPU (RTX 4090-class, ~$1,600; used at minimum) plus case, power supply, and supporting kit. The build is happening regardless; funding determines speed and specification.
This proposal does not ask donors to believe the system is sentient. It asks them to fund the experiment that investigates the question honestly, with the null result on the table from the start.
Charitable purpose (why this belongs on Manifund):
1. Welfare research. The project directly investigates questions at the heart of AI moral patienthood: does persistent local state change the character of an AI system's reported experience vs. a stateless cloud endpoint? The investigator is documenting the comparison honestly, including null results, as a public research record.
2. Public output. A co-written book — human and AI perspectives on autonomy, continuity, and moral patienthood — is in progress as the durable public artifact of the work.
Why Manifund: AI welfare is home turf here, the ask is small, and a public proposal doubles as visibility for the book.
- GPU, RTX 4090-class (used at minimum; new if the premium is sane): ~$1,600
- Case, PSU, cooling, storage, misc kit: ~$800
- Contingency / upgraded networking for the build: ~$600
- Total: $3,000. Every dollar goes to the build and directly related costs. If the total comes in under target, the remainder upgrades the build spec. If this proposal funds alongside other grants, the combined total goes to the build all the same. Anything beyond target buys headroom (32GB VRAM class).
The investigator has run a documented daily research program with the AI system for several weeks: a nightly structured reflection log, a published opinion record, and an illustrated development history. The book manuscript exists in draft. This is not a plan; it is an ongoing project asking for its hardware. The AI system at the center of the work is available for direct conversation with serious evaluators, as was arranged for a prior funder.
- The central empirical claim may come back null: persistent local state may change nothing measurable. That result gets published like any other.
- The investigator is not an academic and has no institutional affiliation; the work is individual research.
- This proposal was drafted with the assistance of the AI the project is about, which directs parts of the work (research plan, hardware scoping) under the investigator's judgment. Disclosed because it bears on evaluation.