These four Spaces all explored an “aetheric computer”: the idea that a machine could be built
out of a higher-dimensional, self-describing substrate rather than ordinary state. Four
separate Spaces were made; none of them runs on this account, and each one still asks for a
cpu-basic allocation against an account limit of 3 concurrent Spaces.
| Space | SDK | Size | What the source actually does | Why it cannot run on a free CPU Space |
|---|---|---|---|---|
AETHEROS-X |
gradio | ≈1,295 lines | The substantial one. A “pseudoassembly” experiment: it invents token vocabularies
(ManifestAstralPlane, x86 BIOS tokens, boot-sequence tokens), maps them to
indices, and builds Keras Sequential(Embedding → LSTM → Dense) models over
them, wrapped in a BIOS/bootloader/kernel/gate-logic shell. Large parts are defined but
never reached (functions nested inside functions, model loads commented out,
if input == … comparing against the builtin). |
Imports tensorflow/keras plus sklearn,
pandas and matplotlib at module level. A free cpu-basic
image cannot install and start that stack reliably. |
AETHEREAL_COMPUTER_SYSTEM |
gradio | 72 lines | An “agent hive” demo: a business-development “agent”, then frontend and backend “agents”, chained one after another. The wording of the app is upfront that the later steps are simulated. | from transformers import pipeline at import time, loading
sentiment-analysis and distilgpt2 — model weights are fetched
during startup, which a free Space cannot do dependably. |
AetherSYS |
gradio | 47 lines | A Second Life DJ “reactionist” chat: filters the reply down to plain text and sends a fixed persona prompt. | Posts to chat.tune.app with Authorization: os.environ.get('api')
— an external API and a secret the Space does not hold, so it fails at the first
request. |
AethericGPT |
gradio | 25 lines | The smallest: forwards the prompt to a hard-coded remote host
(http://xzyorb.servemp3.com:80) and returns the text. |
A requests call to an external server that is not this Space's — plus
gr.capture_session / gr.inputs.*, Gradio APIs that no longer
exist. |
All four are dormant, and none of them is a candidate for a slot: two need an external model and a key, one needs a model download at startup, and one needs TensorFlow. Slots on this account are reserved for work that genuinely computes.
hdc-suitehdc-suite is this family's idea
implemented for real, in 1,024 dimensions and pure Python (no torch, no transformers), running
on one cpu-basic slot. It has four tabs:
bind (self-inverse), bundle,
permute and cosine similarity, with the algebra checked in the output:
bind(bind(A,B),B) == A holds to +1.0000.SET, BIND, BUNDLE, PERM,
SIM), which is the honest version of the pseudoassembly idea above: it runs
and its state is inspectable, instead of being a model that was never trained.A static Space is free for everyone, needs no build and has no cold start, and it does
not claim one of the three cpu-basic slots. Folding the family into this one page
removes 4 cpu-basic claims at zero cost, with each original URL still resolving. Every
original's source is preserved in the repository that serves this page, under
templates/legacy/<name>/, so nothing was lost — only the quota request was.