mirror of
https://github.com/lux-apps/papers.git
synced 2026-10-30 18:15:21 +00:00
zeekayandHanzo Dev
cf44d44756
ADOPT-00 through ADOPT-10, the 18-page omnibus, the decentralization rubric, the whitepaper and the LOI — every paper written for the Morpheus engagement, with the shared cover and proposal styles, versioned here beside the LP series. Co-authored-by: Hanzo Dev <[email protected]>
84 lines
4.8 KiB
TeX
84 lines
4.8 KiB
TeX
\documentclass[11pt,a4paper]{article}
|
|
\usepackage{../shared/luxcover} % black cover (self-contained, TikZ logo, no assets)
|
|
\usepackage{../shared/proposal} % \proposalhead helpers, stake environment, muted colour
|
|
\usepackage[utf8]{inputenc}\usepackage[T1]{fontenc}
|
|
\usepackage{amsmath,amssymb}
|
|
\usepackage{listings}\usepackage{xcolor}\input{../shared/lstlang}
|
|
|
|
\newcommand{\code}[1]{\texttt{\small #1}}
|
|
\newcommand{\prov}[1]{{\color{muted}\scriptsize\sffamily[#1]}}
|
|
\setlength{\emergencystretch}{3em}
|
|
|
|
\title{Real Demand and Attested Supply}
|
|
\author{Zach Kelling\\\textit{Lux Industries, Inc.}\\\texttt{[email protected]}}
|
|
\date{2026}
|
|
|
|
\begin{document}
|
|
\luxcoverpage
|
|
\sloppy
|
|
|
|
\noindent{\color{muted}\footnotesize\sffamily ADOPT-07 \quad\textbullet\quad Hanzo AI \& Lux Industries \quad\textbullet\quad 2026}\par
|
|
\vspace{0.9ex}
|
|
|
|
\noindent\textbf{The number.}
|
|
Morpheus has never published a usage number. No session, price, or revenue series exists
|
|
in any of the 62 repositories \prov{verified}. Today the network routes about zero verified
|
|
inference sessions per day \prov{verified}. Real revenue is Lido and Aave yield, about
|
|
\$476K/yr and falling, down from \$3.24M in the first quarter after the 2024 launch
|
|
\prov{verified}. Emission runs roughly ten to one over it \prov{verified}. The 18 providers
|
|
on \code{active.mor.org} largely resell Venice and OpenAI keys \prov{verified}.
|
|
|
|
\vspace{0.35ex}
|
|
\noindent\textbf{The problem, in their numbers.}
|
|
On chain a provider is a base URL and a bearer token. The node dispatches every request to a
|
|
configured \code{ApiURL} (\code{ai\_engine.go:64}) and POSTs to it (\code{openai.go:64}).
|
|
Settlement is \code{pricePerSecond} times elapsed time (\code{SessionRouter.sol:264}); the
|
|
model and the token counts never enter it \prov{verified}. So a datacenter GPU and a resold
|
|
ChatGPT subscription settle the same, and a 7B served where a 70B was sold pays the same. The
|
|
capital layer is sound: principal is withdrawable one to one, and 75\% of yield funds a real
|
|
buyback, about \$511K over the trailing year \prov{verified}. The honest problem is narrower.
|
|
Dilution outruns revenue, and the meter is fakeable.
|
|
|
|
\vspace{0.35ex}
|
|
\noindent\textbf{The fix, named to a repo.}
|
|
Hanzo connects on both sides with no protocol change. As demand, Enso, Hanzo's learned router,
|
|
adds Morpheus providers to its candidate set and sends real requests to the best-value bid,
|
|
opening ordinary sessions on Base \prov{verified}. As supply, Hanzo runs a provider node whose
|
|
backend is one models-config entry pointed at \code{api.hanzo.ai}, its attested
|
|
OpenAI-compatible endpoint; the node is \code{Morpheus-Lumerin-Node/proxy-router} and nothing
|
|
in it changes \prov{verified}. Hanzo attests its GPUs on a root of trust it hosts itself:
|
|
\code{lux/ai} checks SPDM measurements against local Reference Integrity Manifests, no cloud.
|
|
The node's own verifier trusts three remote roots instead, NVIDIA's attestation service
|
|
(\code{nras\_verifier.go:17}), a third-party quote-parse portal (\code{verifier.go:80}), and a
|
|
measurement registry on a CDN mirror (\code{artifacts\_registry.go:20}), any of which can 404
|
|
or change its answer \prov{verified}. \code{mord} meters capacity, work, and identity as three
|
|
quantities and pays work only on counts the consumer also signed; faked work earns zero and a
|
|
substituted model is slashed. It is built, 40 tests passing \prov{verified}.
|
|
|
|
\vspace{0.35ex}
|
|
\noindent\textbf{The outcome, projected.}
|
|
Assumptions stated. Launch reach across \code{hanzo.chat}, \code{hanzo.app}, \code{hanzo/dev},
|
|
\code{lux.exchange}, and \code{lux/wallet} is 100M \prov{client}. Assume 2\% turn monthly active
|
|
(2M), 2\% of those opt into a private-inference tier (40k), and half route (20k)
|
|
\prov{illustrative}. Supply bounds the near term at 18 providers: 1,000 to 10,000 verified
|
|
sessions per day \prov{illustrative}. The low case, about 10,000 per day, still exceeds all
|
|
currently published Morpheus demand, which is about zero \prov{verified}. Route a small fee on
|
|
verified volume into the buyback Morpheus already runs, and revenue scales with usage, not
|
|
deposits.
|
|
|
|
\vspace{0.35ex}
|
|
\noindent\textbf{Relevance, and first step.}
|
|
No other counterparty brings both a live 100M-reach consumer funnel and an un-fakeable meter to
|
|
point at this network. It adopts today as configuration: \code{postModelBid} is permissionless
|
|
for an active provider (\code{Marketplace.sol:63}), so no facet upgrade and no 5-of-9 Safe
|
|
signature \prov{verified}. Demand and verifiability land first; governance, listing, and lending
|
|
rails second; an L2 and a raise third, cleared only into demonstrated on-chain revenue and a KYC
|
|
base already onboarded through Hanzo IAM and \code{lux/wallet}.
|
|
|
|
\vspace{0.6ex}
|
|
\noindent{\color{muted}\footnotesize\sffamily
|
|
Diamond \texttt{0x6aBE1d282f72B474E54527D93b979A4f64d3030a} on Base. Every citation checked
|
|
against the working tree.\quad\textbullet\quad \texttt{[email protected]}}
|
|
|
|
\end{document}
|