Coinupbtc

local AI · measured builds · open repos

Coinupbtc

I don’t rent intelligence — I host it. Local AI, measured builds, open repos.

GitHub X · @coinupbtc Email
Pseudonymous · hiring-friendly coinupbtc.com

The machine

Hardware behind the proof

GPU memory
121 GB
unified CPU+RAM, on-box
Storage
3.7 TB
NVMe, local weights
Decode
169 t/s
measured coding throughput
Art
on-metal
ComfyUI, no cloud API

Selected public work

Built here, shipped open

all repos
Local LLM serving bakeoff — measured tokens and tuning curve

miaai35-tune

Measured llama.cpp flag/profile bakeoff for Qwen3.6-35B on DGX Spark — pick serving settings with evidence, not vibes. Coding decode measured at 169 t/s, quality 9/10.

Endpoint-agnostic LLM benchmark suite — score dials and pass/fail checks

zwell-bench

Endpoint-agnostic local-LLM bakeoff — executed coding, vision, tool & agentic checks scored pass/fail. miaai35 baseline scored 95.0 objective.

Local GPU fleet dashboard — graphs, model inventory, endpoint health

spark-console

Local GPU & fleet dashboard for DGX Spark and any NVIDIA box — graphs, model inventory, endpoint health, read-only service board. 121 GB unified memory, nothing phones home.

Live Bitcoin mempool visualizer — stacked projected blocks and fee bands

bitcoin-blockfield

Live Bitcoin mempool visualizer — the next 3 projected blocks full-size, fee bands, inscription radar, driven by mempool.space-compatible APIs.

Prompt to 3D-printable STL — local CAD primitives on a slice bed

stl-sandbox

Prompt → 3D-printable STL/STEP, fully local — CadQuery primitives + optional on-box LLM. Zero cloud CAD, bed-fit checks before you slice.

The art

Made on the machine

Every image below was generated locally on my own DGX Spark — no cloud API, no rented GPU. Optical figure studies; the last image is a local systems topology diagram. click any image to view full-size

Elsewhere

Find me in the open