Bkpaine · INDEPENDENT BUILDER

Useful tools.
Continuing minds.
One thing at a time.

I build software that remembers, tools that take work off your hands, and experiments that find out what the hardware can really do.

Explore MindStrix Read the writing →

MINDSTRIX BUILD STUDIO
Building since 1995

THE WORK

Built, used, and still getting better.

OPEN-SOURCE PUBLIC BETA

MindStrix

Rent the Brain.
Own the Mind.

Give an AI a name, a place to remember, and tools to do useful work. MindStrix keeps identity, graph memory, and conversations on your computer, so changing model providers doesn’t mean starting over.

Work through tasks together. Develop reusable skills and tools you can review. Keep the decisions that let you pick up where you left off.

Get MindStrix ↗Read the launch story →

An early beta for Linux, with persistent Neo4j memory. Bring your own model connection; cloud API usage is billed by your provider.

MindStrix: the model supplies the brain; identity and memory form the continuing architecture

Persistent memoryReviewed toolsChoice of models

AVAILABLE NOWLive

StrixScreener

Find the substance beneath the résumé.

Rank a stack of résumés against a job description, see the reasoning, and export a CSV for your own review. Every candidate stays in the list. You make the hiring decisions.

$1.49 per résumé · No subscription · Up to 50 per batch

Visit StrixScreener ↗

NEXT DEVELOPMENT FOCUSPrototype

Pixie Bear

A voice companion, brought into the room.

A warm companion built around dedicated voice hardware. The current work is on the basics that make conversation feel dependable: hearing a wake phrase, listening clearly, and responding aloud.

In development. Real-world wake and hearing reliability are still being tested.

From the workbench · No release date yet

ON THE WORKBENCH, IN WRITING

The story behind the build.

All 13 articles →

MindStrix · Sep 10, 2026

Rent the Brain. Own the Mind.

Why I built MindStrix: a continuing AI companion with persistent memory, useful tools, and work you can inspect.

Read article 11 min

Strix Halo · Jul 24, 2026

I tuned llama.cpp on a Strix Halo mini-PC and it beats Ollama on the same weights — here’s the recipe (and why “Vulkan beats ROCm” is a myth)

Tuning llama.cpp on AMD Strix Halo, with build flags, measurements, and a comparison on identical model weights.

Read article 5 min

Strix Halo · Jun 29, 2026

Ollama Vision Is Slow on Your AMD APU? It’s Running the Image Encoder on CPU — Here’s the 4x Fix

Tracking a slow vision encoder to CPU offload—and testing the difference when the GPU does the work.

Read article 7 min

THE PERSON BEHIND THE OWL

Still curious.
Still building.

I’m Bkpaine. I’ve been building on the internet since the days of dial-up. These days, my work centers on continuing AI companions, useful software, and making capable local hardware earn its keep.

MindStrix is out in public beta. Pixie Bear is the next focus. StrixScreener is available for the work it was built to do. The experiments and lessons live here too, so the next person has somewhere to start.

Explore the technical guides →Find the code ↗

Other work from the studio

MultiScan — credentialed network configuration capture. The storefront remains a development project; it is not currently offered for purchase here.

AMD Strix Halo experiments — local inference, AI video, voice synthesis, and autonomous ML experiments. The writing archive and technical guides hold the findings and links to code.