0xSero

Products and projects

0xSero's main project is the Local AI Registry, which pairs each open model with an inference engine and settings for a specific GPU, keeping one validated recipe per machine with the evidence from the run attached. Recipes are written by a lab script only after the model loads, chats, reasons, calls tools, holds its context window, and keeps pace on the named card. The repository is MIT-licensed and now sits under the Sybil Solutions organization on GitHub.¹⁹

The registry's catalog is published at local.sybilsolutions.ai, where people can look up tested models by GPU for use with coding agents such as Claude Code and Codex. The companion local-ai-images repository publishes the container images that the recipes pin, and Omarchy's Local AI plugin reads the registry's recipe export.¹⁰³¹

Open-source work

0xSero also commits to Local Studio, Sybil Solutions' MIT-licensed control panel for launching, monitoring, and chatting with self-hosted model backends such as vLLM, SGLang, llama.cpp, and ExLlamaV3. The registry's engine profiles cover vLLM, SGLang, and TabbyAPI with ExLlamaV3.⁴⁵²

Sponsorship

In September 2026, the Omacom Foundation announced it would be a premier sponsor of 0xSero's local AI work for three years, through 0xSero's company Sybil Solutions. The collaboration aims to make local models work well out of the box on Omarchy, from picking a model that fits a machine to tuning it and connecting it to agents.⁶⁸

Work and contributions

• Local AI Registry

  Publishes and commits to the registry. Its recipes are written only by a lab script that runs the model on the named card and records the recipe if it loads, chats, reasons, calls tools, holds its context window, and keeps pace. Omarchy's Local AI plugin reads the registry's recipe file, and the catalog is also served at local.sybilsolutions.ai. MIT-licensed.¹⁹¹⁰

• vLLM engine profile in the Local AI Registry

  The registry's vLLM engine profile pins the upstream vLLM OpenAI-compatible server image by digest and defines the serving arguments that the registry's vLLM recipes fill in for each card.¹¹²

• llama.cpp images for the Local AI Registry

  Publishes the container images the registry pins, including one that ships stock, unmodified llama.cpp from the ggml-org release, verified against its published checksum, for standard-quantization GGUF recipes.³

• Local Studio

  Commits to Local Studio, Sybil Solutions' MIT-licensed workstation for launching, monitoring, and chatting with self-hosted model backends such as vLLM, SGLang, llama.cpp, and ExLlamaV3.⁴⁵⁷

Timeline

1. Dec 17, 2025Sybil Solutions created the Local Studio repository, a control panel for self-hosted model backends.⁴
2. Aug 26, 2026Created the Local AI Registry repository of validated, GPU-specific model recipes.¹
3. Sep 2, 2026Created the local-ai-images repository of container images pinned by the registry.³
4. Sep 23, 2026The Omacom Foundation announced a three-year sponsorship of 0xSero's local AI work.⁶

Where they work

• Sybil Solutions 1 person in the wiki

Related people

• Luis Héctor ChávezDeveloper tools · Inference and serving
• Erik GafniDeveloper tools · Inference and serving
• Daniel BeveniusInference and serving · Developer tools
• Matt HicksInference and serving · Developer tools
• Olivier ChafikDeveloper tools · Inference and serving
• Pierrick HymbertInference and serving · Developer tools
• Julien SimonInference and serving · Developer tools
• Dane KnechtInference and serving · Developer tools
• Matthew PrinceInference and serving · Developer tools
• Erik BernhardssonInference and serving · Developer tools