Zerionia / Lab
Tools and field research for technologies that are still becoming obvious.
Interactive research tools, comparative benchmarks and technical investigations exploring the frontier of AI models, agentic workflows and platform infrastructure.
Interactive Systems & Platforms
03 SystemsSmart Glasses Matrix
A source-backed capability matrix for an emerging computing platform.
Open experiment →
AI Transparency Checker
An educational review of the trust surfaces an AI product may need.
Open experiment →
Battery Passport Explorer
A practical map of the EU battery passport timeline, actors and data fields.
Open experiment →
Research & Dispatches
Lab Articles & Technical Breakdowns

Anthropic Ran an AI Across 15 Cheap TV Boxes. The Next Step Is the Scary Part.
Anthropic ran a 26B AI model across 15 cheap Android TV boxes and studied self-replication. Here's what the experiment really proved and what it did not.

Google Says Its New AI Found a Critical Hospital Software Bug Other Models Missed.
Google says Gemini 4 Argon helped uncover a critical bug in hospital software that older models missed. What was actually found, and how far can we trust the claim?

Claude Cowork Quietly Changed Where Your Tasks Run. Here's the Catch.
Claude Cowork now runs new Pro and Max tasks in Anthropic's cloud. Here's the difference between remote execution, local folder access and Claude Code.

CERN Is Building an AI With Mistral to Hunt for Physics Humans Might Miss.
CERN and Mistral are developing Flare, an AI model trained for particle physics. Here's why billion-collision datasets are a different challenge from chatbots.

ChatGPT Is Starting to Build Apps Instead of Just Answering Questions.
OpenAI's GPT-6 Intelligent UI turns some ChatGPT answers into interactive tools. Why the interface matters more than another round of benchmark numbers.

Mistral Says Its Tiny 209M AI Can Out-Forecast Models Much Bigger Than It.
Mistral's new Timestral model uses as few as 5M and up to 209M parameters for forecasting. Its benchmark claims are striking, but real-world validation still matters.

Designing the Commercial Trust Boundary
A practical interface model for keeping answers, sponsorship and measurement legible when AI products monetize conversation.