EXPERIMENTS

A collection of experiments.

Exploring AI capabilities, limitations, and practical usability. Each experiment is documented as a process: problem, hypothesis, experiment, failure, iteration, evidence, result, and lesson.

STATUS

WORKING

EXPLORING

FAILED

STABLE

SELECTED EXPERIMENTS

001

35B MoE Models on 12GB VRAM GPU

Testing whether a 35B MoE model can become usable on a 12GB VRAM GPU by finding the optimal configuration for stable performance.

STABLE

002

Prefill Test for Long Context Inference, 26B MOE Models, Consumer GPU 12GB AMD (RDNA 2) with ROCm Backend.

Exploring context scaling, prefill cost, memory pressure, and practical usability.

STABLE

003

Can an open-source LLM with relatively small weight parameters conduct complex business research?

Connecting LLM, tools, memory, and automation into a working system.

STABLE

004

Dense vs MoE comparison

Comparing model architecture behavior under constrained resources.

WORKING

Each experiment is documented as a technical case study, showing not just results, but the process of thinking, testing, failing, and iterating.

GET IN TOUCH