Updated August 2026 · what is a now page?
Wicklee — pushing toward first revenue: self-serve checkout and a public WES leaderboard so people can compare tok/W and thermal honesty across real setups, not vendor spec sheets. v0.9.0 shipped Runtime Config Surface and a dedicated Models tab. → wicklee.dev If you run Ollama, vLLM, or llama.cpp in production, I’d love feedback.
Writing — developing a thesis on the inference-network control plane: the resource-decision problem that sits below protocols like MCP and above any single model or GPU. First essay, The Layer Below MCP, is up; two more are drafted and coming once they clear an internal review.
ELM work (personal research) — finishing account-intelligence-7b-v1.
Fine-tuned Qwen2.5-7B on a synthetic dataset of 568 examples across six
surfaces: meeting prep, QBR, handoff, renewal alert, onboarding, escalation
triage. LoRA training on DGX Spark. Eval harness running DeepEval +
LLM-as-judge with 8 locked metrics. Q4_K_M GGUF release when it clears eval.
Running — base building. No race on the calendar yet. Keeping it consistent. Following along on Strava.