SMELT: Scaling Laws for Compute-Matched MoE Looped Transformers
Authoritative Technical Breakdown & Systems Analysis of SMELT: Scaling Laws for Compute-Matched MoE Looped Transformers (arXiv:2609.01343).
Articles
Thoughts on Agentic Systems, AI Architecture, and the future of work.
The AI Inference Bottleneck: explaining DFlash, PFlash, and the Lucebox Hub
A comprehensive deep-dive into state-of-the-art AI inference acceleration technologies. We explore the 5 Ws of PFlash and DFlash, their synergy, and an end-to-end teardown of the Lucebox Hub repository.
How to Build a RUST web search ai agent using coding agents
Why the future of enterprise automation lies in stateful, reliable agents, not stochastic chat interfaces.
Building a web search ai agent using rust
Why the future of enterprise automation lies in stateful, reliable agents, not stochastic chat interfaces.
The Manual Data Science Delusion: Why Autonomous Agents Render Intuition Obsolete
Exposing the hard truths of autonomous systems from the iteration volume gap to the feature engineering ceiling that make manual workflows look like statistical waste.
The Shift to Deterministic AI Agents
Why the future of enterprise automation lies in stateful, reliable agents, not stochastic chat interfaces.