Now

Updated June 2025

A snapshot of what I'm doing right now. What is a now page?

Building

RAG recommendation engine @ createxpIterating on retrieval quality and latency, adding re-ranking and experimenting with hybrid sparse/dense search.

LineaAdding a workflow marketplace so teams can share and fork agent pipelines. Building the execution sandbox.

Learning

Distributed systemsWorking through Designing Data-Intensive Applications. Focused on consensus algorithms and replication strategies.

RustLearning for performance-critical infra work. Following the Book + building small CLI tools to practice ownership.

Reading

The Pragmatic ProgrammerRereading. Hits differently after a year in production.

Attention Is All You NeedGoing back to the original transformer paper to understand what modern architectures are actually changing.

Thinking about

I've been thinking about the gap between "AI features" and "AI-native products." Most products bolt LLMs onto existing flows. The interesting companies are designing the product around the model's capabilities from the start, and that's a fundamentally different design problem. Also thinking about where RAG actually breaks down at scale and whether graph-based retrieval (knowledge graphs + vector hybrid) is worth the complexity tradeoff.