Tag
dubbed
Project Management for R&D Teams
Early-stage engineering isn't about building as much as possible. It's about reducing uncertainty before you run out of time and money.
Machines Should Retrieve, Models Should Reason
Many search and matching workloads don't need an LLM or a vector database. Use the cheapest computational primitive capable of solving the problem.
Smaller teams should collect less data
The standard playbook for modern software insists that more data equals better product capability but for small teams more data is more problems.
Glyph — MinHash is all you need
A lightweight content intelligence toolkit. Compact fingerprints, no vector database, and the thought process behind skipping the heavyweight stack.
Glyph Collections — Tricking developers into guiding MinHash
More text is not more signal. Collections steer fingerprints toward the fields that actually matter — without making you learn aggregation.
Glyph Query — Don't just search
Don't make the search algorithm do work the system can eliminate earlier. Fingerprints, LSH, and filters as a deliberate tradeoff; not a smaller Elasticsearch.