Useful questions
How often do edits apply correctly? Which context improves completion acceptance? When does semantic retrieval beat lexical search? What tool policies reduce risk?
Project information
Open engineering questions
Measure before making frontier claims.
El-cro does not run a formal AI research lab. Research-oriented work should be reproducible, benchmarked, and connected to product reliability.
01Evaluation harnesses
02Retrieval quality
03Edit reliability
Publish evidence
Share datasets where licensing permits, evaluation code, model versions, prompts, hardware, costs, and failure cases.
AI coding, under your control
Build with your tools, models, and rules.
El-cro puts you in control of your models, keys, and data.