Renato Garay
Software engineer. I design, build and run systems, from the first version to production at scale.
Here I share what I learn in labs about building with AI, teach at Anku Academy, and help teams through Anku Labs.
Latest labs
All 3 labs →confirmed refuted inconclusive
Does a design skill beat using no skill at all?
The gap between skills was smaller than the gap between two runs of the same skill.
Do subagents cut context use on a legacy migration?
41% fewer tokens in the main thread, with the same test pass rate.
Is hybrid retrieval worth it below 10,000 documents?
BM25 alone matched hybrid on nDCG@10 at a third of the latency.
Most read
- Do subagents cut context use on a legacy migration?
41% fewer tokens in the main thread, with the same test pass rate.
- Is hybrid retrieval worth it below 10,000 documents?
BM25 alone matched hybrid on nDCG@10 at a third of the latency.
Topics
All topics →- Claude Code workflows
Subagents, hooks and skills, measured
- Evals
Judges, blind votes and variance
- Agents
Delegation, cost and when it pays off
- Search and retrieval
Hybrid retrieval, RAG and relevance
Work with me
How I can help →Alongside my job, I take on a few consulting projects through Anku Labs: building something from scratch, fixing a system that already hurts, or helping a team work well with AI.
I also teach: courses at Anku Academy (in Spanish), group training for companies, and one-to-one mentoring. All teaching →