Victor Guillard
Building agentic AI systems. Writing about agents, orchestration, and the tools that make them work.
Writing
Font Evaluation
Testing font pairings against The Desk principles.
Why Agents Fail in Production
Most agent failures aren't model failures — they're orchestration failures. 90 days of testing three architectures.
Checkpoints Should Be Default
A short observation on why agent checkpointing should be opt-out, not opt-in.
Benchmarking Three RAG Architectures
Naive RAG vs hybrid search vs agentic RAG — measured on 500 queries across 4 domains.