Writing · Arjhine Ty

Writing

Notes from the work, written down.

Technical writing on the systems I build — the decisions behind them, what broke, and what I'd do differently next time.

I benchmarked my own agentic skills against plain web search

They lost, and cost 2.6x the tokens losing

120 agent runs across four arms says the skills I built did not beat an agent that can just search the web, and cost 2.6x the tokens doing it. Why the usual skills benchmark hides this, and the contamination bug that nearly invalidated the whole wave.

Read →

Evaluation, honestly

Why your RAG evaluation is lying to you

Series: Evaluation, honestly — part 1

Three mistakes that make RAG evals look better than the system: answerable-only test sets, LLM judges grading their own family of models, and latency hidden from the quality number.

Read →

Latency budgets: what 800 ms actually costs

A field note from the PocketPatient latency work

800 ms feels like a number until you price it: turn-taking psychology, eval timeouts, and the budget that forced the routed-model design.

Read →

Building agents that last

Agent memory is a product decision

Series: Building agents that last — part 1

What an agent remembers is not a plumbing detail: it is the product. Notes from building Mimir and what the recall numbers changed about the roadmap.

Read →