Yaroslav Vikhariev

Yaroslav started AI//COST in March 2025 after spending three months unable to find a single trustworthy comparison of Claude Code, Cursor, and Aider for his own evaluation. Every existing source was either vendor-marketing or single-author blog posts running 3-5 prompts and calling it a review. The gap was obvious; the project was the answer.

Before AI//COST he spent twelve years in software engineering — last as founding engineer at two YC-backed companies (one B2B SaaS, one developer tools), before that as backend lead at a fintech in Berlin, before that as full-stack engineer at startups in St. Petersburg. The technical depth shows up in the methodology: he writes the pricing parsers himself rather than outsourcing to a freelancer, runs SWE-bench reproductions on his own hardware, and has personally installed every coding agent in the directory.

He's not an AI researcher. The site doesn't pretend otherwise. The expertise here is the practitioner kind — what 4,200+ test runs and 12 years of engineering judgment produce, not what an ML PhD produces. That distinction matters for what we cover: tool-use questions, cost economics, real-world workflow fit, agent reliability, integration patterns. Not architecture papers or training-method critiques.

On the editorial side he is editor-in-chief — final sign-off on rankings, corrections approval, methodology versioning. The 3-reviewer panel for editorial rankings includes him plus M. Krause and A. Park. When the panel disagrees, joint re-review is logged. He can be overruled by panel consensus and has been (twice — once on Cursor's #1 ranking in 2025, once on the τ-bench leader call in early 2026; both logged in corrections).

When he's not running the desk, he's contributing to open-source agent tools (aider contributor since 2024) and writing about testing AI tools without becoming one of those people who just runs ChatGPT through a five-question gauntlet and calls it a benchmark.