OLX CPO Built a Company OS on Hermes+OpenClaw: 54% Context Coverage, 31% Accuracy Lift
aakashgupta · x · 2026-11-02
OLX Classifieds CPO Mikhail Shcheglov shares on the Growth Podcast how he built a company operating system on OpenClaw and Hermes over 5 months; his entire product team now runs on it.
Key findings:
- His knowledge graph covers 54% of the company's product, business and customer context — a number he tracks as a personal KPI. At that level the agent already makes backlog decisions; stakeholders pitch feature requests to it before getting near a PM
- It runs his email, calendar, recruiting funnel and design system
- Counterintuitive finding #1: summarizing meeting transcripts costs 20-25% recall, so he stores every one raw
- Counterintuitive finding #2: letting Hermes write its own skills off repeated tasks produced a 31% accuracy lift in controlled testing
Also discussed: what happens to PM headcount when one PM covers four domains, and what CPOs should screen for now. A quoted tweet mocks agents that say "it's super easy, just open your calendar yourself" as "fake helpful."
More from coding & agent
- Install a Gmail connector on ChatGPT to avoid output-misreading mistakes — jxnlco · 2026-11-02
- A weekly competitive-intelligence agent prompt for biotech pipeline tracking — nlarusstone · 2026-11-02
- DeepSeek Harness Super-agent design: the bottleneck isn't smarter models, it's your attention — sujingshen · 2026-11-02
- tito: the hidden tokenizer bugs breaking RL training, and a Lean-verified fix — VoidAsuka · 2026-11-02
- Build an Agentic Knowledge Graph from Scratch — No Graph DB, No Extraction Model — adnan_hashmi · 2026-11-02
- One Vault, Two Sides: An Obsidian + Claude Agent Workflow for Personal Knowledge — adnan_hashmi · 2026-11-02