Mark Sulleza, technical lead. AI tooling that is cheap to run and hard to misuse: coding agents, model routing, MCP servers, plus hardware and apps built on the side.https://mark-web.com/https://mark-web.com/writing/read-only-gmail-mcp/https://mark-web.com/writing/read-only-gmail-mcp/One OAuth scope, a refresh token in the Keychain, and email the agent treats as data. Also why this site has a privacy policy.Sat, 07 Nov 2026 00:00:00 GMThttps://mark-web.com/writing/living-with-hermes-agent/https://mark-web.com/writing/living-with-hermes-agent/Nous Research’s Hermes on my Mac: 295 sessions, 105 skills of which it wrote 41, a bot per role, and why it now asks before it runs code.Sat, 07 Nov 2026 00:00:00 GMThttps://mark-web.com/writing/waypoint-mcp-checked-writes/https://mark-web.com/writing/waypoint-mcp-checked-writes/Waypoint’s MCP server writes as the signed-in user, through the same database checks as the app, and answers every write with what it changed.Sat, 10 Oct 2026 00:00:00 GMThttps://mark-web.com/writing/two-agents-one-memory/https://mark-web.com/writing/two-agents-one-memory/Hermes and Claude Code share one pool of notes. Mining 148 chat threads into it produced about 400 writes; 361 of them needed fixing when checked against their sources.Wed, 30 Sep 2026 00:00:00 GMThttps://mark-web.com/writing/px-deepseek-coding-agent/https://mark-web.com/writing/px-deepseek-coding-agent/A 30-turn session costs $0.0149 off-peak. Most of the saving comes from never changing a byte the cache has already seen.Fri, 11 Sep 2026 00:00:00 GMThttps://mark-web.com/writing/what-deepseeks-api-wont-tell-you/https://mark-web.com/writing/what-deepseeks-api-wont-tell-you/Notes from building px: cache hits at 2% of a miss, a model that spent its whole output budget thinking, images that vanish, and the $16 day that taught me to price a sweep before running it.Fri, 11 Sep 2026 00:00:00 GMThttps://mark-web.com/writing/local-27b-model-on-a-laptop/https://mark-web.com/writing/local-27b-model-on-a-laptop/Qwen3.8-27B on an M4 Max: 20.8 tokens a second, a 20,000-token floor before any work starts, and the defaults that quietly broke it.Sat, 29 Aug 2026 00:00:00 GMThttps://mark-web.com/writing/claude-code-hybrid-model-stack/https://mark-web.com/writing/claude-code-hybrid-model-stack/A local LiteLLM proxy where a strong planner hands bounded coding jobs to cheap executors, with spend tracked per session.Tue, 07 Jul 2026 00:00:00 GMT