Mario design rules for vertical agents: consistent inputs and progressively unlocked capabilities
dragon_khoi · x · 2026-09-26
- The author borrows two game-design rules: Mario always jumps when you press A (users learn inputs from consistent outputs), and you don't get all of Mario's moves at the start (each level teaches a new move and how to use it).
- Applied to AI products: consistent agent behavior builds user trust, and new capabilities should be introduced progressively — e.g., via distinctively named new agents.
- His core interest is vertical agents: a defined task space the user knows about constrains expectations and creates that "100% reliability" feeling; expansion happens by introducing new agents that teach users about new capabilities.
More from coding & agent
- Peter Steiberger says he codes with Codex plus an OC harness, not Claude — steipete · 2026-09-26
- One prompt makes Claude generate a 2-min video game history film entirely in code — prasenx · 2026-09-26
- Run /doctor prompt-audit to find stale paths and conflicting instructions in Claude Code — daniel_mac8 · 2026-09-26
- Repurpose your old M1/M2 Mac as a local test server with this open-source runner script — jasonkneen · 2026-09-26
- DHH says 37signals went "pencils down": 150k lines in August with agents vs 30k/year by hand — Recent-Tangerine2745 · 2026-09-26
- Harness engineering: the key discipline for building reliable AI agents — techNmak · 2026-09-26