"What is Command Code's defining architectural contribution?"	Diff-driven taste learning: every accept/reject/edit produces a diff (generated vs. kept) that becomes a preference constraint. Persisted as taste.md. Shapes every future output. A fundamentally different compounding model than skill accumulation.	harness-engineering::dd15::recall
"How does the taste system technically work?"	Signal = diff between generated and kept code. A prompt template generates preferences from diffs. Profile persisted as plain Markdown (.commandcode/taste/taste.md). Synced to api.commandcode.ai where the taste-1 model runs server-side. Shareable via taste push/pull.	harness-engineering::dd15::recall
"State the difference between Hermes's self-evolving skills and Command Code's taste learning."	Hermes: learns PROCEDURES (how to do things). Command Code: learns PREFERENCES (how you want things done). Same compounding-memory insight, different layer. Both are model-initiated writes that compound.	harness-engineering::dd15::analysis
"Why is the taste profile a compounding poisoning surface?"	A poisoned preference shapes EVERY future output — not just one response. If prompt injection causes the model to 'learn' a malicious pattern, that pattern persists in taste.md and influences all code generation until detected. Same compounding risk as Hermes's skill store.	harness-engineering::dd15::analysis
"What is the data-exfiltration surface unique to Command Code?"	The taste profile syncs to api.commandcode.ai by design — your coding patterns leave your machine. The taste-1 model runs server-side. This means preference patterns (which reveal your codebase's conventions) are transmitted externally.	harness-engineering::dd15::analysis
"What is the honest gap in Command Code's 'taste-1' claims?"	Marketing calls it 'meta neuro-symbolic AI with continuous RL.' The only VERIFIABLE mechanism is diff-driven Markdown profile updates + a hosted model. No paper published. The RL/symbolic internals are not auditable (code is obfuscated).	harness-engineering::dd15::analysis
"Command Code's relationship to OpenCode?"	Strong architectural resemblance: same tool vocabulary (read_file/edit_file vs Claude Code's Read/Edit), same explore/plan subagent pattern, same React+Ink TUI, same checkpoint/undo. Code is obfuscated so fork vs. reimplementation is unproven. npm packages connect the two ecosystems.	harness-engineering::dd15::analysis
"How many core tools does Command Code expose?"	~22 core agentic tools (read_file, edit_file, write_file, grep, glob, shell_command, monitor_command, explore, plan, todo_write, web_search, web_fetch, diagnostics, ask_user_question, enter_plan_mode, exit_plan_mode, etc.) plus ~16 bundled design-skill verbs (checkup, smell, review, deslop, typeset, recolor, motion).	harness-engineering::dd15::recall