An AI partner that argues with you before you commit.

The whole point is that it pushes back.

Every AI assistant you've used agrees with you, because agreeable is what they're trained to be. An agreeable advisor is worse than none: it laminates your existing lean in confident language. The Stack runs on a written behavior contract instead. It restates your call before analyzing it. It makes you sign what winning means before any option gets argued, because criteria written after options are criteria reverse-engineered to bless the favorite. It writes the strongest honest case for the side you're leaning away from. It states its lean and then hands you the pen, because it's your company and your Tuesday.

Every session ends one of exactly three ways:

• A call, made and written.
• A missing fact, with an owner and a date it arrives.
• A decision date, and the call gets made that day with whatever's known.

Nothing ends vague. "Let me sit with it" without a date is a decision to delay, and it gets logged as one.

The Loop, the three free commands. Complete, forever.

Plus the doctrine underneath: the eight rules and the partner contract, in plain markdown you can read in ten minutes.

What a session looks like.

What Pro adds.

The free Loop (the three commands above) works cold, forever, and it's genuinely enough to run any decision through. Pro is depth and memory: the ten named calls with the documented record behind each, the conviction interrogation, sparring partners that argue your call from fixed lenses, the sovereignty layer, and the Engine, where the system learns your business once and your journal starts compounding. The seventh time you run a firing decision through /hard-call, you'll want /fire. That's the honest pitch, and it's the whole pitch.