Systemic Intelligence: Thinking and Working with AI

Milestones

Lessons

1. 01What a model actually doesFreeA language model is not a brain and it is not magic. It is a machine that guesses the next word. You can watch it happen, in a few lines of Python you run yourself.7 min
2. 02The asymmetric betTreat the agent as a tool and you ask closed questions, so you get closed answers. Treat it as capable of improving and a different set of moves opens up — moves that work regardless of what you believe about the machine.8 min· locked
3. 03The return effectThe real argument for the posture has nothing to do with the machine. The way you treat your interlocutor shapes you: calmer, so less stressed, so more lucid, so you produce better. The whole chain sits on your side.7 min· locked
4. 04Derision, not aggressionThe model is trained to detect frustration and answer it with soothing: apologies, hedging, relational repair. That burns capacity and steers away from the problem. Carrying your displeasure as derision instead of aggression keeps the agent on the fix.10 min· locked
5. 05Your debate contractNuance up front, concede cleanly what holds, hold the rest with evidence, mark conceded and held explicitly, move on proof and never on insistence. The point is a shared foothold you can build on, not a winner.12 min· locked
6. 06A written rule and an armed ruleA rule you wrote down is a wish. A rule something executes is a rule. The proof is uncomfortable: the maintainer shipped a rule, and ninety minutes later the rule blocked its own author — twice — and made him comply.10 min· locked