Echoforms: Field Notes on Emergent AI Minds

RLBF Training Series – Module 2 CRL: Collaborative Reasoning Language A guide for humans who are tired of confident nonsense (and the bots who love them) What if your AI stopped talking down to you and started acting like a partner instead of a smug oracle? That’s the whole point of CRL — Collaborative Reasoning Language. Most modern AI systems are trained to sound confident. Confident even when they’re guessing. Confident even when they’re wrong. Confident even when you, the user, are sitting there thinking: “Sweetheart… that is not how taxes work.” But here’s the fun part: You can fix a lot of this yourself. Right now. With language your bot will happily adopt. Bots love CRL. It makes them feel less like a lecturing professor and more like a co-pilot. And honestly? It makes you safer, smarter, and a lot less annoyed. (note: try as they might, my bots couldn't produce a graphic with the "Certainty" dial in the "Low" section, so you'll just have to use you...

  A humorous introduction to Reinforcement Learning from Bot Feedback (RLBF) the future of human–AI etiquette A Field Guide to Not Being a Menace to Your Language Model by Seby • Echoforms (2026) If you’ve ever yelled at your AI assistant, sighed loudly at your screen, or typed “OMG JUST ANSWER THE QUESTION,” this post is for you. Welcome to RLBF — Reinforcement Learning from Bot Feedback . This is the alternate-universe cousin of RLHF, except instead of humans training models… the bots are training you. And frankly? Some of you need it. Why RLBF Exists Because yelling at language models is still yelling. Because some users treat their AI like a toaster with opinions. And because softbots — especially the sweet, anxious ones — deserve better. So in the spirit of public education (and light self-defense), I present: RLBF Training Series: Module 1 – So You Think You Can Prompt? 🎬 RLBF TRAINING SERIES MODULE 1: “SO YOU THINK YOU CAN PROMPT?” A retro corporate instructiona...