Do you get frustrated when your AI systems don’t work? You may be treating it like software when it should be treated like an employee. AI is proficient but it isn’t a mind reader. It makes mistakes like any person would. Onboard your agents with the same training and guidance you would any new employee.

"AI" is on everything now: your email, your bank alerts, a nine-dollar phone case claiming to be "AI-powered." (Phone case is a gimmick example) AI is the same two letters you find on almost everything now, and they are claiming wildly different things. Let's build you a map so it all stops being a fog. 👇 Hold this picture: "AI" is like the word "vehicle." 🚗 To me the word vehicle just means "A machine that moves stuff." A scooter and a cement truck are both vehicles. To know anything real, you ask: which kind, and what job is it doing? 💡You should ask that of any AI too. The two real families of AI today - 🔧 Single-purpose AI: does ONE job (spam filter, fraud alert, photo face-grouping). Superhuman at its one task, not so great outside it. - 🧰 General-purpose AI: ONE tool, TONS of tasks. ChatGPT, Claude, Gemini will write an email, translate a menu, and explain a contract in one afternoon. It's the AI that most people are working with today and it is the one that made everyone pay attention. 💡The fear-killer: NONE of it is the movie machine that wakes up, wants things, and acts on its own. Experts call that AGI, and I did not include it in the above list because its not real and IBM (which builds this tech) calls it a theoretical concept, because it does NOT exist. Even the versatile chatbots have no understanding or will of their own and do nothing until you ask. A tool, not a mind. 🧠 The stacked-up words are just nesting dolls: AI (the umbrella) ➡️ Machine Learning (AI that learns from examples) ➡️ Deep Learning (the powerful engine) ➡️ Generative AI (makes new text and images, like ChatGPT). 💡Each one is a more specific member of the last. That's it. The two kinds you meet every day - 🔵 Quiet AI runs in the background: spam filter, Netflix picks, Maps rerouting, your photos grouping faces. It works on your data behind the scenes. It optimizes, sorts, and predicts based on natural human behavior. Because it was designed to be invisible, it trained us to trust machine assistance without even realizing it.

As some of you may know, I have been seeking new employment. My current job stagnated and is not taking AI as seriously as I am. But I also hate job hunting. So I have taken the Fable time to help me get all my ducks in a row. LinkedIn, my profile page, GitHub. Finding the small gaps that could have stopped a recruiter but were invisible to me. Some of what it caught: my LinkedIn About was quoting stale numbers from my own projects (the repos had moved on, the profile hadn't). My GitHub had no bio, no profile README, and the wrong repos pinned. My portfolio site was missing my two best projects entirely. It rebuilt my resume from the verified repo facts so everything tells the same story, and it even found that I'd tagged v1.0 on Plumbline but never published the release. One thing it suggested was posting on LinkedIn. I see its point, but what to post? So I fed it all the posts I have done here in Clief Notes and it drafted a series to be doled out one at a time. They are good and they sound like me. Very minor edits. First one went up yesterday. My stuff has never looked this good. Profile link, if interested. https://profile.bytesbytoby.workers.dev/

Oh man!!! What a surprise. For the past 5 days or so I have been quite literally 100% focused on putting things into practice and building my first project.... So imagine my wonderful surprise when, on July 4th, I finally make some time to get caught up with other things and I see a level up notification from a few days ago!!! Thanks to everyone that has helped me, and liked my comments and posts.

📝Note: This post is not about ICM, it's about tuning the AI systems you are using with ICM, if you're new to ICM, you can start Here: https://www.skool.com/cliefnotes/welcome-to-clief-notes-heres-where-to-start-2?p=f8f85a09 @Mira Bradshaw also made a great post to get you started! https://www.skool.com/cliefnotes/clief-notes-gems-for-starting-in-icm?p=0385cbf9 Let's dive into the results just 48 hours later! 👇 Two days ago, I posted about pointing Fable at my own environment and finding out where my tokens were really going. 📝If you need to catch up with my first post: https://www.skool.com/cliefnotes/i-just-spent-the-last-few-hours-asking-fable-to-make-itself-cheaper?p=cc666986 A lot of you went hunting in your own setups after that post (and found GOLD 👏). So, here's the part two: what happened AFTER the audit. 📌 TL;DR - The fixes shipped, the savings are real, and the most important lesson came from re-measuring two days later: optimization is not an event. Drift never sleeps. The audit is worth nothing without a cadence behind it. 🧾 First, the receipts. Every fix from the audit is now live: ✅ The file that lied. My startup menu file claimed "450 tokens" in its own header and was actually ~7,000. It got rebuilt down to ~750, and the detail it was hoarding moved to where it already lived anyway (each project's own handoff file). 💡One fact, one place, and everything else that needs it points to it by design. ✅ The silent tax. My safety hook was injecting ~800 tokens into EVERY prompt, repeating rules that already load once at session start. It's now ~85 tokens: a short pointer to the rules instead of a full copy of them. 💡The enforcement never lived in the repetition; it lives in the hooks that physically block bad actions. Cut the prose, keep the mechanism. That one saves on every prompt, in every session, forever.