Understanding degradation through plausible slop
My role is weird. Yours has probably gotten weirder too since AI has gotten democratized. It's a battle for cognitive availability and good judgement. That typically requires deep understanding of the topics we engage with on a daily basis. In my case: strategy, AI, Data, technology in general. Based on this judgement, there is expectation for me to set right priorities, budgets, and deliver results.
I now feel that my understanding is being degraded unintentionally.
I'm sure you've seen what a Claude-generated doc looks like. It's amazingly structured, text formatting is perfect, all of the pompous words make it also sound astonishingly smart.
It tires the mind.
I call it "plausible slop".
It's not "slop" because it's bad. On the contrary. It's slop because the person working on it hasn't yet gone through the steps of understanding what the output represents. It's the result of mostly thinking about something shallowly.
The difference between thinking and understanding.
You know how sometimes you go into "clarifying meetings" or "alignment meetings" and you get out more confused that you went in? That is because there is no good understanding of the topic, which would be the only way of clarifying it.
This is also what happens with the documents and assets we generate and send, with minimal review, minimal depth, minimal pushback, and no understanding. It creates confusion, cognitive drag, and degrades decision.
Being an engineer, I hate complaining without coming up with solutions. That's why I built Thinking OS. A collection of agents and skills that have the purpose of driving understanding.
Agents can only do thinking. But they can do it in such ways that help you achieve understanding. The name is intentionally "Thinking OS". It's still our responsibility to understand.
How will it change the way you work? Let me tell you how it changes mine.
Lately, my Claude-radar has been really sharp. I can pretty much spot every time someones sends Claude-generated content/assets/docs. Therefore, my first step has been triggering Thinking OS on it. This immediately allows me to take content down to first principles which is the easiest way of initiating the understanding process. This steps also humanizes the output. You know...removing all the formatting fluff like em dashes, "not only...but" statements, and so on. Nobody needs that.
The skill name is easy to remember: /understand
Once first principles are understood, we go into asking questions. Questions are the number one way of evaluating if someone understands something. You know how sometimes you can really tell someone is getting your point when they ask really sharp questions? Thinking OS is built to support on both ends: supporting the reviewer with first principles, and then preparing questions for the creator to drive better understanding and deeper research.
Next, your responsibility as the reviewer of the content is to get back to the creator of the content and push for better understanding leveraging the new perspectives that Thinking OS has helped you achieve.
❗️Keep in mind that you should use this on your own documents and assets to ensure you understand the content. Not only plausible AI slop you receive. We bias ourselves into thinking our output is better than that of someone else. When some reads it with no context, our output is just as fuzzy.
In closing, I really hope Thinking OS enables better decisions. If it doesn't, I hope this article increases awareness of the cognitive burden we are all sometimes causing in teams with increased verbosity and decreased understanding.