@adam-

As harsh as it might come across, a lot of people who inhibit the bottom space of the K curve come off as defeatist, and LinkedIn becomes their venue to complain about it. They lean heavy on frontier models, likely because they are unaware that any other options exist. I just responded to someone today who was complaining about Opus 4.8. Their issue was around extending and managing their context window. Except they didn't know what a context window was, or what tools exist to modify it. There's something to be said for "You don't know what you don't know", but there was also no effort on their end to try and find a better way, or impetus to solve it. Its a lot easier to complain than to try and figure out why. IMO changing the trajectory from those on the bottom of the K curve to the top starts with understanding what options exist to improve the experience. That journey begins when they start asking more nuanced and detailed questions. Most people will never make that leap, but without it I don't see that trajectory changing for them. It really depends on the kind of person you are, and also explains why a lot of neurodivergent people excel at A.I The reality is most people treat prompts like a Google search query. Thats understandable because we've been conditioned to approach it that way for the last 25+ years. Thats where most stall, because they treat it like a omni directional relationship. When that kind of person complains in detail on social media it's a strong tell of where they're stuck. Their ability to articulate their frustration is an indicator of where they are on the curve. i.e: going from "Opus 4.8 sucks now and the experience isn't the same, here's 10 reasons why" to: "What options exist to improve my experience with Opus 4.8? It used to help me with business strategy but It's responses are getting worse and hallucinating more." One engages feedback, the other is the sound of the world's smallest violin playing just for them. However, since I'm someone who wants to help people break upwards, my responses look like this: I synthesized their complaint into options they can utilize with bullet points, fed it to ImageGPT2 and then responded with something thats visually pleasing to digest and take into action. Here's what I did in this instance. I'd then follow up and say "Feed this image into Claude or ChatGPT and ask it to interview you about what you're struggling with, along with any details about your set up. Use this image as a guide to help, and add any suggestions based on your feedback" In doing so they usually learn: a) How to better interact with an LLM to help solve a problem based on specific feedback b) That images can be interacted with in this way Time will tell if it sticks and they build upon it, but they now know something they previously wouldn't have considered if they just treated their interaction like a normal query.

3 replies

2 recasts

3 reactions