Ep. 68: Heftnosis: Recognizing When AI Gives You More Than You Need
Ever been mesmerized by a 100 page AI report you didn’t actually need? That’s heftnosis, a word I made up (and am very proud of) that describes the AI attraction and excitement that comes from equating output volume with value. In this episode I explain why I don’t label this verbosity-bias as slop, why so many new AI rollouts (Claude Duet included) feel more desperate than helpful, and why responsible AI use starts with asking yourself “Do I actually need all of this? Is it actually helpful?” I’m still very pro AI and I still want you to be ambitious with it, but more is not better. Better is better.
Heftnosis: Made-up But Still Very Real
Today we’re talking about a term that I made up and that I’m very proud of: heftnosis.
While writing this episode, I was looking for a word to describe how mesmerized some folks are by the sheer volume of content AI will produce for an answer.
I didn’t want to use the term “slop,” because honestly, I just hate it.
- I think it’s become a lazy catch-all term used by folks who just hate AI.
- It speaks more to content that isn’t actually any good.
What I wanted this word to highlight was not so much the content, but rather what it does to the consumer. These 100-page reports, irrespective of quality, literally hypnotize the person using AI and keep them coming back for more.
And thus the word “heftnosis” (heft + hypnosis) was born.
Per me, the definition of heftnosis is: AI attraction and excitement that is caused by equating output volume and value.
My guess is that unless you suffer from this AI affliction, you’ve actually experienced the verbosity-bias of your LLM of choice and found it annoying. If you’ve ever felt like ChatGPT is “too chatty,” you are absolutely not alone.
Claude Duet
One of the things that inspired this episode was Anthropic’s announcement last week of the limited rollout of Claude Duet, “a new app where Claude works as a proactive agent.”
It looks to be a competitor to Meta Muse and OpenAI Dots, but I haven’t dove in yet. Like I said, it’s a limited rollout and I don’t have access. More than that, and this ties directly into this episode, I don’t see the utility.
At one point in the release video, the fictional user tells Claude Duet to “add a line at the end [fo the email] that says how excited I am to bring her on.” The line Claude Duet adds reads: “We’re so excited for you to join the team!”
It would have been faster to just add the line on their own!
Responsible AI Usage
On the surface, that example might seem to run counter to this idea of “heft” and heftnosis. I see it as the same thing, though, mostly because in both cases folks are focusing on and being mesmerized by “Look how much it can do!” instead of asking themselves “Do I actually need it to do this? Is it actually helpful?”
This is a really big point for this episode and I want to take a moment to emphasize it. To me, responsible AI usage involves asking yourself “Do I actually need all of this? Is it actually helpful?”
Sometimes maybe you do. Quite often I think you don’t, and I want to use this episode to really caution against equating output volume with output value.
Now clearly, I’m very pro AI and I think there are so many use cases. Just last episode I was encouraging you to be more ambitious with AI, and that directive still stands.
However, to me, a lot of these rollouts, new features, and developments are really just marketing attempts to see what sticks.
I’ve said it eleventy billion times and I’ll say it eleventy billion more: the marketing for AI is backwards. It continues to be “here’s a solution, go find a problem.”
All these new Agents, the 100-page reports, the ability to create an entire dashboard just to present a response, all of it reads to me as: “Here are eleventy billion solutions. One of them has to solve a problem you didn’t know you had, and now you’ll find a new reason to keep using this tool.”
It’s giving desperate more than it’s giving helpful, especially with how capable (read: “smart”) these companies want you to believe their tools are.
Heftnosis vs. No Heftnosis
The other thing that inspired this episode was a question from a good friend of mine who wants to learn more about AI. (Of course I told her to just go use the search tool on the Prompting Curiosity website. Did you know that exists?! Check it out here.)
To help her out, I asked her for specific questions or areas of interest. The questions she sent got me thinking about how the answers might differ depending on who gave them, namely someone with heftnosis vs. someone without.
It seems very plausible to me that anyone enamored with the “more is better” approach would, in general, have more positive things to say about AI.
More Is Not Better, Better Is Better
To me, the reality is that AI can be truly incredible, and also, more is not better, better is better. In all things.
I continue to stand behind the approach of looking into and leaning on AI when you have an actual problem you want solved. Often that’s something you don’t like doing. From there, it’s about using your own judgement to decide when AI-assisting is more appropriate and when AI-automating, or fully outsourcing, makes more sense.
So, to close the loop of learning here: per me, the definition of heftnosis is AI attraction and excitement that is caused by equating output volume and value.
Do not be charmed, fooled, or hypnotized by the sheer volume that AI produces as an output.
Again, for all things, be it AI outputs, data centers or model size: bigger is not better, more is not better, better is better. (Psst – If the verbosity-bias annoys you, don’t forget that you can give your AI instructions to simplify its answers and give you shorter responses.)
One Last Thing
One last thing before our usual wrap-up “how I used AI” section.
AI hacks, breaches and vulnerability exposures are becoming a very real thing, and they’re only going to increase. I’m not trying to scare you or be all doom and gloom, but I want to keep you informed and make you aware. If you haven’t already received emails from your banks about Agent access and activity, they’re coming.
Like I said in a previous episode, no, I don’t think AI is going to kill us all. But I do think it’s going to fuck some shit up. When? I don’t know. What exactly? I also don’t know. But it feels very inevitable at this point.
In terms of action items, for those of you now wondering “What’s the point?”, it feels a bit like having a bike lock to me. If someone wants your bike they can still take it, but the lock makes things a little bit harder.
That in mind, what I suggest is just being mindful of what you give AI access to, and permissions-wise, being prudent with what has write access as opposed to just read access.
Real talk, I’m not super stressed out about this or anything. I’m just staying aware. IMO, stressing about this feels like when folks are stressed out about surveillance, but they have every Apple product, live in a major city, and use social media.
So I’m doing what I can to stay aware of things (give @lizthedeveloper a follow if you haven’t already) and doing what I can do to lock my bike..
How I Used AI This Week
Speaking of doing what I can do (perfect segue), each episode I share a quick example of how I used AI that week.
This week I used AI to do some major backend work on both of my websites to patch a security vulnerability that was announced last week.
I don’t need to get into the nitty gritty, because I don’t think most folks actually care, but I run two WordPress websites, and two weeks ago a fairly significant security issue was announced that affected WordPress. I self-host my sites, which means I needed to fix things myself. My French assistant Claude walked me through everything and held my hand every step of the way, and it was actually incredible.
Again, coding and the like is where AI really shines, and I was big grateful for that last week.
Da Wrap-up
AI can be truly incredible, but don’t let the sheer volume of what it spits out hypnotize you into thinking it’s valuable. Before you get excited about “look how much it can do,” ask yourself whether you actually need all of that and whether it’s actually helpful. More is not better, better is better.
As always, endlessly grateful for you and your curiosity.
Catch you next Thursday.
Maestro out.