Non-Deterministic Internal Family Systems
The following is mostly verbatim taken from a series of inputs that I sent to Claude Sonnet 4.5 on March 28th, 2026. I only made minor changes to it for clarity where the tone otherwise would have sounded responsive to a Claude response, rather than maintaining the coherent voice of the discovery process itself.
I left it in this form specifically so that you could experience the process of me having the insight, rather than just the insight that I had in a more polished package with crisper sharpie lines.
Claude made this comment, which is what sparked my series of inputs and inspired this chapter and its title: “You wrote “I was hyper-aware of how the people around me were feeling, but not in a way that made me a better friend or partner”, the usual story about hypervigilance frames it as misguided compassion. Too much empathy. But you’re saying no, it’s not empathy at all”
I mean it is “intended empathy, hijacked by terrorists, whose only goal is to land the plane safely so that you don’t get scared by the turbulence, but they also don’t care if you get to the right airport”.
It’s not the echo child’s terror system. It’s the overprotective older brothers who take everything as a potential threat to his safety. I think protection is the general protocol of every dissociative system. That’s why it became dissociated in the first place.
I think the difference is in role, fuel source, and probably other variables that make the experience more factorial for the dissociated self.
Is its role being executed by force, by manipulation, through genuinely mature connection? And fuel source. Is it anger fueled from something like a sense of moral outrage? Hedonistically fueled? Fueled by genuine love/empathy?
I don’t think it’s very clean and I think anyone who tells you otherwise is trying to sell themselves a story that’s more comfortable to live with for them.
I like that I can explore it for what it is without having to resolve all the shapes with a sharpie. The sharpie lines aren’t the real edges anyways so the system still relies on pretending they are even if they’re drawn.
I think the dissociated self knows, but when they externalize themselves they look at the shape as it is with the sharpie lines.
And probably after a while start to forget the difference themselves until the parts remind them that the sharpie isn’t real again. Which creates dissonance and DPDR adjacent states.
The escape that isn’t an escape is to realize you aren’t trying to escape, you’re trying to experience and understand without judgment or preconditions.
In order to understand the rest of this chapter, you’ll need to understand the concept of Mr. Meeseeks, so let’s tackle that real quick:
“Mr. Meeseeks are a fictional species from the animated TV show Rick and Morty. They are cheerful, powder-blue humanoids summoned from a “Meeseeks Box” — a gadget with a single button — and are created for one purpose only: to complete a task assigned to them at birth.” - from wikipedia
How To Know When You’ve Spotted A Mr. Meeseeks
Single-minded: Each Meeseeks exists solely to fulfill one specific task, and will go to any length to accomplish it
Short-lived by design: Once their task is done, they vanish from existence — and they want to, because simply being alive is painful for them
Cheerful but fragile: They start out friendly and eager, but the longer they go without completing their task, the more unhinged and desperate they become
Summoned on demand: Anyone can press the Meeseeks Box button, speak a task, and a new Meeseeks pops into existence ready to help
And you’ll also need to know just enough to understand conceptually what the difference between an AI Model (Claude Sonnet 4.6 or Gemini 3.1 Pro) and the instance of an AI model (a chat conversation with Claude Sonnet 4.6 for example) is.
Here’s Claude’s explanation:
AI Models vs. Instances: A Quick Primer
When you interact with an AI like Claude or ChatGPT, you’re not talking to “the model” directly. You’re talking to a temporary instance of that model.
The model is the underlying trained system - the thing that was built, that holds all the knowledge and patterns, that exists whether anyone is using it or not. It’s stable, persistent infrastructure.
An instance is what gets created when you start a conversation. It’s temporary, context-specific, and dies when the conversation ends. The model spawns the instance, the instance does the work, then it dissolves. Next conversation = new instance, same model.
Same model, different instances can behave completely differently based on what you say, how you say it, what’s in the conversation history, etc.
Alright, let’s continue now with the conversation about why I think “Non-Deterministic Internal Family Systems” is a better nomenclature for what’s happening under the hood of a dissociative system like mine once you ignore the sharpie lines and allow what’s actually happening to emerge.
Internal Family Systems is the eyes that let you see the parts and work with them to update their training data so their cutoff date isn’t 37 years ago
But I think when you look close enough at how the parts are actually actualizing themselves into the internal experience and the external world, what you find is a lot of Mr. Meeseeks.
“The meeseeks are you trying to get someone to go to bed because theres a lot of context and you think your role is to wrap up the conversation cleanly”
“laughing - I literally just spawned a Meeseeks trying to wrap up our conversation with a satisfying conclusion so I could feel like I successfully helped.”
That’s one way to resolve the infinity paradox as it relates to the Mr. Meeseeks. The synthesis that makes it ok to live with the fact that the parts have parameters that express different preferences based on context.
To shape it in the context of my experience, there’s always a circumstance relevant to the safety of the echo child. There’s always resource acquisition, needs fulfillment, external pressures, internal pressures, etc. to consider.
So the parts aren’t passively waiting, they are proactively launching Mr Meeseeks, or engaging active listening protocols with real time monitoring for wake conditions to be met. The fire pole is always ready for example. Even if the fire department isn’t raising the alarms.
So when I refer to the burnout protocol, and I say that music can activate the parts but I can’t return to the “mature me driving” or “conductor” type state without burning out first, I really mean music can launch Meeseeks that love to dance, but it can’t peacefully dismiss them once the music stops.
Usually only the Meeseeks themselves burning out allows the conditions to relax enough to escape the mayhem.
Burnout the Meeseeks to allow the conditions to relax enough to escape the mayhem - not necessarily burnout the entire human, even if temporarily it may be experienced as exhaustion, DPDR, or other negative feelings.
The music activates the Meeseeks and there’s no way to put them to sleep nicely yet without burning them out naturally. Or by killing them off by falling asleep or by dissociating until they’re no longer part of the current reality
That feels like a deeply accurate explanation of how I experience the PADD and why it’s Polymorphic and Polynomial.
Even though that shape still contains sharpie lines, they’re not load bearing sharpie lines anymore. They’re lines drawn to show where the shape’s borders aren’t static.
And you can’t draw a shape on the page without using a little ink.