Stimpunks Foundation

Mutual Aid and Human-Centered Learning for Neurodivergent and Disabled People

Playing pieces arranged on a network of lines

AI and Competency Networks

🗺️

Home » AI at Stimpunks » AI and Competency Networks

A competency network is how a community knows its own strengths. It is the living map of who knows what, who can help with which thing, who to sit beside when you are stuck. It is not an org chart. It is built from disclosure — people showing each other their actual thinking, their rough drafts, their half-formed skills — until the group can make sense of itself.

Generative AI is both an asset and a compromise to that network. It can widen the web or quietly cut its threads. The same tool does both, and the difference does not announce itself. This page names the fork and offers a test for telling the two apart — so we can use AI to make each other more legible, not less.

Our stance here is the stance that runs through all of our AI work: human-led, community-grounded, counter-deficit, and clear-eyed about power. AI helps generate possibilities. The network decides what belongs.

Competency Networks Run on Legibility, Not Visibility

We often confuse being seen with being understood. They are not the same, and the confusion makes our thinking about AI poorer. As Tressie McMillan Cottom draws the line: visibility means you can be perceived, the way a lamppost or a dog can be perceived. You are a category. You exist. Legibility is different. Legibility means other people have the capacity to make sense of you — that you are not only a thing that exists, but a thing that is supposed to exist, in this space, this conversation, this exchange. Legibility is relational. It happens between people. And that makes it a kind of situational privilege, unevenly granted.

Being legible is a kind of situational privilege.

Tressie McMillan Cottom, essaying

This is why our whole culture points at legibility. We default to open. We show our work. We iterate our rough drafts in public. We treat legibility as part of belonging. None of that is decoration. It is how the network stays able to make sense of the people in it.

Vulnerability Is the Currency

Legibility is not free. It is paid for with vulnerability. You show a rough draft of your reasoning. Someone meets it with theirs. Over time, each of you learns what the other actually knows and actually needs. That gradual, mutual disclosure is the competency network being built. There is no shortcut around the paying-in.

Here is where AI can quietly cut a thread. When one person outsources their side of the exchange — using AI to compose a reply that hides how they actually think — they collect the other person’s disclosure without paying in themselves. The responder still exposes their reasoning. The AI user does not. That is not merely opacity. It is extraction. And it destroys the thing gradual disclosure was for: the slow, reciprocal building of trust and mutual knowledge.

We prize communication that is authentic, candid, and vulnerable. The DEI backlash has already eliminated many spaces where people were allowed to be anything other than neutral, professional, and neuronormative. When AI also removes the labor that used to force genuine contact, authentic space gets squeezed from both directions: no permission to be real, and a tool that lets everyone avoid the cost of being real. We need space to not be neutral. To be authentic, transparent, and legible to each other on purpose.

Authenticity and vulnerability are revolutionary.

The Journey of Undoing, SITI Girl Miami

The Same Tool, Two Functions

The line we are drawing is not AI-use versus no-AI-use. It cannot be. For many Autistic and otherwise Disabled people, AI is the scaffold that finally lets them say the true thing they could never get out unassisted. That use makes a person’s actual thinking more visible to the group, not less. It closes a double empathy gap instead of widening one. It increases legibility. It is augmentation in the fullest sense.

The plausible-deniability use does the opposite. It screens the person’s thinking behind fluent output so that no one can quite tell what they know, believe, or need. From the outside, both look identical — “using AI to reply.” The difference is entirely in the relational function. One reveals a person to the network. The other conceals them from it. The problem was never the tool. It is what the use does to the web.

The Legibility Test

Because the two functions look alike, we need a running check rather than a rule about tools. One question does most of the work:

Does this use make our competencies more legible to each other, or less?

Augmentation increases mutual knowledge of who-knows-what. Replacement quietly reduces it. When AI use is standing in for the network instead of feeding it, the balance has tipped — and it will not tell you it has tipped. You have to watch for it. Some early signs the thread is being cut rather than woven:

  • You know someone is present and productive, but you can no longer tell what they actually think.
  • The vulnerability in an exchange runs one direction. One person discloses; the other returns polished, unreadable output.
  • People surface less half-formed thinking to each other, because they process privately with AI instead of together.
  • The map of who-knows-what stops updating. New strengths in the group go undiscovered.

Practices: Using AI to Increase Legibility Across the Team

The point is not to police anyone’s tools. It is to use AI in ways that feed the network rather than replace it. These practices grow directly out of how we already work.

  • Show the prompt, not just the polished output. Making your AI use visible is a form of showing your work. It lets the group see your thinking, not only your result.
  • Iterate in public. Publish the rough draft. Iteration and defaulting to open are how legibility accrues. A finished artifact that appears from nowhere teaches the network little about you.
  • Use AI to surface your reasoning, not to stand in for it. If the output would let a colleague understand how you think, good. If it would let you avoid being understood, that is the screen — put your own reasoning back in.
  • Do not outsource the first vulnerable move. Match disclosure to disclosure. If someone shows you their rough thinking, meet it with yours — not with something composed to look invulnerable.
  • Name competencies out loud. Affirmation loops and a culture of props are not fluff. They are how the map of who-knows-what gets drawn and redrawn. Say who helped and what they are good at.
  • Body-double with AI in the open. Working alongside AI is a real access practice. Doing it visibly — narrating what you are using it for — keeps it augmentation rather than a private substitute for the people around you.
  • Run the legibility test before you send. Does this make me more legible to my team, or less? Send accordingly.

AI can widen the web or quietly cut its threads. We choose the web — legible, authentic, and built by people who keep paying in.

Read Alongside