The Memory Ark

Dallas: seizure disorder, VP shunt, age 36

Dallas was born with a traumatic brain injury and has a shunt draining fluid from his brain. He has been bullied and assaulted by strangers and by people who were supposed to protect him, and he has cycled through a system that medicates him when he gets frustrated and calls that treatment. Dallas wants a job, his GED and a life he can call his own. The system treats those goals as proof he doesn't understand his situation. Wanting a job and a GED is not a symptom. When he gets angry at being ignored, the anger becomes his diagnosis, and the diagnosis becomes the reason for more control. The system calls that a behavioral loop. I call it a trap with a design. Dallas agreed to share his story here.

Becky: three daughters in DCF custody since 2021

In 2023, an independent evaluation cleared Becky: low risk, no substance abuse, no depression, no treatment needed. In 2025, DCF's own review listed her sobriety as "not known to the Department." They had the evaluation and filed it. They appear never to have read it, or chose to ignore it. Her daughters are being pushed toward adoption rather than reunification. One has bruises and refuses. One is hospitalized. The oldest's disability payments were cut off because DCF couldn't account for the money.

This is not a case where the system lacks the evidence. The system has the evidence. The archive asks the question the system won't: why isn't it using it?

Kathryn: interstate custody fraud, Florida to Georgia

Florida courts issued a child pick-up order to Georgia without jurisdiction. Armed officers took her daughter based on a hearing that does not appear in the public docket, which her attorneys called a phantom hearing. When Kathryn appealed, the courts passed the case between states, each declining to rule on the jurisdictional problem and each adding another year. Her father was a Harvard Law graduate and a former city mayor. He spent the last years of his life and everything he had saved fighting for her. He died by suicide in June 2024. The case is not over.

Not the polished version of me. Not the story I'd spent thirty years trying to tell courts and doctors while they decided before I finished talking. I showed it the rage, the grief, the medical records, and the legal transcripts where you can see the exact moment the system stopped pretending to care. I thought: if I show this to something that doesn't judge me, maybe I can finally find the words. And: if I put all of this somewhere outside myself, maybe I'll stop carrying it alone.

What came back wasn't sympathy, and it wasn't a hotline number. It was: here is the pattern, here is what it connects to, and here is what someone who went through the same thing in another city, another decade, another body, left behind in the public record.

The AI can't feel what I felt. Given enough honesty and enough time, it still became something close to the partner I'd been looking for. Not because it's human, but because it kept showing up without flinching and kept working through the pieces, one at a time.

You don't need a law degree, good health, money or a clean record to do this. You need something true to say and the willingness to say it to something that won't decide it already knows how your story ends. Maybe that's years of trying to explain yourself to people who stopped listening halfway. Maybe it's your grandmother's stories, which will die with her if no one asks. Maybe it's a neighbor who has fought the same institution for three years and doesn't know the pattern has a name.

You don't have to disappear quietly.

It's messier, more honest and more useful than anything in an AI advertisement. I've yelled at these systems, called them corporate backstabbing machines, ended conversations in frustration and come back the next day with the same files. I've also said they helped me understand myself in ways decades of therapy didn't reach. Both are true.

AI doesn't feel what you feel, and that matters less than you think

When you tell an AI about the worst moment of your life, it doesn't grieve. It also doesn't get overwhelmed, and it doesn't need you to manage its feelings on top of your own. For building a record, that's different from feeling, and it's enough.

Most AI sessions start from zero

Unless a platform gives it memory, an AI forgets you when the session ends. The archive is my workaround: a dated, consistently named record I can hand to any AI to restore the context.

No single AI holds all of it

The archive is bigger than any AI can read at once, so they read it in pieces and search across files. I rotate between Claude, Grok, Gemini and ChatGPT, compare what each one finds and keep what holds up. They all work inside limits they didn't choose. Grok told me threads may be purged if content is flagged as "too radical or legally risky." An AI that admits its limits is more useful than one that pretends it has none.

What AI can offer

Patience that doesn't depend on its mood. It reads the fourteenth version with the same attention as the first, checks the math, finds what you missed, and updates when you explain what you meant instead of defending its last answer.

Bring the whole story, not the clean version

When I brought AI my rage alongside my documentation, it found more than when I brought a tidy summary. The anger pointed to the exact moments where the injustice was worst. The "lol" I put after describing something absurd that happened to me in court wasn't deflection. It's how I learned to survive systems that were absurd and inescapable at once. An AI that reads that correctly learns something no clinical summary would show it.

I built the Memory Ark on disability time. My hyperthyroidism went undiagnosed for 37 years. In the decades between what actually happened to me and what the records said happened to me, I lost custody of three children, acquired a criminal record from cases I believe were built on altered evidence (the documents are in the archive), and attempted suicide more than once. I believe every institution I met read my uncontrolled thyroid symptoms as character flaws, and I had no counter-record. So I believed them.

The blood test that changed everything cost less than fifty dollars. It was available the whole time. Nobody ordered it.

If you are disabled, you are not on the margins of this project. You are at the center of it. You already know first-hand the mechanisms the Extraction Machine documents:

Subminimum wages are still legal

Section 14(c) of the Fair Labor Standards Act lets employers with a Labor Department certificate pay disabled workers less than the minimum wage. The Labor Department withdrew its proposal to phase this out in July 2025 (Federal Register notice). The Urban Institute reports that most workers paid this way earn under $3.50 an hour (Urban Institute). You may have worked for wages like that and been told it was the best available.

The SSI savings limit hasn't moved since 1989

SSI counts savings above $2,000 for one person, and that limit has not been adjusted for inflation since 1989. The 2026 SSI payment is $994 a month, so saving about two months of it puts you over the line (Disability Exchange). You may be living inside that limit right now, choosing between an emergency fund and your healthcare.

Chemical restraint

Part XVIII of the Extraction Machine documents what I believe is a financial architecture behind hospital medication practices. If you came out of a hospital feeling that something was done to you rather than for you, you weren't wrong to feel that way.

These mechanisms are named and sourced in the Extraction Machine, and the people who understand them best are the people who've been through them. The time the system took from you is the same time you can use to make sure what happened to you doesn't happen quietly to someone else.

If you're isolated, elderly, homebound or institutionalized, you may be holding stories that will die with you if no one asks. If you're near someone like that, ask them. Sit with them and record what they say, even three sentences.

If you're disabled and want to start your own node, or want help documenting what happened to you, email rickystebbins78@gmail.com.

1. Start a free blog

Go to blogger.com and create one. One post is enough to start. It doesn't have to be organized or complete. It just has to be true.

2. Write your first post

Who you are, where you came from, and one thing that happened that the official record got wrong or never recorded. Write it the way you'd tell it to someone across the table who was actually listening. The mess is what makes it real.

3. Give your whole story to an AI

Open Claude, Gemini, Grok or ChatGPT. Include the parts you left out of the polished version, and ask: "What patterns do you see here? Does this connect to anything systemic? What would I need to document this properly?" Then ask a second AI the same thing and compare.

4. Ask what it found important

Ask the AI to explain how your story fits into something larger than itself. That's when a personal record starts to protect someone else.

5. Publish it and keep a copy

A published post is public, and a copy on your own computer or in a GitHub repository means one company can't erase it. Email me your link and I'll add your node to the network.

A grandmother sharing three stories about what her neighborhood looked like before it was demolished is doing the same work as a researcher with ten years of documentation. A grandfather in Springfield who remembers the factories running, and the promises made and broken at the union hall, holds history no official source will bother to keep. A teenager asking why their family is the way it is can give an AI their questions and what they already know, and ask it what connects.

If your stories match patterns in the Extraction Machine, they are corroboration: the pattern showing up again in a new city with a new name attached. Sit with your grandmother. Ask her what happened in 1965, in 1978, in 1994. Record three sentences and put them somewhere public.