WE BUILD. THEY TALK.
What Sierra Special Investigations and Project Milk Carton Actually Are — and Why We Fund Our Own Mission
WE BUILD. THEY TALK.
A Project Milk Carton Statement of Work · July 2026
They say we don’t do anything. So here’s the record — the code, the patents, the data, and the receipts. Read it, then tell us again who isn’t doing the work.
The milk carton, reborn — the missing found through technology and transparency.
It started with a question nobody had answered in thirty years
Somebody in a room asked out loud: whatever happened to the milk carton kids?
You remember them — the faces of the missing on the side of the carton at breakfast. Somewhere along the way, that stopped. The cartons went blank. The kids didn’t. Every year hundreds of thousands of children are reported missing in the United States, and the machine built to protect them — child protective services, foster care, the billions of federal dollars that flow through it — became a place where accountability goes to disappear.
The honest answer to that question was: nobody did anything. So we decided to be the ones who did.
First we had to build the thing itself
The mission needed a home on the internet. People said they’d build the website. They talked about it. They never did it. So it got built anyway — funded out of pocket, stood up, and put online. Then the organization was made legitimate the right way: Project Milk Carton was established as a 501(c)(3) public charity (EIN 33-1323547), with a real board, real governance, and a real mission — child-welfare transparency and missing-children awareness.
That’s the pattern you’ll see over and over in this story: less talking, more building.
The record — with dates — because this predates the noise
People make claims. Timestamps don’t. Here is the dated record of this work, and the important part isn’t any single line — it’s that the building was well underway long before it became something anyone argued about.
November 2024 — It begins. The first code is written and published: public-records and IRS-990 accountability analysis — the foundation everything else is built on. (This date is fixed in public version-control history. It cannot be moved, edited, or backdated after the fact.)
December 2024 – January 2025 — Project Milk Carton goes online. Missing-children data, NCMEC feeds — the mission takes its shape.
March 2025 — The core architecture (internally, “the Guardian Stack”) is designed. Everything built since was born from it.
June–July 2025 — The decision-mapping engine (hundreds of decision nodes across every U.S. jurisdiction) and the financial-intelligence database (340+ million records, $148 billion in grants tracked) come online.
August 2025 — A federal-funding transparency investigation is built, flagging billions of dollars in accounting irregularities.
November 2025 — The current public platform, GuardiansWatch, goes into production at projectmilkcarton.org.
By late 2025 — The defense-grade, fully-offline AI work (MERLIN) begins.
By early 2026 — The autonomous research agents mature, the child-safety education system (ECHO) begins, and a portfolio of provisional patents is filed to document the inventions.
That is nearly twenty months of continuous, documented building. Whatever anyone wants to claim about who did what and when — the record is public, it is timestamped, and it starts in 2024. Set your calendar next to ours and see whose work came first.
What we actually built — the receipts
This is the part the critics skip, because it’s inconvenient. Here is a factual inventory of what has been built, and it is all real, running, and inspectable:
A live public platform. GuardiansWatch — a working website that maps how child-welfare decisions are actually made across all 51 U.S. jurisdictions (50 states plus D.C.), tied to the money that funds them.
Roughly half a million lines of working software. Across the organization’s systems there are on the order of 463,000 lines of hand-written code and hundreds of thousands more in supporting infrastructure — autonomous research agents, investigation tools, and public-facing applications. This is the output of a software company, built to serve a nonprofit mission.
A financial-transparency intelligence layer. A database of 340+ million public records — federal grants ($148 billion tracked), nonprofit tax filings, campaign-finance contributions, foster-care funding flows, and missing-children data — cross-referenced so the money can be followed.
252 custom investigative tools built to query all of it.
Municipal-scale coverage. Public records crawled from more than 19,000 U.S. municipalities.
Custom-trained AI models. Not just using AI — training it. Multiple language models fine-tuned in-house on child-welfare law and decision data, running on the organization’s own hardware, offline, under its own control.
You do not have to take our word for any of it. Parts of it are open-source and public right now, including an open legal knowledge graph of the entire U.S. child-protection decision system (over 2,000 mapped decision points across every jurisdiction). Go read the code.
This is built to a defense-grade standard
The architecture behind this work was engineered to a standard far beyond a typical nonprofit’s website — because the same core problem shows up in national security. By late 2025, one internal system — MERLIN — was built to run fully offline, under its owner’s control, in environments where the network is denied, degraded, or unavailable — what the defense world calls DDIL. AI that listens, organizes, engineers, and remembers, without ever “calling home.”
That principle — decentralized AI that you own and run yourself — is where serious capability is heading, in defense and in public accountability alike. The reason it matters here is direct: the same architecture that can stand up in a contested, comms-denied environment is exactly what lets a small nonprofit run a nationwide transparency operation without depending on anyone else’s servers, funding, or permission. Self-owned technology is self-owned mission.
The invention record: 23 patents
Talk is cheap. Invention is not. The organization’s technology is protected by a portfolio of 23 provisional patents — a documented, dated invention record spanning:
Child-welfare and anti-fraud systems — models to detect diverted funds, optimize foster-care matching, and score the transparency of the agencies handling children.
Tactical and defense AI — offline deployment architectures, human-machine teaming, and no-code AI configuration for the field.
Educational AI — adaptive systems to teach children safety.
This is a real body of intellectual property, owned by the company and dedicated to the mission.
Here is the part that actually matters: how we’re funded
This is where we are the opposite of what has become normal in this space.
We fund our own mission.
The technology, the patents, and the infrastructure are owned by Sierra Special Investigations, LLC — a private company owned by the same people who serve the mission. SSI is an active, revenue-producing enterprise — its members hold real-world credentials and perform contracted work in national security, including as a federal subcontractor supporting U.S. Government anti-terrorism assistance efforts. That real revenue, and the members’ own labor and intellectual property, is what funds the mission. SSI licenses the entire technology stack to Project Milk Carton for one dollar a year.
The board underwrites the work. A real company that earns its own money underwrites the work. The people building this put their own resources into it — not the public’s.
What we do not do is take public donations and route them into the pockets of a for-profit board. We do not run a charity as a pass-through that converts good-faith donor dollars into salaries and consulting fees for a private, for-profit entity sitting behind it. That model exists in this industry. It is the exact model we were built to expose — and it is the exact model we refuse to operate.
The difference is simple enough to put on a bumper sticker: the people who talk about the children raise money off them. We spend our own money on them.
What this has to do with saving children — directly
Fair question, and here is the honest, quantitative answer.
Child welfare in America is not primarily a compassion problem. It is an accountability and money-flow problem. Tens of billions of federal dollars move through the system every year. When that money is diverted, wasted, or spent on everything except the child, kids fall through — into neglect, into the pipeline that feeds trafficking, onto the missing lists that used to be printed on cartons.
Transparency is the intervention. Every dollar you can trace is a dollar that is harder to steal. Every decision you can map is a decision someone can be held to. Every agency whose failures are made visible is an agency under pressure to fix them. The tools described above exist to do exactly that at national scale: follow the $148 billion, map the decisions in all 51 jurisdictions, surface the failures, and put the receipts in the hands of the people — journalists, investigators, legislators, and parents — who can force change.
We are not going to insult you by claiming a database saved a specific child this week. What we will tell you, plainly, is this: you cannot fix what you cannot see, and for thirty years the people running this system have counted on no one being able to see it. We built the thing that lets everyone see it. That is how this saves children — not with a slogan, but by ending the darkness the failures hide in.
To the ones doing the talking
You have accused the builders of not building. So here is the standing invitation: come look. The code is real. The IP’s are real. The database is running. The website is live. The models are trained.
We are not asking you to believe us. We are telling you where to look.
We build. You talk. The children can tell the difference — and soon, so will everyone else.