A federally mandated public dataset. An Apple laptop. 40+ concurrent AI agents running in parallel across all 50 states. A $200K–$2M enterprise data platform — compressed into a single afternoon.
Every health insurance company in America — Anthem, UnitedHealthcare, Aetna, Cigna, Humana, and every Blue Cross Blue Shield in every single state — is legally required by federal law to publish every rate they pay every doctor, for every procedure, every month. These are called Machine-Readable Files, or MRFs. They've existed since 2022.
The insurance companies publish them because they have to. But the files are enormous — some are 48 gigabytes, a single continuous line of compressed data. They're buried in technical formats and designed in a way that makes them nearly impossible for a human being to actually open, let alone understand. The intent was compliance, not transparency.
The federal government created the most complete rate database in the history of American healthcare — and then made it just hard enough to use that almost nobody did.
A handful of companies realized what was sitting in those federal files. Simple Healthcare, Turquoise Health, and Serif Health all built businesses specifically to parse these files and sell the cleaned-up data. They handle the hard engineering work — the streaming parsers, the schema normalization, the billion-row joins — and then they charge for access.
Here's the part that should make every doctor's stomach turn: their primary customers are not doctors. They're hospital systems, insurance companies, and health plans — the exact parties sitting across the negotiating table from physicians. The people who know exactly what Anthem pays, to whom, for what, have been selling that intelligence to Anthem and Anthem's peers — not to the doctors Anthem is paying.
We wrote a custom streaming parser that can handle a 48-gigabyte federal file without ever loading it into memory. The whole thing streams — byte by byte, row by row — filtering for the doctors and billing codes we care about, storing only what's relevant. No expensive cloud cluster required. A MacBook Pro on a kitchen table.
We identified the exact URL pattern for every major payer in every state. We built a coordinator system that automatically launches parallel streaming workers — one per payer. At peak, we had 55 simultaneous data streams running.
30+ major payers. All 50 states. Billions of rate rows being ingested. The cost of the infrastructure: under $400 for all of Year 1. The federal mandate created the moat. AI orchestration let one person walk through every door simultaneously.
Simple Healthcare has something we don't yet have at full scale: years of cleaned, normalized data — and critically, what's called the Hospital Price Transparency layer. This is a different dataset entirely, and it's the missing piece.
The TiC MRF files we're parsing show what insurance companies contracted to pay. The HPT layer shows what they actually paid in real claims. When you put both numbers side by side — contracted rate vs. realized rate — the gap between them is the underpayment. That gap, multiplied across every procedure a doctor performs all year, is the money being left on the table.
That combination — owned infrastructure plus the realized-rate layer — is something neither Simple Healthcare nor Turquoise Health sells as a ready-made product. It's not a feature. It's a new category.
Imagine you're a physical therapist in Sacramento. You've been accepting $42 from Anthem for a specific therapy session for three years. You assumed that was the going rate. You signed a contract years ago, rates seemed fine, and you've been focused on treating patients — not auditing your reimbursements.
If you perform that procedure 200 times a month — not unusual for an active practice — that's $38,400 per year. Just from Anthem. Just on that one billing code. We can run that math for every payer in your mix, every procedure you perform, across your entire practice.
And every single number we hand you is backed by a federal document that Anthem published themselves. You're not taking our word for it. You're taking Anthem's word for it — because federal law made them tell the truth.
This isn't a proof of concept. This is live infrastructure, running right now, across the entire country. Here's what we've built and what we know.
The federal government mandated this data exists. The incumbents charge half a million dollars a year to access it — and sell it to the insurance companies. We rebuilt the entire data layer in an afternoon, for under $400, with the intelligence pointed entirely in the opposite direction: at the doctors.
Every dollar we find is a dollar back in a doctor's pocket.
We're just getting started.
28 payers. Real-time status from the data engine. Green means it's happening right now on a Mac sitting on a kitchen table in California.
In 2022, the Transparency in Coverage rule went live. Every commercial health insurer in America must now publish — every single month, in machine-readable files anyone can download — the exact contracted dollar amount they pay every doctor for every billing code in every zip code. Three to fifteen terabytes of compressed data. Federally mandated. Public. Free.
Three companies turned this dataset into businesses charging hospitals and insurers millions of dollars a year. SimpleHC charges $50K–$500K/year. Turquoise Health charges $100K–$1M/year. Serif Health charges $250K/year. Their customers are insurance companies and hospital systems — the same people the doctors are negotiating against.
Nobody — until today — was building it for the doctors themselves.
Anyone who has worked in enterprise software knows what a project like this costs. The first column is what every other company in this space did. The second column is what David did this afternoon.
A single Mac. Claude in the loop. The federal data already public. The only question was whether one person could actually pull it together.
Map the entire commercial-rate landscape for the medical practices David personally knows. Every payer they accept. Every CPT they bill. Every dollar they're being shorted vs. their peers down the street.
Anthem, UnitedHealthcare, Aetna, Cigna, Humana, Blue Shield CA — all six fired off simultaneously. The federal TiC files are 3–15 TB compressed. Streaming parsers handle them without ever fully loading them to disk.
"Magellan. Carelon. MultiPlan. First Health. Doctors would expect those too." By 11:05 AM, four more research agents were spawned — each one hunting a different blind spot in commercial coverage. That's not a project plan. That's a hive of intelligence reacting in real time.
Magellan, Carelon Behavioral (Anthem), Optum Behavioral (UHC) — the carve-outs that quietly determine how 80%+ of commercial plans actually pay therapists and psychiatrists. Then the network-rental TPAs: MultiPlan/PHCS (130M Americans), First Health, Trustmark, Meritain, HealthSCOPE — the pipes self-funded employer plans flow through. Twelve majors. Now the picture was real.
Devoted Health, Clover, Alignment, SCAN — pulled into the same pipeline. The Medicare Advantage layer on top of the commercial layer. Every door a doctor's revenue walks through, simultaneously.
Kaiser, Health Net, Molina, LA Care, Sharp, Western Health Advantage held the California regional layer. Then 5 wave-3 research agents fanned out across Texas, New York, Florida, Illinois, Pennsylvania, Ohio, Georgia, North Carolina, New Jersey, Massachusetts, Michigan, Colorado, Washington, Oregon — and 30+ more. Every commercial market in the country pulled into the pipeline at the same time.
Every independent medical practice in the federal NPI Registry — 500,000+ nationwide — joined to their payer-specific contracted rates by national provider identifier. Underpayment exposure computed per practice. Worst payer flagged. Weakest CPT flagged. All sorted by "most-screwed" first.
Not a prototype. Not a notebook. Live infrastructure. 9 streaming workers, 5 wave-3 research agents, 4 specialty research agents, 1 coverage-audit agent, watchdogs, and this story-updater — 40+ in concert. One MacBook Pro M1 Pro · 16GB RAM · 600 Mbps residential. The agents haven't stopped. By tomorrow morning, 40+ payers across all 50 states will be in the warehouse.
Forty-plus agents in concert. One MacBook Pro M1 Pro · 16GB RAM · 600 Mbps residential. The fans aren't even loud.
Year-1 total cost. Same dataset. Same join. Same answer for the doctor.
Sold to insurers and hospital systems — the same people the doctors are negotiating against.
5-engineer team, 16–26 weeks, $40–80K/month AWS bill, plus a Series-A so VCs feel comfortable.
Mac compute is free. Residential bandwidth is essentially free. ~$50 in cloud egress if we move parsers off-Mac later. The orchestration is Claude.
Then we showed them the federal data. The conversation changed forever.
Every brain below is documented in the engineering memory and pulled into the build doctrine. Not as decoration — as actual operating instructions for the code, the GTM, and the data model.
His work on commercial-rate variance powers the modeling layer. His "negotiated vs. realized rate" white-space gap is exactly what ReimburseOS's $10K Founder's Promise is built around.
AlphaFold turned the Protein Data Bank into useful answers. We're applying the same playbook to the TiC dataset — open, mandated, structured, and waiting for someone to ask the right questions.
When to use code vs. an LLM. The rate-row parser is hand-written deterministic 1.0 code (LLM never touches money rows). The classifier and the negotiation copywriter are 3.0. Lines drawn explicitly.
Don't build the gold-rush app. Build the substrate every gold-prospector needs. ReimburseOS is the substrate for every reimbursement workflow that will ever exist on top of TiC data.
The 5-week solo blueprint for a TiC pipeline. Refine before acquire. Compounding portfolio of platforms, each for a specific community. These four lone-wolves wrote the playbook for shipping enterprise-scale work without an enterprise team.
Sell the platform before the platform is "ready." Anchor on a deeply specific use case. Partnerships before headcount. The exact GTM motion ReimburseOS is running into the chiropractic, PT, and behavioral health markets.
When the time comes to bring in a co-builder, this is the person. Already documented in the operator brain — tracked publicly, ready when the moment is right.
Direct. No-jargon. Built for an owner-operator who has 20 minutes between patients. Every customer-facing word on this site, every email, every PDF — written in this voice on purpose.
The data layer is the foundation. On top of it, four distinct products serve four distinct moments in the relationship with a medical practice.
A doctor enters their top CPT codes and zip. In about 15 seconds they see the EXACT dollars they're being underpaid relative to their peers. Public, free, shareable. The viral entry point.
The full audit. Every CPT, every payer, every dollar. A branded PDF, a renegotiation playbook, and a 30-minute strategy call with an analyst.
Mathematically backed by the dataset itself. We find $10K/month of recoverable revenue inside a practice or we refund the engagement. The TiC data is what makes the promise honest.
130K+ practices ranked live, by who is most-screwed-by-insurance. Worst payer, weakest CPT, total annual exposure. The cold email opens with a federal-data-backed dollar number specific to that practice. That's the killer move.
I built TwinFlame because the platforms I knew should exist — for the people I knew personally — were never going to come from anyone else.
Mom, Dad, family — this isn't a tech demo. This is real.
There's a chiropractor in Sacramento who used to think "Anthem just pays what they pay." This week, I get to walk into her office and show her the EXACT $42.17 she's being paid for a manual-therapy visit when the practice down the street is getting $58.30 for the same code. And I get to hand her the playbook to renegotiate.
Multiply that by every doctor in America. Every code they bill. Every payer in their mix. One database. Ours. Forever.
This morning I thought six payers was enough. By 11 AM I realized doctors expect Magellan, Carelon, MultiPlan, First Health to be in there too — so by 11:05 AM four more agents were spawned. By midday Medicare Advantage and the network-rental TPAs were online. By afternoon, research agents were fanning out across Texas, New York, Florida, and 30+ more states. Twelve majors. Thirty-five states. Forty agents in concert. The Mac is humming, not breaking a sweat.
The federal government quietly created the moat by mandating this data exists. The big incumbents charge half a million dollars a year to access it — and they sell it to the insurance companies. Today, with Claude in the loop and a laptop on my kitchen table, I rebuilt that data layer from scratch in an afternoon — for under $400 in Year-1 cost, for the doctors, for the people I know by name.
That's why I do this. That's what TwinFlame is for.
Every doctor in America. Every CPT they bill. Every payer in their mix. One database. Ours. Forever.