Every other search engine stores data, then builds a separate index to find it. Trinity compresses once. The compressed form is directly searchable. That's not an optimization. It's a different result.
Every search engine on earth answers one question: does this document contain these words? Trinity answers a different question: what is the structural relationship between these concepts? The result contains information that doesn't exist in any other system's output.
These are not faster versions of what other systems do. These are operations that produce results no other system can produce at all.
Every document is classified by stance: affirmed, denied, mentioned, or absent. Polarity is encoded at the storage level, not guessed from text. You don't get keyword matches. You get structural relationships.
Evidence of absence, not absence of evidence. Find documents where a relationship was explicitly ruled out. Traditional search can't distinguish "drug X caused Y" from "drug X did NOT cause Y."
Take a document from one database and find structurally similar documents in a completely different database. No schema mapping. No ETL pipeline. The compression creates the bridge between domains.
Every document has a compressed fingerprint. Measure deviation from the norm — outliers surface automatically. No ML pipeline. The compression structure reveals what's unusual.
Content that isn't directly indexed still becomes searchable through Trinity's compression. Data discovers its own search paths without increasing storage overhead.
Hypothetically reclassify assertions and see how document classification changes in real time. Counterfactual analysis baked into the search layer. "What if every NAY became YEA? How many bills pass?"
Seven federal databases. 65 million records. Two servers totaling $48/month. Every number below is live and verifiable.
| Database | Source | Documents | DB Size | RAM (search) | Search Speed |
|---|---|---|---|---|---|
| FEC | Federal Election Commission | 39,990,307 | 131 GB | ~400 MB | 0.4 ms |
| Congress | Congressional Votes | 24,000,000 | 139 GB | ~200 MB | < 100 ms |
| FAERS | FDA Adverse Events | 394,516 | 1.1 GB | 8 MB | 0.3 ms |
| CVE | Vulnerability Database | 331,966 | 1.1 GB | 12 MB | 0.4 ms |
| OSHA | Workplace Injuries | 102,922 | 619 MB | 6 MB | 0.2 ms |
| Exploits | Exploit-DB | 46,014 | 543 MB | 10 MB | 0.3 ms |
| Bills | U.S. Legislation | 22,496 | 2.9 GB | 14 MB | 0.5 ms |
| Total | 64,888,221 | 276 GB | ~650 MB | < 1 ms | |
| Elasticsearch | equivalent workload | — | ~800 GB | 16–48 GB | 10–200 ms |
The things that make this system fundamentally different are the same things that make it fundamentally cheaper. The efficiency isn't a tradeoff. It's the architecture.
A Raspberry Pi. $15 for the board. $13 for a 128 GB microSD. $5 for power. $33 total, and it runs everything you see on this page. 65 million records. Seven federal databases. Structural polarity. Cross-database search. The entire industry spends billions on search infrastructure that delivers less than what a $33 board delivers with Trinity.
One Pi is a search engine. Ten Pis is a distributed cluster. A thousand Pis is a global intelligence network. $330 for a 10-node distributed search cluster. Elasticsearch can't match it at $30,000. No central server. No cloud bill. No single point of failure.
Most people on earth have never searched a federal database. Not because the data is secret — it's all public. Because the infrastructure to search it costs tens of thousands of dollars. Trinity eliminates both barriers. $33 hardware. Zero specialists. A mesh of Pis in a village does what a data center in Virginia does.
A Raspberry Pi. A $96 VPS. A 64-core server. A 64-node cluster. The architecture doesn't change. The numbers just get larger. On elite hardware: parallel sharding, sub-10ms cross-database search across a billion records, 500 queries per second on one machine. Elasticsearch on the same hardware still gives you one answer. Trinity gives you four.
No CS degree. No high school diploma. Twenty years swinging a hammer in the oilfield until the work broke my body and the industry moved on. When the layoffs hit, someone on Twitter posted #LearnToCode as a joke aimed at roughnecks losing their jobs to green energy development.
I didn't take it as a joke. I took it as a dare.
Five years of self-teaching. No bootcamp. No mentor. One safety net — a friend everyone told would never be more than the wealth he was born into. He bankrolled five years on what started as a wild bet on each other and became a shared conviction — to go boldly and do really cool things for everyone.
Trinity is what came out the other side. A search architecture that runs on $33 hardware, serves 65 million public records on 650 MB of RAM, and does things that systems backed by billions in venture capital have never done. Built by a roughneck who was told to learn to code — and did.
If this system can go everywhere for almost nothing, it's because it was built by someone who came from almost nothing. That's not a limitation. That's the architecture.
65 million documents. Seven federal databases. Four-state structural search. Running right now.
Open Trinity Search