// operational data infrastructure · tampa, fl
From first line to production. Verified by replay.
Your data is born on physical equipment, inside a compliance boundary, and the record has to hold up afterward. We build the layer that captures it at the point of work — and proves it by replay.
> capture --line 04 --src ttyS2
✓ 1,142 events · comms DOWN
> sync --on-contact
✓ 1,142 / 1,142 · ledger appended
> verify --replay ledger.chain
✓ chain intact · 0 discrepancies
// illustrative — not customer data
Direct US Navy and Air National Guard work — 100% US-based engineering
// vice chair, ieee ras fl west coast · coder radio, 650+ episodes since 2012
// the problem
The record gets reconstructed after the fact.
On most lines, operational truth is assembled later — from memory, paper, and spreadsheets. A reconstructed record fails the same three ways: it fails the audit, it fails the dispute, and it fails the AI initiative that assumed the data existed.
The fix is not another report. It is capturing the record at the moment of work, on the equipment where it happens — and being able to prove it afterward.
// the platform
Three layers. Your hardware. Your keys.
Gateways, sensors, serial equipment, printers, scanners. Kernel-level timestamps, store-and-forward through outages.
captures with comms down
PostgreSQL, Parquet, and DuckDB on hardware you own. One system of record for everything the line produces.
system of record, your keys
Process capture, provenance ledger, compliance outputs — AS9110 evidence, MIL-STD-129 labels, custody trails.
answers, not raw feeds
NOTHING CROSSES THIS BOUNDARY THAT YOU DID NOT DECIDE TO EXPORT. SYNC ON CONTACT, NEVER BY DEFAULT.
// verified by replay
A record you can verify. Not a report you have to trust.
Every event is captured deterministically at the point of work and appended to a hash-chained ledger. Anyone with a laptop can re-verify the entire record independently — no vendor in the loop, no dashboard to believe.
// success criteria — checkable, not adjectival
> run acceptance-tests
✓ zero loss through comms outage
✓ every record replay-verified
✓ air-gapped, on a laptop
3 passed · 0 failed · 0 adjectives
// illustrative — not customer data
// "we'll have our it guy do it with ai"
Fine. Then run this test.
Pull the network plug mid-shift for four hours. Count the records that are missing when it comes back.
Ask for the hash chain proving no record was edited after the fact. Wait for the answer.
Run it in the room where your data actually lives — the one with no internet and an ITAR sign on the door.
You cannot prompt your way to a kernel-level timestamp on a serial device behind a compliance boundary. The model was never the hard part. The room is.
// proof
Running now. On real equipment.
// independent coverage
The Mad Botter has been covered since 2018 for custom software, aerospace engineering, legacy-data work, and its founder-led approach.
// custom engineering
Your equipment isn't standard? Good.
When the Fit Check finds a device the platform doesn't capture out of the box, the same senior engineers who built the platform build the driver. Scoped in writing, priced fixed, delivered onto your line — then it becomes platform capability you keep.
No customer-specific engineering begins without an agreed scope and price. Never hourly.
// the offer
One price to prove it. One price to run it.
Send your equipment and protocol list. Get a one-page written fit verdict within two business days. Keep it even if you never talk to us.
One site, one line, thirty days on real equipment. Acceptance tests published before we start.
On commodity hardware you own. The full $1,500 pilot credits against your first platform invoice. Expand from evidence — one line at a time.
// free · written verdict · no meeting required
// founder
The founder is not a black box.
// michael dominick — founder
I'm Michael Dominick. I've hosted Coder Radio since 2012 — 650+ episodes — and I write at dominickm.com. If you've listened, you already have a decade of evidence about how I think about software, which is more than you'll get from any pitch.
Founder, The Mad Botter · Vice Chair, IEEE Robotics & Automation Society, Florida West Coast Section · Host of Coder Radio since 2012
// coder radio · latest episode
655: Homebrew Mike McQuaid// live from dominickm.com/feed
// faq
Straight answers
What does The Mad Botter actually do?
We build the layer that captures operational data where it is created — on gateways, sensors, serial equipment, printers, and scanners — stores it on hardware you own, and makes the resulting record provable by replay. Capture, a sovereign data layer, and the pipelines that turn it into compliance outputs.
How is this different from a workflow tool or an integration platform?
Those move data that already exists in software. Our problem starts one step earlier: the record has not been created yet, because it lives on a machine, on paper, or in someone's memory. We capture it at the point of work, through outages, inside boundaries where cloud tools cannot run at all.
Why can't we just do this ourselves with an LLM?
Sometimes you should. But the model was never the hard part. Kernel-level timestamping on a serial device, store-and-forward through a four-hour outage, and a hash-chained ledger that re-verifies on a laptop inside a room with no internet are not prompting problems. Run the three tests on this page against any alternative.
What does "verified by replay" mean?
Every event is captured deterministically at the point of work and appended to a hash-chained, append-only ledger. The whole record can be independently re-verified afterward — by you, on your own hardware, with no vendor in the loop. That is what makes it hold up in an audit or a dispute.
Can you work inside ITAR, CMMC, NIST, GxP, or air-gapped environments?
Engineered for ITAR, EAR, CMMC, and NIST environments. Air gaps, no-cloud policies, and facilities where data cannot leave the building are normal operating conditions here rather than exceptions. We are designed to support your compliance posture; we do not certify compliance.
What does it cost to find out whether it fits?
Nothing. The Equipment Fit Check is free: send your equipment and protocol list and a senior engineer returns a written one-page verdict within two business days — what captures out of the box, what needs a scoped driver, and what is out of scope. No payment and no meeting required.
What if our equipment is old, undocumented, or on paper today?
That is the normal case. Serial devices with no documentation, unlabelled PLCs, and processes that are a clipboard today are the work. Where a device is not captured out of the box, the driver is scoped in writing and priced fixed before any engineering starts.
What do we own?
Your operational data, the infrastructure it runs on, your configuration and integrations, and the deployment artifacts — in open formats, readable without us. The Mad Botter retains and licenses the reusable platform core. There is no kill switch and nothing phones home to disable itself.
// start with the line you already have
Send us the equipment list. We will tell you what we capture.
A written one-page verdict within two business days: supported, needs a scoped driver, or out of scope, with a line of reasoning for each. Free, no meeting required, and yours to keep whether or not we ever talk.
The Mad Botter builds operational data infrastructure for environments where the cloud is not an available answer — capturing the record on the equipment where work happens, inside your compliance boundary, and proving it by replay.