Evidence
before hardware.
Our first product ambition did not survive its opening resource budget. That correction did not shrink the mission. It made the mission real.
The dream was too easy to describe
A small, quiet system. Long endurance. Lower cognitive burden. An underwater life-support platform that feels less like managing a machine and more like doing the work you came to do. It is an exciting picture—and a dangerous place to begin engineering.
A rendering can hide every assumption. A physical budget cannot. Oxygen demand, carbon-dioxide production, absorbent capacity, breathing resistance, energy, pressure, temperature, reserve, escape, maintenance, and certification do not negotiate with the shape of a backpack.
A serious program begins when the idea gains the ability to say “stop.”
The first useful result was “the budget does not close”
We built a deliberately simple screening model from public technical references. Its purpose was not to size a dive system or recommend a duration. Its purpose was to make the first-order resource flows visible and reversible. Under the initial combined ambition, the mass and volume logic did not close with enough evidence to justify moving toward hardware.
That result removed a seductive product promise. It also created the first real product requirement: every assumption, boundary, correction, and failure response needs a place where another reviewer can inspect it.
Evidence v0.2
Evidence v0.2 is our internal, non-human research workbench. It freezes a declared mission boundary, shows a reversible first-order budget, runs one implemented synthetic-fault demonstration, keeps five additional scenario entries explicitly marked as planned, connects hazards to requirements, and exports a watermarked local research snapshot. If an input leaves the declared domain, the tool stops and produces no budget result.
Its state labels describe only what happened inside a particular software model. They do not describe a device as safe, certified, ready, or fit for any dive. It has no device connection and no control authority.
What changed this week
- The product idea now has explicit non-human boundaries and stop conditions.
- A preliminary hazard analysis maps ten catastrophic hazard families to requirements and planned verification.
- Ambiguous “pass” and “accepted” language was removed after a red-team safety review.
- Public interaction, wearable hardware, breathing tests, pools, open water, pressure chambers, and performance claims remain blocked.
The future we are trying to earn
We still believe underwater life support can become more legible: limits easier to see, faults easier to diagnose, maintenance state easier to trace, and task burden easier to understand. But the route is sequential. Evidence first. Independent review next. Non-human fixtures and qualified laboratories after that. Certification, insurance, medical and ethics gates long before any human phase.
The most exciting progress is not a promise that the hard constraints disappeared. It is a system that makes those constraints visible early enough to change the design.
Boundary notice. BlueLoop is an OpenFirefly research program. There is currently no certified or human-usable product. This note is not diving guidance, a certification or safety claim, or a launch commitment. M0 includes no animal testing, human breathing, pool, open-water, or pressure-chamber testing, and no pre-orders are accepted.
Primary references
- NAVSEA General Specification for the Design, Construction, and Repair of Diving and Hyperbaric Equipment, 2023
- U.S. Navy Experimental Diving Unit, unmanned testing methods for underwater breathing apparatus
- Molecular Products, Sofnolime 797 technical data
- Regulation (EU) 2016/425 on personal protective equipment
- BSI, BS EN 14143:2013 scope and status