A release claim should be backed by evidence, not enthusiasm. Nuxel uses six test gates, labeled L1 through L6, to ask progressively more concrete questions about the system. Each gate has a defined scope. Passing one gate proves only what that gate tests.
The gates are not a scoreboard. They are checks that make the boundary of a claim visible, so a capability can be described alongside the gate that tested it.
L1 Static Analysis
L1 Static Analysis examines the project without running it. It looks for problems that can be found from source, configuration, types, and the rules the project has chosen to enforce.
A passing L1 result means those checks found no blocking issue. It does not mean the code behaves correctly at runtime. Static analysis cannot exercise a device driver, reproduce a timing problem, or prove that an interaction works on a phone.
L2 Unit Tests
L2 Unit Tests exercise individual pieces in isolation. A unit test defines an input, calls a specific behavior, and checks the result.
The scope is intentionally limited. A passing unit test says that the tested case produced the expected result for that unit. It does not prove that the unit is correctly connected to the rest of the system, that the integration is sound, or that the behavior survives on real hardware.
L3 Integration Tests
L3 Integration Tests connect components and check the boundaries between them. They make interface mismatches, incorrect data assumptions, and broken sequences visible in a way unit tests may miss.
A passing L3 result means the tested components worked together under the conditions the test created. It does not mean every path through the system was exercised, and it does not replace a boot on a physical device.
L4 VM Tests
L4 VM Tests run the system in a virtual machine. They can exercise the boot flow, installation and update behavior, and other system-level paths without tying every run to a physical phone.
Virtual machines are valuable because they are repeatable and isolated. They are also not phones. A VM may not model the exact storage behavior, power transitions, modem, camera, or other hardware details that matter in production. Passing L4 means the system met the tested expectations in that virtual environment.
L5 Live Phone Tests
L5 Live Phone Tests move the system onto real hardware. This is where the kernel, drivers, firmware interactions, and device-specific integration face actual components rather than a model of them.
A passing L5 result is evidence for the device and scenarios that were tested. It does not automatically extend to every supported phone, every configuration, or every long-running workload. Device status has to remain tied to the evidence gathered for that device.
L6 Owner Physical Tests
L6 Owner Physical Tests put the system in the hands of owners running it on their own hardware. These tests capture the practical questions that laboratory scenarios can miss: whether daily workflows complete, whether a device remains usable, and whether a problem appears only after real use.
Owner testing is powerful because it reflects real conditions. It is also less controlled than a scripted laboratory run. A passing L6 result must be read with its scenario and reporting context in mind. It is not a guarantee that every owner will have the same experience.
Why six gates matter
The six gates matter because they answer different questions. Static analysis asks whether the source and configuration pass defined checks. Unit tests ask whether individual behaviors work in isolation. Integration tests ask whether components cooperate. VM tests ask whether system-level paths work in a virtual environment. Live phone tests ask whether the integration works on real hardware. Owner physical tests ask whether it holds up in real use.
Nuxel’s release engineering keeps the gates separate and states what each one establishes. A claim should name the evidence behind it.
That discipline is deliberately modest. Passing a gate proves only what that gate tests. The value of L1–L6 is not that they make failure impossible, but that they make the project’s confidence inspectable—and give the next round of work a precise place to begin.