Critical systems are increasingly software-defined. Software is faster to write than the hardware it replaced, and coding tools have raised that rate again. Writing it is not the constraint. Showing that it does what the system requires, with evidence a certification authority will accept, takes longer than writing it. That gap is widening.
When these workflows were conceived, software was small, static, and bounded. The F-4A Phantom II carried on the order of a thousand lines of onboard code, and around 8% of its functions were performed in software. Everything else was mechanism, hydraulics, and analogue circuitry. All of it could be inspected, bench-tested, and reasoned about physically.
Today, the F-35 carries roughly 8 million lines of code onboard with about 24 million more in the ground systems that keep it flying. The aircraft became a machine whose behaviour is almost entirely expressed in software.
The tools for writing that software improved by orders of magnitude. The way intent is captured, how it becomes an implementation, how that implementation is shown to be correct has not meaningfully changed. And the gap widens every year.
Kronus runs a single pass from source artefacts to an evidenced system. Every output is bound to the configuration that produced it.
Kronus reads the requirements, interface documents, and design data a programme already holds. No migration step, and no new modelling language to adopt first. Conflicting sources are reported rather than silently reconciled.
States, interfaces, contracts, and timing become explicit and machine-checkable before any implementation exists. The specification reports conflicting requirements, unreachable states, and timing with no satisfying schedule.
One specification produces the source, the tests, the documentation, and the trace links back to intent, in a single pass. Every output carries a configuration identity, so evidence cannot drift from the build that made it.
Generated software runs on an execution model Kronus designs and controls. Static scheduling, explicit ordering, no dynamic allocation, no cache. Timing becomes a property of the architecture, not a measurement taken at the end.
Kronus is organised around certification objectives rather than around code that gets certified afterwards. The evidence model is common to every standard.
Kronus is not, and does not replace, a certification authority.