Sanguis 3.0 · Iteration:Determinism
VolkanaDEV is building AI that remembers everything it learns, and gives the same answer every time. It's written in Sanguis, the language we built for it.
Be patient while we tinker. More info and sign-ups soon.
Take the same program to a second computer and you can get a second answer. Almost all software has this problem, and it's unsolved across the whole industry. Most tools settle for looking the same.
We mean answering the same. The same program gives back the same answer on the graphics card, on the software, and on the compiler that builds itself. Every number is a real run you can check.
BloodCode: 29 instructions. Fixed numbers, fixed spellings, fixed meanings.
Assigned slots never change. Deprecated ops keep their number. Changing a slot's spelling fails the build.
A new executor (another GPU, an FPGA, a microcontroller) is admitted by passing one public per-operation test suite. Adding hardware never splits the language.
Artifacts are text. Open one in an editor and read exactly what your program became.
Write it out. Pick it back up somewhere else. It comes back bit-identical.
A core writes itself to a small package: its wiring, its weights, its clock. Insert it into a fresh process and it resumes where it was, the same program.
Because the settle is deterministic, any behaviour re-runs bit-for-bit years later, on hardware that didn't exist when it was written.
Every result has a computed cause. Debugging stops being an anecdote and becomes a measurement.
Prove a behavior on one machine and the proof stands on all of them, so a single audit covers every deployment.
What passed your tests last year passes them next year, on machines that didn't exist yet.
The program is the file. Same sentient instance, desk to bench to rack, nothing lost in transit.
Autonomous weapons, mass surveillance, facial recognition without consent, tools built to exploit people. All forbidden. Break the rule and your license dies on the spot.
Open and inspectable for individuals, students, and researchers, and commercial deployment is a conversation.