// TECHNOLOGY / 5-DST

Time is a place,
not a line.

3 dimensions of time + 2 of space = 5

Anomly exists to compute in five-dimensional space-time. No existing arithmetic could do it: a 5-DST computation is a path through time, and it needs numbers that give the same answer in any order, on any machine. So we built them. The exact arithmetic and the receipts INVAR ships are the first pieces of that machine.

Drag to orbit · scroll to zoom · Space to pause · R to replay Open full screen ▸

What you just watched

  1. 0 a point · a single moment
  2. 1 a line · time, the one dimension you already know (t₁)
  3. 2 a sheet · time gains a second direction (t₂)
  4. 3 a block · a third: time is a place, not a line (t₃)
  5. 4 a ball · space emerges from time, first as distance (ρ)
  6. 5 the globe · then as direction (θ): one world

In the usual picture, space is fundamental and time is a single line running through it. 5-DST (five-dimensional space-time, Kletetschka 2025) turns that around: time is primary and has three dimensions, and the two dimensions of space, distance and direction, emerge from it. Each point in the animation is a value of a multi-temporal field on that manifold. The picture is computed, not drawn.

Why computing 5-DST needed new arithmetic

A 5-DST computation is a worldline: a path of events through three dimensions of time, with space carried along as distance and direction. To compute one, two things have to hold. Independent events combined in any order must arrive at the same destination, and any event must be replayable, same inputs, same bits, on any machine. Floating point fails both. It rounds at every step, so the answer depends on the chip, the thread count and the batch. That made 5-DST uncomputable, as a matter of engineering rather than physics.

So we built the arithmetic first. Every product is formed exactly and accumulated in a 256-bit register, the quire, with one rounding at the end: Kulisch's exact accumulation on Gustafson's posit operands. Order cannot change the result. The same datapath runs bit for bit on x86, NVIDIA GPUs, ARM boards, Tenstorrent, a Xilinx FPGA and a fabricated die. And because every step is replayable, a computation can carry a receipt: INVAR records an inference as a worldline of hash-chained receipts that anyone can re-execute.

Those are the first products of the program, and they stand on their own: deterministic AI, verifiable inference, one result on every substrate. The program is 5-DST computing itself: the framework where time is the primary coordinate, the quire hardware built to run it, and the physics (Kletetschka, 2025) as its open frontier. See the technology overview for the measurements and INVAR to run it today.

References

  1. U. Kulisch, Computer Arithmetic and Validity: Theory, Implementation, and Applications, 2nd ed., De Gruyter, 2013.
  2. J. R. Shewchuk, "Adaptive Precision Floating-Point Arithmetic and Fast Robust Geometric Predicates," Discrete & Computational Geometry 18, 1997.
  3. J. L. Gustafson and I. Yonemoto, "Beating Floating Point at its Own Game: Posit Arithmetic," Supercomputing Frontiers and Innovations 4(2), 2017.
  4. G. Kletetschka, "Three-Dimensional Time: A Mathematical Framework for Fundamental Physics," Reports in Advances of Physical Sciences, 2025.