Roadmap
Implementation status
Entries in the first list have a committed run artifact. Entries in the third are open questions with no scheduled date.
Works today
- DEC incompressible solver on fixed meshesGhia cavity, cylinder Re 100, graded MMS at order 2.00.
- QTT compressible marchers, 1-D through 3-DSod to L1 ≤ 0.027 against the exact Riemann solution.
- Fitted closures and the reacting-air kernel suiteRAM-C II peak electron density: the uncalibrated finite-rate network lands within about a factor of two of the flight anchor (+0.35 dec).
- The CfdFlow study grammar with compile-time phase disciplinesweep, gates, record, verdict; four compile_fail doctests pin the ordering.
- O(1) copy-on-write forks of a paused marched stateMeasured fork setup at 42 ns; continuation cost 0.67–1.04× the trunk.
- Precision as a type parameterf32 / f64 / Float106 through the same physics, switched at one alias.
- Plasma retropulsion descent, blackout exit to touchdownSixteen gates in 355.6 s. The mid-burn fork of the marched plume-coupled state departs a frozen-drag prediction by 139.4 m/s, which is what forking state rather than parameters buys.
Building
- Re-pinning plus exact Rankine–Hugoniot interface for marched rankThe measured lever: radial flux with the front as a tracked interface holds bond 8 flat in resolution, where flux-through-front grows 25 → 35.
- Grid-convergence study for the cylinder16 → 24 → 32 cells/D, Richardson-extrapolated, plus C_L,rms, θ_sep and C_pb. Required before any DNS-grade accuracy claim.
Open
- Turbulence, staged on the DEC solverLES after the spectra observables land. The wake is currently reported and never gated.
- A tractable 3-D fitted marcherMeasured >3× over the ten-minute budget at 16³ with bond cap 16, which is why the shipped corridor runs a 2-D layer.
- Plume-resolved compressible SRP on tensor trainsRank behaviour of a colliding-shock system on a quantized tensor train is unmeasured. Open research, named as such.
- Off-axis SRP, thrust vectoring, 6-DOFExcluded by validity discipline: off-axis leaves the Cordell–Braun validated envelope, so a surprising result would be unattributable.
How this list is maintained
An item moves from open to building once a de-risk study exists, and from building to works today once a self-verifying target or a committed example output covers it. Four items were removed after measurement refuted them; they are recorded on the boundaries page.