ITER · July 2026
6 / 9
Six of nine tokamak sector modules were installed, placing two-thirds of the toroidal core in the pit. ITER remains an assembly-and-integration project, not a source of experimental plasma results yet.
A compact public record of three stages: where fusion stands in the world today; what the frozen ATLAS-Σ V6 campaign actually demonstrated in computation; and how the private V200–V208 line moved the architecture toward worst-subsystem robustness, material survivability, thermal routing and adversarial falsification.
The frontier is no longer a single record. The public program spans burning-plasma evidence, long-pulse confinement, D–T operating scenarios, heat exhaust, superconducting magnets, plasma-facing materials and machine assembly.
Six of nine tokamak sector modules were installed, placing two-thirds of the toroidal core in the pit. ITER remains an assembly-and-integration project, not a source of experimental plasma results yet.
JET released 69.26 MJ over six seconds in its final D–T campaign, while also developing ITER-relevant integrated scenarios and exhaust regimes.
WEST sustained a hydrogen plasma for more than 22 minutes with actively cooled tungsten plasma-facing components and real-time density/heating control.
EAST maintained steady-state high-confinement plasma operation for 1,066 seconds and has continued work on density limits, edge stability and core confinement.
In April 2025 NIF reported 8.6 MJ yield from 2.08 MJ delivered to the target, a target gain of 4.13. This is inertial confinement and must not be equated with tokamak gain metrics.
CERN is not an ITER-like fusion reactor program. Its fusion relevance is technology transfer and collaboration: superconductors, magnet technology and engineering synergies with EUROfusion.
The earlier ATLAS-Σ documents describe a real executed program built around a synthetic, clipped, zero-dimensional plasma surrogate. Its scientific value is therefore computational and methodological—not evidence of physical ignition, net-electric gain or hardware readiness.
Target: Q ≥ 60 at every simulated time step.
Target: Ln < 1 MW/m².
Target: B ≥ 6 T.
One deterministic candidate was fixed before the 25 holdout cases were opened.
Matched negative controls failed when the Σ operator was removed—evidence of causal dependence inside that implementation.
Algebraic closure; the dossier also reports 40/40 formulation checks and 12/12 computational gates.
The protocol, grid, seeds, scoring rule, holdout generator and negative controls were frozen before selection. Q, wall-load proxy, B and thermal clipping had to satisfy their limits simultaneously—not as separate post-hoc victories. The selector did not receive holdout outcomes. The operator was then removed in matched ablations, and a stuck-actuator negative control was required to fail.
The strongest wording from the later V6 dossier is deliberately narrow: the ablation establishes causal dependence within the implementation, not causal authority over a physical plasma.
The private V200–V208 sequence marks a change in architecture. The public claim is intentionally structural: Q, neutron-wall loading, magnetic confinement, temperature, density and input power cease to be treated as isolated optimization targets and become components of a coordinated reactor state governed by a worst-subsystem criterion.
Moves from a set of passing thresholds toward a coordinated state whose quality is limited by its weakest subsystem.
Extends the architecture from plasma-side proxies toward whether the material system can survive the selected state.
Heat becomes not only a limit to respect but a routing problem: where energy appears, where it travels and whether it can be evacuated.
The public version name records a restart rather than a cosmetic iteration—an important marker of revision in the development path.
The validation philosophy advances from fixed tests toward actively searching for the condition, perturbation or subsystem that can invalidate a candidate.
Q ≥ target AND Lₙ < ceiling AND B ≥ floor AND no thermal clipping
A state passes when every declared threshold passes.
reactor_state_quality ← worst-performing critical subsystem
This is a conceptual description of the published “worst-subsystem criterion”, not disclosure of the private formula.
| Layer | World public frontier | ATLAS-Σ V6 | V200–V208 public evolution |
|---|---|---|---|
| Evidence type | Physical / engineering Experiments, hardware, machine integration. | Executed synthetic computation Frozen protocol, holdout, ablation and negative controls. | Private artifacts + public fingerprints Architecture described publicly; implementation remains private. |
| Primary question | Can fusion-relevant plasma and subsystems work physically at required regimes? | Can one frozen candidate satisfy all declared synthetic bounds and survive unseen tests? | Can the coordinated reactor state remain acceptable when judged by its weakest critical subsystem and attacked for failure? |
| Plasma | Measured confinement, D–T operation, long pulse, edge/core behavior. | Synthetic 0D surrogate. | One component of a larger coordinated reactor state. |
| Wall / materials | Real tungsten surfaces, divertor heat exhaust, irradiation/material programs. | Wall-load proxy and no-clipping condition; no measured material response. | Explicit material survivability milestone at V205. |
| Thermal | Real heat exhaust and actively cooled components. | Thermal clipping boundary inside the surrogate. | Explicit thermal-routing line at V206/V207. |
| Failure philosophy | Operational faults, disruptions, component limits, experimental campaigns. | Holdout + Σ ablation + stuck-actuator rejection. | Adversarial falsification as an explicit generation. |
| What cannot be claimed | A synthetic ATLAS-Σ Q must not be ranked against experimental fusion gain as though the quantities represented the same physical evidence. Public hashes do not turn private computation into physical validation. | ||
Anyone can save this page—or download the plain-text/JSON registry above—and later compare a disclosed artifact against the corresponding digest. Matching bytes produce the same digest; any change produces a different one.
SHA25680c854d84c814843c7dad65b908c76ddd2fbf33efdb3be4b21671c7fd9a0c845SHA3-5121abeaa1b68164b82d1b843047fdef64535f2c010ec99f09c5a61f41b26da8c14a50678d5d4bc2e8824254d281aa42c7177ef3691bc2361c0ccff1e31b60a3898SHA2564ee95164586004985422e10053dffdc1307b64b3ec8adee32c899272d5d2eca6SHA3-512debdfaf0323ab3b93320f11a49cfb4dcd6780f620f393f619e7a937659feb1c1e7aba187f1d344944c9c8f90347594fc5103f975c0c9461f20cbf5e3515ad54fSHA2566b1f9eeb13cb5d5e0da9557677597fa801ad4fcf3297e366f4519e1ea4347189SHA3-5121545abdbf406847b16827228886bef10558570ba395fd392f17852b5a469bbea37ed86ab88120cde9c357e55e3e20bdf86937de310e198923436acc620e6216aSHA256f0f56f7762a49706e26671a553c50a6298872eaf2c42e79f4d78220233568073SHA3-512b155dfc64169e7f16828542128c3785d5c1b77c40a40c0087b261be01620d8e523731b175f22d941648c32b386cecfd95df10ba5c790239c6a235888280f4f04SHA256406f132abb442f4710f7a9c5b85bacd95d99d1ff158c5487fca24dbc31042bb5SHA3-51239aab70af447dc770faf151ea9a64ba0ff4e3295eaeff8b219fd9bc65cd7eff60b235d5a6a34578c484edfb6c2a92a2b03c546bae3b70cc7dd37a388a70611c7This statement describes the claimed architectural evolution without exposing the private mechanism. It should be read together with the evidence boundary on this page.
Primary or institutional sources were used for the public fusion landscape. Accessed for this evidence-pack draft in September 2026.