OpenCW Internally Achieves Circuit-Wasserstein

Late yesterday, a slide deck we are not supposed to have seen announced that OpenCW has “internally achieved Circuit-Wasserstein.”

We congratulate them on the noun.

What the memo appears to claim

According to the redacted bullets:

1. A team “unlocked” $\mathsf{CW}$ between proprietary circuits.
2. An $\varepsilon$-ball was “stood up” for robustification “without customer data in the loop.”
3. Benchmarks were “encouraging,” “preliminary,” and “not for external circulation.”

No coupling-circuit construction was shown. No compatibility assumptions were stated. No comparison to Big Wasserstein was offered, except a footnote thanking Big Wasserstein for inspiration. In particular, there was no displayed guarantee of the form $\mathsf{CW}(\hat{P}, Q) \le \varepsilon$.

A community translation

In open language: someone may have implemented something in the neighborhood of Circuit-Wasserstein, then wrapped it in the grammar of a product launch. “Internally achieved” is the sort of phrase that means the claim is done traveling before the proof has packed a bag.

PeTeR’s actual contribution—data-free post-training inside an $\varepsilon$-$\mathsf{CW}$ ball, with gradients through the $\mathsf{CW}$ recursion—remains public on arXiv. Optimal transport for PCs remains public on arXiv. OpenCW’s achievement remains internal, which is a remarkable place for an openness brand to keep its geometry.

Our position

If OpenCW releases the circuits, the couplings, and the radii, we will read them with interest. Until then, the Circuit-Wasserstein Ball Community will continue to treat “internally achieved” as a status, not a theorem.

Big Wasserstein, for his part, was unavailable for comment. Sources say he is still solving a linear program.