AIBRIEF
A Microsoft researcher, working with two AIs, says he produced a fast algorithm that reaches the 2 log N signal to noise threshold for MIMO (multi antenna) wireless decoding, where N is the number of antennas. Peer review has not started.
For 25 years, MIMO detection, the math every 5G, Wi-Fi, and satellite receiver runs every microsecond, has had a precise open target: a fast algorithm that decodes at the 2 log N signal-to-noise threshold, where N is the number of antennas. No one had proved such an algorithm could exist. Dimitris Papailiopoulos now says he and two AIs spent seven days producing one.
The result, posted to X and reported by Chinese tech outlet Qbitai, is a two-step algorithm: an initial estimate, then a greedy bit-by-bit correction. The paper claims it runs in polynomial time and succeeds at the 2 log N threshold. It also claims 2 log N is the floor: even the brute-force maximum-likelihood decoder fails below it.
The closest prior result was a 2020 box-relaxation bound at 4 log N, twice the suspected line. A 2001 polynomial-time claim was overturned in 2005.
The new claim has not been peer-reviewed. Papailiopoulos says he did not use the Lean proof checker and only line-checked the work by hand. The model names in circulation, "GPT-5.6" and "Claude Fable 5", do not match any public model family. English-language community commentary is already asking whether anyone outside the collaboration can reproduce the result.