Microgpt
(news.ycombinator.com)
1.
2.
Show HN: Per-instance TSP Solver with No Pre-training (1.66% gap on d1291)
(news.ycombinator.com)