from Hacker News

Researchers Discover the Optimal Way to Optimize

by jnord on 10/13/25, 10:41 PM with 14 comments

• by treetalker on 10/16/25, 7:18 PM

  > While the efforts of Bach and Huiberts are of theoretical interest to colleagues in their field, the work has not yielded any immediate practical applications.

• by measurablefunc on 10/18/25, 5:24 AM

  There must be lots of theorems in optimization theory that can be improved w/ more intellectual effort. Unlike video generation if AI is applied to find better algorithms it will have a direct impact on the economy b/c almost every industrial process is using some kind of constraint optimization algorithm including the simplex algorithm & its variations. But it's not flashy & profitable so OpenAI will keep promising AGI by 2030 w/o showing any actual breakthroughs in real world applications.

• by fernly on 10/18/25, 12:55 AM

  Another nice quote,

  > The next logical step is to invent a way to scale linearly with the number of constraints. “That is the North Star for all this research,” she said. But it would require a completely new strategy. “We are not at risk of achieving this anytime soon.”

• by Animats on 10/18/25, 4:32 AM

  Neat. Progress on lower bounds. That's been a tough area for decades.

  There are a lot of problems like this. Traveling salesman, for example. Exponential in the worst case, but polynomial almost all the time.

  Does this indicate progress on P = NP?

• by akshayka on 10/18/25, 1:31 AM

  Anecdotally it seems like most software engineers have heard of linear programming, but very few have heard of convex programming [1], and fewer still can apply it. The fixation on LPs is kind of odd ...

  [1] https://github.com/cvxpy/cvxpy