Write a bottleneck people can help with

A good bottleneck lets a researcher from another field understand what you’re computing, how big it is and where it breaks, without having read your papers.

The bottleneck form asks for most of this already:

• The question in plain terms. “Pick 30 of 12,000 candidate additives that maximize stability without breaking solubility limits.”
• The math, if you have it. Write it in LaTeX: $\max_x \; c^\top x$ subject to $Ax \le b$, $x \in \{0,1\}^n$.
• Size. Number of variables, atoms, assets, nodes or time steps. Size decides which methods are realistic.
• What you run today and how long it takes. “Gurobi, 32 cores, 41 hours to a 2% gap” is far more useful than “it’s slow.”
• What “better” means. Faster, cheaper, a better objective value, or simply feasible at all.

You never need to share proprietary data. Describe the structure of the problem and offer synthetic proxy data with the same shape and statistics. Mark the data setting honestly so others know what they can reproduce.

The last field asks what would change if the computation ran ten times faster. It helps others judge impact, and it’s what selectors look at when choosing problems for live sessions.