Docs

Bigger problems, tighter deadlines

Measurement-efficiency is not just a cheaper run. The same budget or deadline can buy a bigger problem — or more problems.

The headroom compounds

Reaching target with fewer measurements means that at a fixed budget you have headroom left over. Spend it on scale: a larger n, a harder instance, or another run. The measured shots-to-target gap widens as problems grow, so the headroom compounds — where a conventional optimizer runs out of budget or collapses under drift on a large instance, the runtime can still finish.

Three ways to cash it in

  • Run larger. Reallocate the saved budget toward more parameters or a bigger instance in the same window.
  • Run more. Fit additional instances into a fixed deadline — higher throughput per shift, per quote, per machine-hour.
  • Run where others can’t. Under drift or one-measurement-per-round, the alternatives stop producing usable answers entirely; finishing at all is the win.

Batch related work

If your problems come as a stream of related instances, keep one session open and carry the prior configuration into the next instead of cold-starting each one. Warm-started sequential regimes are where the runtime is strongest — throughput climbs across the batch. See the warm-start pattern for the loop, and the deadline demo to watch the collapse regime live.

Honest framing: these gains concentrate in the winning regimes (tight budget, drift, latency). At ample budget on a static problem the headroom shrinks toward zero — the envelope will tell you.
← The envelopeRelated-task warm-starting →