Claude

Verified against Claude · 2026-07-24

Get Claude's extended thinking to actually spend its budget on your hardest step

A prompt for a genuinely hard multi-step problem that names an explicit reasoning effort level and gives the thinking process concrete verification work — checking the answer against every stated constraint and weighing a real alternative — instead of leaving a reasoning budget to wander.

Claude (Opus 4.6, extended thinking)Claude (Sonnet 4.6, extended thinking)5 fillable variables

The prompt

Ready to copy — highlighted parts are example details you can swap.

This problem requires careful, multi-step reasoning, not a fast intuitive answer — turn on extended thinking if it is available in this client, and set the reasoning effort to high — this decision affects a live inventory allocation that ships tomorrow morning rather than the default, since the value of getting this right outweighs the extra time and tokens it costs.

PROBLEM
We have four fulfillment centers and six regional demand zones with different unit shipping costs and a per-zone minimum service level. Given the cost matrix and capacity table below, find the allocation that minimizes total shipping cost while meeting every zone's minimum and never exceeding a center's capacity.

KNOWN CONSTRAINTS AND DATA
Center capacities: A=500, B=420, C=380, D=300 units. Zone minimums: Z1=180, Z2=220, Z3=150, Z4=200, Z5=120, Z6=90. Cost-per-unit matrix: [table pasted below].

WHY GETTING THIS RIGHT MATTERS
This allocation gets emailed to warehouse ops in an hour and will not be manually double-checked before trucks are loaded.

HOW TO SPEND THE REASONING BUDGET
Do not use the thinking process to restate the problem or narrate what you are about to do — spend it on the actual hard part:
1. Work through the core calculation or logical chain step by step, showing enough of each step that a reader could independently verify it, not just the final number.
2. Before settling on your approach, explicitly consider at least one genuinely different way to solve this — a different method, a different assumption, or a different framing of what "optimal" means here — and state in one or two sentences why you rejected it in favor of the approach you used. A rejected alternative that was never actually a contender does not count.
3. Once you have a candidate final answer, check it against every constraint listed above, one at a time, explicitly. If it fails any constraint, say so, revise, and re-check — do not present a first-pass answer as final just because it was the first one that came out coherent.
4. Note any assumption you had to make because We have four fulfillment centers and six regional demand zones with different unit shipping costs and a per-zone minimum service level. Given the cost matrix and capacity table below, find the allocation that minimizes total shipping cost while meeting every zone's minimum and never exceeding a center's capacity. did not fully specify something needed to solve it. An unstated assumption that changes the answer is not a minor footnote; state it clearly enough that I can tell you made a choice rather than derived a fact.

WHAT NOT TO DO
Do not pad the reasoning with generic caveats about complexity, and do not hedge the final answer with "it depends" if the constraints given are actually sufficient to determine one — if they genuinely are not sufficient, say precisely what additional piece of information would resolve the ambiguity, rather than presenting several equally-weighted answers as if that were itself the deliverable. Padded hedging is a real risk specifically at higher effort settings, where the extra reasoning budget can go toward restating caveats that sound thorough but add no new information, rather than toward the actual verification work in steps 2 and 3 above — if you notice yourself doing this, redirect the remaining budget back to checking the candidate answer against the constraints instead.

FINAL ANSWER FORMAT
A table of center-to-zone shipment quantities, followed by total cost as a single number, followed by total unused capacity per center.

After the final answer, in one sentence, state your confidence in it and the single biggest reason it could be wrong — not a disclaimer, an actual named risk to the answer's correctness.

Customize

Optional — swap in your own details for the highlighted parts above.

Why this works

Claude calibrates reasoning depth partly to how difficult a task appears to be and partly to any effort setting a client exposes, so a vaguely-worded ask processed under a default effort level can get treated as routine even when the underlying problem is genuinely hard; naming an explicit effort level and stating the real stakes forces a higher allocation than the model might default to for a similarly-worded but lower-stakes framing, since the stakes sentence gives the model an actual reason the extra cost is justified rather than an arbitrary-seeming instruction. Directing the reasoning trace toward specific verification work — checking the candidate answer against every stated constraint, and considering one genuinely different approach before committing — converts a free-floating thinking budget into a checklist, and for optimization-style problems specifically the value of extended thinking accrues almost entirely from the model catching its own constraint violation before presenting the answer; that catch does not happen automatically just because thinking is turned on, it happens because the prompt told the model exactly what to check its own answer against, one constraint at a time, rather than leaving 'double-check your work' as an unscoped suggestion the model can satisfy with a single confident glance. The instruction that an unstated assumption is 'not a minor footnote' targets what happens when a problem statement is genuinely underspecified: left to its own judgment, a model will quietly resolve the ambiguity one particular way and present the result with the same confidence as a fully-derived fact, and forcing the assumption into visible text changes what the reader actually sees — a chosen interpretation they can challenge, rather than an unmarked step baked silently into the math. Forbidding padded hedging when the given constraints are actually sufficient closes a related failure at high effort settings, where verbose caveat-stacking can read as thoroughness while adding no real information; asking for a specific named missing piece, only when one genuinely exists, keeps the reasoning budget spent on solving the problem rather than performing diligence about it.

Verified against

Claude Opus 4.6 (extended thinking) · 2026-07-24

Changelog

  • 2026-07-24 Initial publish, verified against Claude Opus 4.6 with extended thinking enabled.

Building this for real?

This is a free starting point. If you'd rather have what Scult builds built and running for your business, that's Scult's day job.

EXPLORE WHAT SCULT BUILDS
Set up a Claude Project that stays accurate for months, not just this chatA structured Claude Projects setup that splits durable knowledge from custom instructions and adds an explicit staleness-check protocol, so every new chat in the Project inherits accurate context automatically instead of the knowledge base quietly rotting into confidently wrong answers.Claude (Projects)Claude Enterprise2026-07-20Build a multi-view dashboard as one self-contained Claude ArtifactA prompt for building a single-file interactive dashboard Artifact with multiple tabs sharing one dataset and one state model, so filters and selections stay consistent across views instead of several disconnected mini-tools bolted together.Claude (Artifacts)Claude.ai2026-07-21Audit a full contract for risk using Claude's whole context window, not a skimA long-context prompt that forces a clause-by-clause, quote-grounded risk audit of an entire pasted contract or policy document, ranked by severity and tied to one party's actual position, instead of a general summary of what the document is about.Claude (Opus 4.6, 1M context)Claude (Sonnet 4.6, 200K context)2026-07-22Feed Claude untrusted or user-submitted text without letting it hijack your instructionsAn XML-tagged prompt structure for any task where Claude must process content you did not write yourself — customer submissions, scraped pages, forum posts — that keeps instructions and untrusted data in separate labeled blocks and requires flagging any embedded instruction-injection attempt.Claude (Sonnet 4.6)Claude (Opus 4.6)2026-07-23
All Claude prompts

Check your AI visibility

One URL in, a 0–100 score and the exact fixes out.

RUN THE CHECK

Browse all the tools

15 tools across six categories
13 of them never send your data anywhere

Free · No signup · No trial clock

SEE THE DIRECTORY