Skip to content

[Anchor Proposal]: Fermi Estimation #608

Description

@raifdmueller

Proposed Term

Fermi Estimation (also: Fermi Problem, Fermi-Schätzung, order-of-magnitude estimation, back-of-the-envelope estimation)

Context

A method for estimating quantities that seem unknowable, attributed to Enrico Fermi: decompose the question into factors you can roughly estimate, estimate each to the nearest order of magnitude, and multiply. Overestimates and underestimates partially cancel in the product, so the result typically lands within a factor of 2–10 of the truth — enough for sanity checks, feasibility decisions, and capacity planning. Canonical example: "How many piano tuners are there in Chicago?"

Proposed category: problem-solving (alongside First Principles Thinking, Occam's Razor, Morphological Box). Fills a real gap: the catalog has no anchor for quantitative estimation yet.

Quality criteria check

  • Precise: a clearly bounded method (decompose → order-of-magnitude per factor → multiply → sanity-check); no ambiguity with other meanings
  • Rich: activates decomposition, powers-of-ten thinking, error cancellation, upper/lower bounding, plausibility-over-precision
  • Consistent: densely documented in EN and DE; the piano-tuner example reproduces reliably across models
  • Attributable: Enrico Fermi; physics teaching tradition; Weinstein & Adam, Guesstimation (Princeton UP, 2008); standard in tech-interview and capacity-estimation culture

LLM Activation Test Result

Asked "What concepts do you associate with 'Fermi estimation'?" — strong activation: decomposition into estimable factors, order-of-magnitude arithmetic, the Chicago piano-tuner problem, error compensation, back-of-the-envelope framing, use in interviews and feasibility checks. Recognition, accuracy, depth, and specificity all high.

Notes for implementation

Per the contribution convention from #603/#605: research documented criticism before the PR (literature exists on calibration/overconfidence in Fermi-style estimates — verify and link sources by fetching), EN + DE files, changelog entry, AgentSkill catalog update.

Proposed by Ralf D. Müller.

Pre-submission Checklist

  • I have searched existing anchors and this term is not already included
  • This term refers to a well-established concept (not something I invented)

🤖 Generated with Claude Code

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions