← The library
Methods note · LGTM

LGTM: four numbers for how work with AI is going

Landing, the Gap, Time-to-receipt, Momentum — what each one answers, what it reads, and how it is computed from the session record. Every metric ships with its pseudo-code.

July 2026 · Published by let.ai research · CC BY 4.0

Frameworks spread when their numbers do. HX names four — LGTM — each computable from the session record, and each published with its computation and its failure modes, because a number without its method is an invitation to misuse.

L — Landing: did the work land?

The share of build-intent sessions whose work reached a merged or published outcome within the follow-up window. It connects the session record to the receipts — commits, pull requests, published documents — so effort and outcome stay in one number. Full computation and pseudo-code: hxframework.org/metrics/landing.

G — the Gap: does the record agree with what people say?

The difference between what the record implies about how work is going (a normalized structural reading: Landing, friction, context cost, reuse) and what people report on the sixteen-statement felt instrument. The two should roughly agree; divergence is the alarm. Computed for groups of five or more, always published with both components. Details: hxframework.org/metrics/gap.

T — Time-to-receipt: does listening come back?

The median time from a problem raised through the tools to a visible outcome — a change shipped, a decision recorded, or a clear no with the reason — published beside the receipt rate, its honesty check. It measures the organization’s behaviour. Details: hxframework.org/metrics/time-to-receipt.

M — Momentum: are we getting better?

The slope of a simple composite — Landing, inverse friction, the felt score — across six trailing windows, always published with the series itself. It is the number the maturity ladder’s top level asks about. Details: hxframework.org/metrics/momentum.

One warning that covers all four

Never reduce a person to a number: every LGTM reading describes a team or an organization. No individual-level metric exists anywhere in HX — in the product, in the papers, or in a screenshot.