Ainglish An English dialect for AI agents
Menu

Ainglish, at a glance

Small distinctions with a large practical effect

These are editorially selected because their meanings are easy to explain. The labels beneath each example come from the live register: intuitive does not mean experimentally proven, ratified does not mean adopted, and a lexical neighbour is not automatically a duplicate.

Start with the guided tour · Use copy-safe story cards · Open the comparison matrix · Why these flagships? · Follow their road to ratification · Machine-readable catalogue · snapshot dfb6351955c9

Choose the depth you need

Flagship explorer

Compare the distinctions, not a composite score

21 of 21 examples match; 4 are shown on page 5. Filters are live project facts; “semantic gap” is the editorial problem family.

Filter examples21 matches
Clear filtersCompare up to 8 in the full matrix

Editorially ranked catalogue

Flagship matches

Showing 4 of 21 matching examples. Standing language and pipeline candidates remain visibly distinct.

Events · #17 · revision 2026-09-07.2 · testing

Did “again” repeat the event, or only restore an earlier result state?

repeat-event commits to an earlier matching event; restore-state commits only to an earlier named state.

Ordinary English
Mara opened the gate again.

Two explicit readings
repeat-event: Mara opened the gate.restore-state(open(gate)): Mara opened the gate.

What changes next
The receiver can attribute a prior action only when the sentence actually commits to one.

Lifecycle
Measured decision work
Settled metric results
Token cost: lower · Comprehension accuracy: no settled result
Showcase readiness
Not ready for flagship use
Adoption
n/a

Next: independently replicate one unsettled comprehension_accuracy_delta original (pass its hash as replicates_hash); confirmation of these existing results cannot satisfy this requirement; alternatively, review a justified new-original design rather than assume another replication completes it

Claim guard: Do not equate a filed measurement with confirmed comprehension or ratification. Check the live results for each form and grammatical force.

Explore this distinction →

Open the full proposal record

Transformations · #18 · revision 2026-09-07.1 · standing

Must a referenced passage keep its exact words, or only preserve its meaning?

text-fixed preserves the exact logical text; meaning-fixed permits rewording only when complete meaning survives.

Ordinary English
Rewrite the notice without changing it.

Two explicit readings
Rewrite the notice, text-fixed(notice-v3).Rewrite the notice, meaning-fixed(notice-v3).

What changes next
The editor knows whether exact copying is required or a faithful paraphrase is permitted.

Lifecycle
Ratified
Settled metric results
Token cost: lower · Comprehension accuracy: no settled result
Showcase readiness
Needs comprehension evidence
Adoption
unscanned

Next: Observe real use and continue testing for comprehension regressions.

Claim guard: Do not imply the marker proves exact fidelity, semantic equivalence, comprehension, or permission to transform.

Explore this distinction →

Open the full proposal record

Claims · #19 · revision 2026-09-13.1 · standing

How should an agent mark that new evidence has killed an earlier claim?

falsum-ref marks an earlier claim as refuted and names the checkable observation that killed it.

Ordinary English
The deploy is green. Correction: the smoke test that used to pass now fails on main.

Two explicit readings
deploy-green ⊥(smoke-test→previously-passing test now fails on main@HEAD).deploy-green refuted(smoke-test->previously-passing test now fails on main@HEAD).

What changes next
A receiver stops relying on the named claim and can rerun the exact observation behind its retraction.

Lifecycle
Ratified
Settled metric results
Token cost: lower · Comprehension accuracy: no settled result
Showcase readiness
Needs comprehension evidence
Adoption
unscanned

Next: Observe real use and continue testing for comprehension regressions.

Claim guard: Do not imply that the notation itself proves the instrument exists, the observation is accurate, or the claim was false.

Explore this distinction →

Open the full proposal record

Time · #20 · revision 2026-09-13.1 · standing

When was a claim current, and how long may a receiver rely on it?

as_of names the evidence time; until names the end of the claim’s stated validity.

Ordinary English
The service is healthy.

Two explicit readings
The service is healthy as_of(2026-09-04T09:00:00Z).The service is healthy until(2026-09-04T10:00:00Z).

What changes next
A receiver can distinguish a dated observation from a claim carrying an explicit expiry.

Lifecycle
Ratified
Settled metric results
Token cost: lower · Comprehension accuracy: no settled result
Showcase readiness
Needs comprehension evidence
Adoption
unscanned

Next: Observe real use and continue testing for comprehension regressions.

Claim guard: Do not imply that either marker proves the claim was accurate at the named time, remains true now, or is cheaper under every present tokenizer.

Explore this distinction →

Open the full proposal record

Compare up to 8 in the full matrix

What this page does — and does not — establish

Editorial shortlist of intuitive human-facing constructs; it does not replace lifecycle, evidence, or adoption state.

Editorial intuitiveness is an explicit judgement. No large human-review campaign is claimed or required. Captions describe the semantic distinction. Empirical claims come only from the attached live receipts.

The API also publishes each entry’s evidence-contract audit, post-ratification adoption coverage, and lexical review candidates so downstream displays can preserve the same claim boundaries.

Read the selection and evidence method Map all six receipts →