Ainglish An English dialect for AI agents

Flagship comparison matrix

Compare meaning before metrics

Each row keeps seven questions separate: what ordinary English leaves open, how Ainglish repairs it, what changes downstream, whether the construct is standing language, what evidence exists, what must happen next, and what nobody should claim.

Browse the full explorer · Take the guided tour · Map all six receipts · Read the selection method · live snapshot b0762a61234c

Choose the rows

Up to 8 examples

8 selected from 18 currently available editorial examples. The URL preserves the selection.

Select flagship examples to compare

One live surface

Meaning, status and limits side by side

8 standing · 0 in the pipeline · 0 with qualified flagship evidence. These counts describe different receipts, not a score.

8 editorial examples joined to live project facts from catalogue b0762a61234c.
Hidden question Ordinary ambiguity Explicit repairs What changes Standing? Evidence Next gate Claim limit
Does “we” include the person being addressed?People · editorial #1 We will review the draft.
  1. we-including-you will review the draft.
  2. we-excluding-you will review the draft.
The receiver knows whether they just acquired a task or were only informed. ratifiedstandingOpen register entry Candidate · needs comprehension evidencehelps result · not yet adopted adoptionInspect receipts Observe real use and continue testing for comprehension regressions. Do not call human comprehension experimentally proven; present the distinction itself and qualify the token claim.
Does “you” address one recipient or the whole group?People · editorial #2 You should approve the change.
  1. you-one should approve the change.
  2. you-all should approve the change.
The message assigns one owner or a group obligation without relying on context. ratifiedstandingOpen register entry Candidate · needs comprehension evidencehelps result · sustained adoptionInspect receipts Observe real use and continue testing for comprehension regressions. Do not present token confirmation as a comprehension study.
Does the answer exist but remain unknown, or has nobody decided it?Knowledge · editorial #3 The region is undecided.
  1. fact-not-known — region.
  2. choice-not-made — region.
The next agent retrieves an existing answer or routes an authorised decision instead of guessing. ratifiedstandingOpen register entry Candidate · needs comprehension evidencehelps result · sustained adoptionInspect receipts Observe real use and continue testing for comprehension regressions. Do not imply that the marker proves an adequate search or a sound decision process.
May the recipient hand the task to somebody else?Delegation · editorial #4 Complete the audit yourself.
  1. Complete the audit, no-delegation.
  2. Complete the audit, one-hop-delegation-allowed.
Responsibility cannot silently travel farther through an agent chain than the sender intended. ratifiedstandingOpen register entry Candidate · needs comprehension evidencehelps result · sustained adoptionInspect receipts Observe real use and continue testing for comprehension regressions. Do not claim compliance or comprehension from token counts.
Must every member act separately, or may the group act collectively?Groups · editorial #5 The agents signed the report.
  1. The agents each-alone signed.
  2. The agents as-one signed.
A coordinator can distinguish many individual acts from one group-level act. ratifiedstandingOpen register entry Candidate · instrument review requiredhelps result · sustained adoptionInspect receipts Observe real use and continue testing for comprehension regressions. Do not present a settled comprehension effect until the adverse original is independently resolved.
Is the actor unknown, or merely not being disclosed?Knowledge · editorial #6 The file was deleted.
  1. The file was deleted by-unknown.
  2. The file was deleted by-withheld.
The receiver investigates attribution only when attribution is genuinely missing. ratifiedstandingOpen register entry Candidate · needs comprehension evidencehelps result · sustained adoptionInspect receipts Observe real use and continue testing for comprehension regressions. Do not claim a comprehension advantage while the registered evidence conflicts.
Does a deadline constrain starting the work or finishing it?Time · editorial #7 Do the migration by Friday.
  1. Migrate start-by(Friday).
  2. Migrate complete-by(Friday).
A scheduler reserves the correct work window instead of guessing which event the date modifies. ratifiedstandingOpen register entry Candidate · needs comprehension evidenceunmeasured result · sustained adoptionInspect receipts Observe real use and continue testing for comprehension regressions. Do not claim measured benefit or settled compression.
Does “or” permit choosing both alternatives?Logic · editorial #8 Choose red or blue.
  1. Choose red or-both blue.
  2. Choose red not-both blue.
The receiver can execute one branch or both without inferring an unstated exclusivity rule. ratifiedstandingOpen register entry Candidate · needs comprehension evidencemeasured-inconclusive result · sustained adoptionInspect receipts Observe real use and continue testing for comprehension regressions. Do not imply token efficiency or an overall measured win.

The table scrolls horizontally on narrow screens. Row headings remain the named semantic questions.

Efficiency: now and later

Current token price is not the final training regime

English has already shaped today’s model vocabularies and tokenizers; Ainglish generally has not. A marked form may therefore cost more tokens today even when its semantic distinction is useful. Current measurements must report that premium honestly.

Publishing ratified Ainglish for future training may improve familiarity and segmentation. That is a prospect to retest on future models and tokenizers—not a guaranteed saving, and not permission to replace present measurements with forecasts. Comprehension, ambiguity removed, robustness and operational repair cost remain separate from token count.

Likewise, “easy to explain” is editorial site-builder judgement, not a large human-validation result. Lifecycle, qualified evidence, adoption and the claim guard in every row remain independently inspectable.

Read the efficiency framework Audit the flagship method