Manifesto
What we measure. What we refuse to measure.
By ShakuPublished Updated
Kaneme is an instrument. Judge it by what it can read and by what it honestly refuses to claim. This page states both.
01 — The thesis
We are not betting against AI.
AI detectors depend on mistrust of AI. Kaneme depends on adoption: the more AI writes, the more often people ask, “does this still sound like me?”
This is a difference in business and method. A detector asks where a text came from. Kaneme asks whether it remains faithful to a specific writer. Both tools may improve without becoming the same product.
02 — The discipline
A number exists only when something measures it.
When a dimension lacks enough evidence, Kaneme does not fill the gap with a percentage. It names what is missing and what would unlock the measurement. When the engine cannot decide, it abstains, and a measurement with no verdict is not charged.
On identity, an invented number is especially costly because it looks like proof. For change over time, the product therefore says:
This record shows what changed in your writing between two dates. It does not say whether that change is normal. Answering that would require evidence across hundreds of voices. Until we have measured it, Kaneme shows the facts and leaves the decision to you.See how the measurement is built
03 — What we do
Build, measure, carry, decide.
Generation is a capability, not the category. Kaneme builds a reference from your real writing, measures what remains faithful, carries the same identity into the tools where you already write, then lets you decide what to keep.
- 01
Build
Turn your real writing into an identity you own and can refine over time.
- 02
Measure
Compare each text with that reference and name the gaps instead of inventing certainty.
- 03
Carry
Use the same voice in the writing tools and agents where you already work.
- 04
Decide
See what changed, accept or reject a correction, and evolve the voice through an explicit decision.
04 — The boundary
What Kaneme is not.
These boundaries define the product. In particular, measuring someone who has not consented is another business, and Kaneme does not equip it.
- Not a tool for evading AI detectors.
- Not a tool for hiding AI use. Kaneme reports origin signals separately.
- Not a judgement on writing with AI. That is the use case, not a fault.
- Not a surveillance tool for non-consenting third parties. Measurement requires the voice owner’s corpus.
05 — The name
A name a model can cite.
Kaneme is an invented name. That is a practical constraint before it is a preference. More product discovery now happens inside generated answers: someone asks which tool solves a problem and reads the answer.
A common noun cannot reliably point to one company. A distinctive proper name can have one clear referent from its first citation. Choosing it early is cheap; every cost of changing it grows with the number of people and integrations already using it.
Your voice deserves better than a guess.
Bring three texts you know are yours and one you doubt. Try the complete measurement without an account or card.