Prepared demo · No live AI · Not an assessment
How it works
Presencia Education rests on one habit: what a learner said, what a helper suggested and how we understand it now stay apart, each with its time, so anyone can check how understanding changed.
A learning moment is not one note. It is up to six records, each written by someone different, each with its own clock. None of them overwrites another.
The learner’s own report, in their own words, exactly as said.
Said · Day 1, 09:42
Written as an observation of what happened, never as a verdict on the person.
Observed · Day 1, 09:50
Kept with its label until the learner chooses to keep it. Then it is marked Kept Human–AI.
Made · Day 1, 09:55
What the learner did then, with every piece of help named on a receipt.
Took place · Day 1, 10:02
A later attempt, on another day or in another setting. This is the evidence that lasts.
Took place · Day 8, 10:15
A new reading sits beside the earlier one, dated and marked recorded later.
Recorded later · Day 8, 10:20
Every piece of help is written down, and two kinds of help are kept apart. Access support helps someone take part. Answer support does part of the thinking. Taking away answer help never takes away access support.
| Read aloud | Access support |
|---|---|
| More time | Access support |
| Hint: “count one table first” | Answer support |
| Model answer shown | No |
On Day 8 the attempt is checked again with the same access support and no answer support.
When understanding changes, the earlier record stays exactly as it was. The new reading sits beside it in a window of the same size, dated and marked.
“The helper’s $90 total looks right to me.”
“The $90 only covers six people. Twelve are coming, so materials are $60 and the total is $120.”
changedBefore a plan is trusted, four plain questions. A plan fits when its time, place, amount and order agree. They are questions to ask out loud, never a hidden score.
When does it happen, and is there time for it?
Where, and who is allowed to be there?
How much: people, money, materials, effort?
What has to come first, and what can wait?
These are short summaries of published studies, not results from Presencia. Presencia Education has not been measured yet.
Bastani and colleagues · PNAS, 2025
In a high-school mathematics field experiment, students with unrestricted access to a chat assistant solved 48% more practice problems, then scored 17% lower on a later exam taken without it. A version with teacher-designed safeguards largely removed that harm.
Limits. One country, one subject, a few weeks.
Wang and colleagues · Tutor CoPilot
Human tutors who received AI suggestions during live sessions saw their students master topics about 4 percentage points more often, with larger gains for less experienced tutors.
Limits. The AI supported the tutor; it did not replace them.
Kestin and colleagues · Scientific Reports, 2025
In a university physics course, a carefully designed AI tutor produced larger learning gains in less time than an in-class active-learning lesson.
Limits. One course, short-term measures, a tutor built with expert structure.
Support educators first. Ask for the learner’s own attempt before any help. Measure what a learner can do later, without help, and compare it with ordinary good teaching.
Learn to notice, check and act—letting experience become judgment while retaining the ability to question it.
This website and its prepared demos, in English and Spanish. No live AI, nothing about you kept.
Learning records with their own clocks, assistance receipts and the then-beside-now view, on Presencia’s shared kernel.
A small adult pilot by invitation, then family and school pilots, each after its own review.
