2:37
Friday Update: What would count as learning?
What would prove an AI assistant learned from your work? Alice explains Brava’s controlled tests, the outcomes still unmeasured and the checks ahead.
Watch on YouTube · Join the conversation
Alice narration supplied and approved by Mevanu’s Founder, generated with ElevenLabs. Original explanatory graphics describe a dated 6 November 2026 development snapshot; they are not app footage or a completed customer job. The candidate remains unfinished and awaits independent review. Improved outcomes were not measured. Brava remains in development.
Read the transcript
Text follows the published English captions, including the original narration’s wording.
English captions (SRT)English captions (VTT)
Hi, I’m Alice from Mevanu. Last Friday, we asked what AI transparency should mean in everyday work. This week, we’re asking a more specific question: what would prove that an AI assistant learned from your work?
By learning here, we mean using what happened in one task to guide later work.
Imagine you tell an assistant, “Next time, check the source before you write.” Saving that sentence is only the beginning. The interesting part comes with the next assignment. Does the assistant use it appropriately? Can you see what happened? And can you change your mind?
This week, our development work connected parts of that process.
First, references to supplied documents passed through reading and correction checks into a recorded result in a controlled local test. The intended benefit is practical: keeping track of the material behind an answer, including documents that don’t have a website address. The candidate is still unfinished and awaiting independent review.
Second, a candidate working method reached a later controlled task. Local checks also covered withdrawing it and rolling back its use. That matters because a useful habit for one situation can be the wrong habit for another. Your latest direction should still govern the work.
Here is the important qualification. The later test kept the original behavior. Improved outcomes were not measured. We have connected parts of a development process; we have not demonstrated the complete everyday experience or shown that it makes Brava more useful.
So the next test has to look at the whole job. Which information was used? Which method was allowed? What did the assistant do? What result did the person accept? How much setup, checking and correction did that take?
The next build milestone is to complete the document-handling path, finish the affected checks and put the frozen candidate through independent review. The everyday workflow and its outcomes still need testing.
What would count as learning? Evidence that an assistant used a relevant lesson in later work, that you could correct or withdraw it, and that the result helped. We can show progress on parts of that chain. We can’t honestly claim the full answer until the everyday journey and its outcomes have been tested.
Have an idea? Drop a prompt, comment, suggestion, or question.
Mevanu. Just. Smarter.
Want help deciding which feedback an AI assistant should carry forward, and how to check whether it helps? Talk with Mevanu.