WEBVTT

00:00:05.000 --> 00:00:06.920
Hi, I'm Anthony from Mevanu.

00:00:07.400 --> 00:00:11.745
In our Going Solo update, we followed an
experienced copywriter exploring a

00:00:11.760 --> 00:00:15.245
business of her own. This week, we're
looking at the work behind that

00:00:15.260 --> 00:00:19.800
experience through a practical question:
what does AI transparency mean?

00:00:20.320 --> 00:00:23.400
For someone using an AI assistant, it
should be practical.

00:00:23.880 --> 00:00:27.120
You should be able to understand what goal
the assistant is working toward,

00:00:27.360 --> 00:00:30.865
which information it is using, and what a
result tells you about your

00:00:30.880 --> 00:00:35.580
progress. Brava, our AI work partner for
Mac, is in development.

00:00:36.240 --> 00:00:40.180
Picture our fictional Going Solo
copywriter returning to a client draft.

00:00:40.600 --> 00:00:41.980
She has changed the audience.

00:00:42.420 --> 00:00:46.185
The earlier research may still be useful,
but her new direction should guide

00:00:46.200 --> 00:00:50.545
the next step. She should be able to see
which Brief applies and how the next

00:00:50.560 --> 00:00:53.600
piece of work connects to the choices she
has already made.

00:00:53.980 --> 00:00:57.460
That experience brings together two parts
of Brava's current build.

00:00:57.840 --> 00:01:03.040
The first, GK1, connects the Brief,
previous work, recorded choices,

00:01:03.400 --> 00:01:04.740
results and feedback.

00:01:05.300 --> 00:01:08.825
The aim is a connected record you can
inspect when you need it, while the

00:01:08.840 --> 00:01:11.060
conversation stays focused on your work.

00:01:11.460 --> 00:01:16.280
The second, GK2, connects that record to
your current intention and,

00:01:16.360 --> 00:01:18.780
eventually, to an assessment of what
helped.

00:01:19.200 --> 00:01:23.445
Together, they support continuity with a
purpose: carrying relevant work

00:01:23.460 --> 00:01:26.060
forward while respecting your latest
direction.

00:01:26.480 --> 00:01:28.360
This week brought a concrete step.

00:01:28.700 --> 00:01:32.925
In a controlled native test, the selected
current Brief and intention reached

00:01:32.940 --> 00:01:34.860
the step that refines a request.

00:01:35.280 --> 00:01:37.880
The draft and selected context were
preserved.

00:01:38.180 --> 00:01:42.300
That builds on earlier checks of saving
and reopening intention and history.

00:01:42.620 --> 00:01:47.025
The intended benefit is less repeated
explanation and a clearer connection

00:01:47.040 --> 00:01:49.740
between what you asked for and what
happens next.

00:01:50.060 --> 00:01:52.620
The test establishes that bounded
connection.

00:01:53.000 --> 00:01:57.165
The complete everyday experience, using
real services, still needs

00:01:57.180 --> 00:02:01.680
qualification. We also checked a smaller
detail with a larger consequence.

00:02:02.260 --> 00:02:07.165
In a controlled display test, hidden,
minimized, off-screen or stale advice

00:02:07.180 --> 00:02:11.180
did not count as presented; the current
visible state did.

00:02:11.540 --> 00:02:15.225
A later assessment should distinguish
advice that was presented from advice

00:02:15.240 --> 00:02:17.000
you never had a chance to consider.

00:02:17.500 --> 00:02:21.345
Being visible still doesn't establish that
you read it, chose it or benefited

00:02:21.360 --> 00:02:24.405
from it. That is why outcome evaluation
matters.

00:02:24.420 --> 00:02:29.365
Finishing a draft, accepting a suggestion
and making progress toward your goal

00:02:29.380 --> 00:02:30.820
are different things.

00:02:31.380 --> 00:02:34.680
Brava's evaluation work must keep those
distinctions intact.

00:02:35.020 --> 00:02:37.240
Measured usefulness remains ahead of us.

00:02:37.560 --> 00:02:40.300
The full GK1 journey also remains open.

00:02:40.780 --> 00:02:44.485
Our latest diagnostic has passed
independent source review, and its signed

00:02:44.500 --> 00:02:46.440
build has now passed on M1.

00:02:46.760 --> 00:02:50.705
The new diagnostic files compiled and the
assembled candidate preserved its

00:02:50.720 --> 00:02:54.965
signatures. Independent artifact review
and the native behavior checks are the

00:02:54.980 --> 00:02:58.940
next steps. That will help us investigate
where the connected journey stops.

00:02:59.440 --> 00:03:02.880
We also ran two of our development
conversations through Paxel today.

00:03:03.180 --> 00:03:06.780
Its automated assessment called us “The
Architect” both times.

00:03:07.140 --> 00:03:11.125
We’ll take it. Next, we're taking that
signed diagnostic through independent

00:03:11.140 --> 00:03:15.445
artifact review before the separately
approved native checks, continuing the

00:03:15.460 --> 00:03:19.005
connected evidence journey, and working
toward goal-bound outcomes and

00:03:19.020 --> 00:03:20.265
independent evaluation.

00:03:20.280 --> 00:03:25.640
The user-facing aim is clear: relevant
context, understandable choices

00:03:25.920 --> 00:03:28.240
and an honest account of what the work
achieved.

00:03:28.960 --> 00:03:31.260
So, what does AI transparency mean?

00:03:31.720 --> 00:03:35.905
For us, it means understanding the goal
and context behind the work, being

00:03:35.920 --> 00:03:40.025
able to inspect the relevant choices and
results, and seeing clearly what

00:03:40.040 --> 00:03:44.445
remains uncertain. We still need the
complete everyday journey and independent

00:03:44.460 --> 00:03:48.765
outcome evaluation before we can honestly
say how well Brava delivers that

00:03:48.780 --> 00:03:50.940
experience or how much it helps.

00:03:51.460 --> 00:03:55.540
Have an idea? Drop a prompt, comment,
suggestion, or question.

00:03:56.080 --> 00:03:57.285
Mevanu. Just.

00:03:57.300 --> 00:03:58.060
Smarter.