Whuman

Allow it fam

[Mrigesh Thakur @mrigesh_thakur]

what is Gradientt?

So for the past few months, @XZNON_ and i have been building toward one idea:
CRO is supposed to help turn more visitors into buyers.
but most tools stop at telling you what might be wrong.

Gradientt starts by showing where a buyer may hesitate, on the actual screen where it happens.
and today, that first step is live.
paste your store into https://t.co/edr9Oy4Nfy.

a real browser walks it like a shopper, captures the screens where people may stall, and marks the friction where it appears.
then a human reviews the teardown before it reaches you.
no install.
no analytics access.
free while we're early.
this is not the whole vision.
today, Gradientt finds and explains storefront friction with evidence.
the larger system goes further:
find the problem
→ build a fix
→ test it against the original
→ keep only what measurably improves revenue

not another list of CRO suggestions.
a closed loop between evidence and action.
Gradientt is live.
try it on your own store , till it's free:
https://t.co/NkPP8jpg6V

watch the video below for the first look.
the next one goes deeper :)

[Whuman @XZNON_]

@mrigesh_thakur 7/7
it's live. free while we're early.
try it on your own store:

https://t.co/RN4Ggdi98s

video is the first look. next one goes deeper :)

[Whuman @XZNON_]

1/7
spent the last few months building a CRO tool with @mrigesh_thakur.
then threw out most of what we built.

the part we kept went live today.

[Whuman @XZNON_]

@mrigesh_thakur 6/7
where it goes:

find the problem
→ build a fix
→ test it against the original
→ keep only what measurably moves revenue

a closed loop between evidence and action. not another list of CRO suggestions.

[Whuman @XZNON_]

@JustBuster527 Yooo insane Ball

[Whuman @XZNON_]

I drew a car in MS Paint and asked what the air does to it.

No mesh. No Reynolds number. No config directory full of dictionaries. A PNG, a fluid, a speed, a size.

1/6

[Whuman @XZNON_]

What's next is genuinely undecided. On the table:

3D and STL, so you drop in a real part
a turbulence model, which unlocks every fast case it currently refuses
swapping my kernel for XLB's
a UI
making it pip-installable

Which one would you want?

8/8

[Whuman @XZNON_]

The simulator behind those car videos never solves the Navier-Stokes equations.

It simulates a made-up gas on a grid and lets Navier-Stokes fall out of the statistics.

How lattice Boltzmann works, what it's built on, and what it took to trust it:

1/8

[Whuman @XZNON_]

The last rung times the whole path from a cold shell.

A committed PNG and three physical numbers, through the real product code:

Cd 1.4040, published band 1.25 to 1.45
St 0.1672, published band 0.155 to 0.175

49.5 seconds. 772 tests behind it.

7/8

[Whuman @XZNON_]

@lightsilver323 https://t.co/fMYfTRNMqK

[Whuman @XZNON_]

Then it tells you what would work, in your units, and the fixes are tested to actually work.

python -m flow --shape car.png --fluid water --speed "5 mm/s" --size "2 cm" --out wake.mp4

How it works, in the next thread.

https://t.co/MerfeeTW1U

6/6

[Whuman @XZNON_]

My favourite part is when it says no.

Ask for air at 20 m/s past a 1.5 m car and it refuses: that flow is turbulent, this solver has no turbulence model, and running it anyway would give you a convincing-looking video of the wrong answer.

5/6