Vitark

a reasoning canvas · goal to next step

AI will give you a confident answer. Vitark gives you a defensible one.

The model does the reading. Everything it puts on the board carries the findings underneath it — each with a source and a reliability.

nothing runs until you ask it to
or try
the difference

Same question. Only one of these you can check.

a chat answer

Should I take the Church Street lease at £38k a year?

Yes — take it. Church Street has strong foot traffic, and £38k is a fair rent for the area. It’s a great opportunity.

sounds sure · shows nothing
the same question on Vitark
understanding
£38k sits inside the street’s going rate
supported by 2 findings · medium confidence
backed by
evidence
Council footfall count, 2025
1 source · high
evidence
Two agents’ comparable rents
2 sources · medium
still assumes
understanding
Footfall holds once the bypass opens
no findings · an assumption
every claim shows what it rests on
how it works

One sentence in. A board you can argue with.

  1. 01Say what you’re working outa sentence, or everything on your mind
  2. 02Answer what only you cana few questions — or skip them
  3. 03Run the questions you chooseeach comes back with its sources attached
Why does retention drop after week three?
goal
Why does retention drop after week three?
question
Which cohorts drop, and do they share an entry point?
question
Did anything change in the April release?
evidenceremove
Onboarding completion by cohort, Jan–Aug
2 sources · high reliability
evidenceremove
Three support threads mentioning price
1 source · low reliability
understanding
The week-three drop tracks onboarding depth, not plan
supported by 2 findings · high confidence
path
Rebuild the second-session empty state
backing
2
assumes
1
open
0
path
Add a guided checklist to week one
backing
1
assumes
1
open
1
swipe for the rest · drag any card · clicktap evidence to remove it
under every card

Open anything, and you get its argument.

Not a summary — the chain. What led to it, what backs it, and what it still assumes. Remove a source and everything resting on it is marked, rather than quietly left standing.

see a finished board
path3 links
Rebuild the second-session empty state
led to bythe week-three drop tracks onboarding
backed byonboarding completion by cohort
assumesthe April release didn’t change pricing
the rules

Three things the model isn’t allowed to do.

question
What would a cheaper plan change?
not run · waiting for you

Run a lane you didn’t press.

understanding
Annual plans churn less
no findings · an assumption

Claim something without attaching what backs it.

evidence
Cohort export
source removed
understanding
Drop tracks onboarding
high low confidence

Keep its confidence after you remove the source it rested on.

before you start

Fair questions.

Which model does the thinking?

Google Gemini reads and writes the cards, and Exa searches the web for the sources that become evidence. Neither gets to decide how sure the board is — that is worked out from what is attached.

Where does what I type go?

Your boards stay in a database on the machine running Vitark. Gemini and Exa see only what a run needs, and only when you press it — the privacy page says exactly what.

What does it cost?

Vitark itself is free. Runs use your own Gemini and Exa keys, so those services bill for what you use — and nothing runs unless you press it.

What if a source is wrong?

Remove it. Everything resting on it is re-weighed on the spot: a conclusion loses the confidence that source gave it, and one with nothing left under it is marked an assumption.

Does it go off and research on its own?

No. It breaks your goal into questions and waits. You choose which to run, and each comes back with its sources attached.

next step

or open your boards →