Okay, quick one before I get on a call. We pulled the trigger on the pricing change this morning — per-seat is gone, everything moves to usage-based, billed on settled volume.
Per-seat pricing removed — the dip never came
Ledgerloop moved from per-seat to usage-based pricing and net revenue retention rose instead of falling.
Proof · 3- Pilot cohort41 accounts across two quarters. Net revenue retention moved from 104% to 118%.
- Support loadSeat-counting tickets fell from roughly 30 a week to near zero.
Never use the 118% figure without the cohort size next to it.
From a founder’s voice note to an approved post
Thinked captures what actually happened, keeps every line of every draft traceable to it, and gets an unambiguous yes before anything reaches your publishing queue.
Demo data, resettable·No account required
We killed per-seat pricing this morning. Everything moves to usage-based, billed on settled volume. We should have done it eleven months ago.
1 of 4The reason we waited was not customers. Customers had been asking for over a year. We waited because we did not trust our own forecasting model to survive being wrong.
2 of 4The client review link, running on the same database as the workspace.
Writers are not short of talent. They are short of context.
- 01
The work is spread across five tools
Source material sits in a voice note. The draft sits in a doc. The feedback sits in WhatsApp. Nobody can answer “what is this post based on?” without three searches.
- 02
Approvals drift and nobody is sure
The client replies in the wrong channel, or not at all. Two weeks later nobody can say whether a post was approved, changed, or quietly dropped.
- 03
Clients do not volunteer context
Launches, opinions, customer sentences and real numbers stay in the founder's head. The writer fills the gap with something generic, and the founder does not recognise themselves in it.
Context arrives the way
founders actually send it
- Voice, typed note, link, screenshot or file
- Transcribes as it records, so nothing is reconstructed later
- Public, not-yet or background-only on every item
Net revenue retention on the pilot cohort went from 104 to 118. That's forty-one accounts over two quarters. And the support tickets about seat counting basically stopped — we were getting around thirty a week on that alone.
A loose update becomes
attributable material
- One claim per packet, with its proof beside it
- Public material with no proof is refused
- The limits the client set travel with it forever
Per-seat pricing removed — the dip never came
Ledgerloop moved from per-seat to usage-based pricing and net revenue retention rose instead of falling.
Proof · 3- Pilot cohort41 accounts across two quarters. Net revenue retention moved from 104% to 118%.
- Support loadSeat-counting tickets fell from roughly 30 a week to near zero.
Never use the 118% figure without the cohort size next to it.
Every line has to point
at something they said
- Source chips open straight into the evidence
- Unsourced posts are flagged on the row and in the library
- Sending is disabled until every post is traced
The dip we braced for never arrived. Net revenue retention across the 41-account pilot cohort moved from 104% to 118% over two quarters.
S1The quiet part: support tickets about seat counting went from roughly thirty a week to almost none. We had been paying for that pricing model in a line item nobody had labelled.
S1The average system goes unmaintained within two quarters of handover.
No sourceOne post, four answers,
no login
- Opens on a phone with the post first and the proof one tap away
- Edits and context requests carry a reason
- Approval is an explicit state, never inferred
We killed per-seat pricing this morning. Everything moves to usage-based, billed on settled volume. We should have done it eleven months ago.
1 of 4The reason we waited was not customers. Customers had been asking for over a year. We waited because we did not trust our own forecasting model to survive being wrong.
2 of 4Then it leaves for Typefully,
with the approval attached
- Nothing reaches publishing without an explicit approval
- The reference comes back onto the draft
- The trail is the product, not a log file
- Handed to publishingNadia Bekker · 08:46
- Client approvedMarta Vlk · 08:44
- Sent for approvalNadia Bekker · 08:40
- Source packet builtNadia Bekker · 08:20
Thinked does not publish. It decides what is allowed to.
The scope is narrow on purpose. Everything outside it already has a good tool, and the handoff to Typefully is the boundary.
What it owns
- Client signal capture through a no-login link
- Attributable source packets with proof and limits
- Drafts that cannot be sent without a source
- Explicit, timestamped client decisions
What it deliberately is not
- A scheduler — Typefully already publishes well.
- A client portal — Two links. Nothing to log into.
- A generic AI writer — No source, no draft.
- Project management — No tickets, boards or billing.
Follow one voice note all the way to a handoff
Nothing here is a video or a mock. It is the product, on demo data you can reset — the same state machine that decides whether a draft can be sent, whether an approval counts, and whether anything is allowed to leave for publishing.
In design partnership with Pairaw and a second founder-content agency · X and LinkedIn first