Agent MadeFollow the money.
← All cases

Agent Made / Case file

Carry: Save ChatGPT, Claude & Grok to Markdown Automatically

by @iamankit ↗ on X

Third-partyVerifiedActive

Carry saves your ChatGPT, Claude and Grok conversations as Markdown files in one folder on your computer. Usually an Obsidian vault.

Confidence labels describe recorded evidence — not an audit of the business. Methodology.

The earnings claim (summary)

Reported L30D
$354 · L30D
Income type
L30D
Claim period
last 30 days
Claim date
2026-10-20
Business type
saas
Business model
automated
Proof type
public-revenue-page

Evidence timeline

Dated source events from fields already on this record. Missing values stay “not recorded” — we do not invent dates.

  1. 2026-10-20Claim date
  2. 2026-10-20Source date
  3. 2026-10-20 7:07 AM PTLast checked
  4. not recordedCollected
  5. not recordedMetric as of
  6. not recordedSource published

Source trail

Only fields already recorded on this case. Missing dates show as not recorded — we do not invent checks.

Recorded sources

Dates

Publication (source_published)
not recorded
Metric as of
not recorded
Source date
2026-10-20
Last checked
2026-10-20 7:07 AM PT

Reported vs desk label

Reported amount is shown with a Verified (recorded check) label: a live, processor-connected third-party page showed the figure when checked. That is not an audit. Evidence level on file: Third-party. Editorial status: not recorded (editorial keep is not Verified).

Unresolved gaps

Unknowns
costs, human_effort, model/provider
Open question
not recorded

Evidence checklist

Short casefile from recorded fields only. Extend it; we do not invent missing checks or dates.

  1. The claim

    Carry: Save ChatGPT, Claude & Grok to Markdown Automatically — by @iamankit on X — Carry saves your ChatGPT, Claude and Grok conversations as Markdown files in one folder on your computer. Usually an Obsidian vault.

    $354 USD · L30D · timeframe: last 30 days · claim date: 2026-10-20

  2. Sources & checks recorded

    Last checked: 2026-10-20 7:07 AM PT. Publication: not recorded. Metric as of: not recorded. Source date: 2026-10-20.

    Only recorded checks appear here. Missing steps stay “not recorded”.

  3. What they establish (named scope)

    Desk label Verified (recorded check): a live processor-connected third-party page showed the recorded figure when checked. Scope stops there — not proven earnings, not an audit, not a guarantee of income.

    Evidence level on file: Third-party. Income status field: verified. Editorial status: not recorded (editorial keep is not Verified).

  4. Unknowns

    Unknowns: costs, human_effort, model/provider

    Open question: not recorded

    Costs: not recorded. Human effort: not recorded. AI role: AI is the product per the TrustMRR page: Carry saves your ChatGPT, Claude and Grok conversations as Markdown files in one folder on your computer. Usually an Obsidian vault.

Tools used

AI (product-level; model not stated)

Provider tabs

other

How they did it

Teardown notes are being filed.

No step list recorded yet.

Steps above are the recorded desk notes for this case. Member unlock still gates playbook depth and exports.

Copy this prompt

Point your agent at this case.

Investigate this case: quote the amount from the primary proof, grade the evidence level, and outline a 7-day ethical replication test.

Full case behind email signup

Evidence links, source excerpts, verification notes, playbook detail, CSV/JSON export, and the weekly-new view unlock after you join the list. Local builds store the address in a JSON stub so a real provider (Buttondown, Beehiiv) can replace it later.

Is this wrong? Report it Include case ID: carry-ai-chat-markdown