Codex to update a four-month cash forecast
Maikeru described a household money workflow that could be updated from a phone during a commute. Instead of waiting to enter every purchase at a desk, the author supplied screenshots of QUICPay transactions and photographs of utility bills to Codex through ChatGPT Work.
Original prompt
Shared by the source author. See source notes for context.
これも入れといて
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source author
- 二拠点生活の、まいける (@uneta)
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- An updated next-card-debit estimate and four-month bank-balance forecast; no account transfer is reported.
- Human involvement
- The author provided evidence and discussed outputs; a separate validation checkpoint is not described.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- Retain household financial context in Drive files and a spreadsheet.
- Supply transaction screenshots and bill photographs from the phone.
- Update the next card debit and four-month bank-balance forecast.
The author @uneta described a household money workflow that could be updated from a phone during a commute. Instead of waiting to enter every purchase at a desk, the author supplied screenshots of QUICPay transactions and photographs of utility bills to Codex through ChatGPT Work.
The working context lived in Google Drive Markdown files and a spreadsheet. New information updated the estimate for the next credit-card debit, while past bank inflows and outflows supported a four-month bank-balance forecast. The useful output was a forward view of upcoming cash, rather than only a retrospective spending chart.
The author's short instruction accompanying a bill photo was to include it too. That makes the recurring interaction concrete: provide another piece of evidence, update the existing records and carry its effect into the forecast. The source does not publish the forecast's actual balance values or a measured accuracy result.
The account also discusses future questions about education costs and investment capacity. Those are possible uses of the records, not evidence that the agent made an investment decision or moved money. No separate human validation checkpoint is described beyond the author supplying the material and discussing the results. The copyable workflow is the combination of retained financial context, lightweight phone capture and a clearly bounded forecast horizon.
Prompt
Original prompt
Source: note.com · 二拠点生活の、まいける (@uneta)
Frequently asked questions
What inputs were used?
QUICPay screenshots, bill photographs and historical bank inflows and outflows.
What did it forecast?
The next credit-card debit and bank balance over four months.
Does the account prove forecast accuracy?
No actual balance values or measured accuracy result are published.
More details
About this account
Codex used supplied records to update debit estimates and a four-month forecast.
Firsthand source opened and read. Prompt is a verbatim excerpt of the published original instruction. The excerpt is the short instruction accompanying a bill photo, not a complete system specification.
Credit: 二拠点生活の、まいける (@uneta)
Source published
Added to Agents In Use
Source checked
Tools and topics
ChatGPT Work · Google Drive · Spreadsheet
Cash flow · Bill photographs · Card debit estimates
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.