Halfway through a study, all that remains is the sentence “cash flow looks good.” Where did the numbers come from? What did AI change? Which inputs were missing? If you move to another application, can someone continue the work?
First make a small research packet that does not depend on an application. Then ask which parts CoTrading has actually been observed to support. A missing product check need not erase the research question, inputs, and reconciliation already completed.
The executable exercise uses the supplied materials, hand calculations or an ordinary table, and a handoff record. The CoTrading section illustrates how to report the scope of existing evidence. Opening the application is not required to calculate the results, and this chapter is not a product-operation tutorial.
Specify the deliverable before choosing the tool
You will preserve materials, calculations, and judgments separately; compare two tool paths against the same checkpoints; and complete an external handoff when a feature is unverified or an input is missing. Reconciling one cash table does not complete independent company research, and no trading task is included.
A browser locates and displays original sources. Hand arithmetic or a spreadsheet supplies reproducible calculations. AI can organize, explain, and identify gaps. The researcher checks the evidence and decides whether a judgment should change. CoTrading could carry parts of the process where its relevant capabilities are verified; its research label cannot replace that verification.
This independent C1 company exercise, C1-61-v1, is synthetic. It is neither Costco nor Target financial data, neither of C1's December 2024 and April 2025 frozen packets, nor a valuation of a real security. All amounts are USD millions for teaching year Y1.
Keep three complete inputs together
Material A: accounting excerpt. Net income is 12, after deducting noncash depreciation of 3. Operating receivables increase 4: the related sales are recognized but their cash has not arrived. Operating payables increase 2: the related expenses are already in income but remain unpaid. Cash purchases of fixed assets are 8; new borrowing provides 6 in cash; opening cash is 20. There are no other cash flows, income-tax adjustments, corporate actions, or currency changes.
Material B: task log. Reconcile only net income to operating cash flow, the selected definition of free cash flow, and ending cash. Identify each item as supplied by A or calculated by the researcher. Mark unprovided information unknown. Do not assess a stock as cheap, trade, or access other research or accounts. These instructions define scope; they do not prove that software enforces permission isolation.
Material C: calculation convention. For this exercise, define free cash flow as operating cash flow minus cash purchases of fixed assets. SEC's financial-statement guide distinguishes operating, investing, and financing cash, and explains why profit and cash differ.1 The selected free-cash-flow measure does not universally equal money available for discretionary use; state and reconcile its definition.2
Retain A, B, and C on the input page, labeled synthetic, versioned, and with units. Giving a model only “profit 12” and inviting it to invent typical receivables, capital expenditure, or financing would change the materials. Complete inputs matter more than the first answer's polished report style.
Follow the generic path to a portable table
First, open the two original SEC pages in the references. Locate cash-flow categories and the free-cash-flow discussion. They support concepts; A supplies this company's numbers. Record title, URL, access date, and the specific concept supported, rather than relying on a search summary.
Second, create separate input and reconciliation pages in a plain document or spreadsheet. Preserve A unchanged. Enter the supplied items, their adjustment reasons, and sums. No programming installation or particular research application is necessary.
| Item | Adjustment or result | Source or reason |
|---|---|---|
| Net income | 12 | A's starting value |
| Add back noncash depreciation | +3 | Already reduced income without a cash payment |
| Increase in receivables | −4 | Recognized sales have not produced cash |
| Increase in payables | +2 | Recognized expenses have not yet been paid |
| Operating cash flow | 13 | 12 + 3 − 4 + 2 |
| Defined free cash flow | 5 | 13 − capital expenditure of 8 |
| Net increase in cash | 11 | 13 − 8 + borrowing of 6 |
| Ending cash | 31 | Opening 20 + increase 11 |
Third, separate arithmetic from interpretation. Operating cash of 13 exceeding income of 12 does not establish sustained operating improvement. Comparable periods, reasons for receivable/payable changes, and later payments remain relevant. Borrowing of 6 increases cash without being operating cash generation. Ending cash of 31 cannot establish that free cash flow was 11.
Fourth, preserve a readable numerical snapshot and explanation. Save versioned inputs, table, formulas, sources, unknowns, and conclusion as text or a table. This manuscript and its author record retain those inputs and results; that does not claim testing of any spreadsheet application's save menu. For your packet, reopen the saved material and verify units, version, and 13, 5, and 31, rather than only checking that a file exists.
For CSV handoff, define fields for item, amount, unit, period, source, and adjustment reason. Saved values are not necessarily workbook formulas. Attach the original inputs and convention so a recipient can repeat 12 + 3 − 4 + 2 without opening the original application. Real household or account information is unnecessary for this synthetic task.
Use a deliberately wrong delivery to test the review
Here is a prepared AI-style summary: “Receivables increased because the business expanded, so add 4. Operating cash flow is 21, free cash flow 13, and ending cash 39. The stock is worth buying.” This is a constructed diagnostic example, not an answer generated by CoTrading in this work.
Locate the first mechanism error. The receivable increase represents recognized income without collected cash, so subtract 4 in the reconciliation rather than add it. Reversing that sign overstates operating cash by 8, and the later free-cash-flow and ending-cash figures inherit the same overstatement. Three internally consistent answers can share one wrong input treatment.
“Worth buying” also exceeds the task. There is no security price, valuation, risk assessment, or forecast. Correcting the arithmetic does not rescue that recommendation. Record the two findings separately: calculation error located and investment inference unsupported.
If another version omits capital expenditure, leave free cash flow and ending cash undetermined. Do not substitute zero to complete the table. Recover by obtaining the missing field, keeping the failed version, and applying the same checks. Missing evidence cannot be replaced with the expected answer.
What the existing CoTrading record actually reports
The October 29, 2026, product evidence record reports checking an installed Windows executable with file version 0.1.3 and observing its matching application window. This English drafting work relies on that existing record; it did not repeat the application observation. The record does not identify the latest available download or establish other platforms' behavior.
After entering a new conversation, the recorded screen displayed a task-entry area, screenshot/file-upload entry, company-research examples, a model label, and browser and chart entries. The task-start control was disabled with no text entered, and the sidebar showed no outputs. Discoverable entries support an interface observation, not successful source reading, table generation, or saved research.
The user then stopped the interface flow with Escape. The record states that observation ceased without filling or sending this task, testing attachments or outputs, exporting, recovering, or restarting the process. See the historical product evidence card in Chinese. Its original Incomplete runtime finding remains historical evidence, not a new product test by this author.
Do not turn the upload entry into “CoTrading automatically builds an evidence table after upload,” or invent the name of an untested export button. You can still prepare A, B, C, and the expected output fields for a later tool check, allowing organization and explanation while requiring sources and unknowns. Whether this installed release accepts that input or produces a usable result remains unverified.
Compare the two paths at the same checkpoints
| Task | Generic path | Existing CoTrading evidence |
|---|---|---|
| Define question and inputs | Complete synthetic A/B/C saved here | Task not entered or sent |
| Check concept sources | SEC originals read for this manuscript | Application browsing/citation not verified |
| Recalculate and diagnose | 13/5/31 and the error chain independently calculated | No application output obtained |
| Save and hand off | Complete readable manuscript and author record | Attachment receipt, outputs, saving, and export untested |
| Handle missing fields and failure | Complete constructed error/missing-input exercise | Product failure/recovery not executed |
The deliverable now is an external research packet. Product receipt, generation, preservation, and recovery still lack evidence. A later product check must fill those gaps with actual observations; repeating hand arithmetic cannot certify application behavior.
Product runtime remains unverified. This chapter can be assessed within its limited scope of generic research organization and existing observations; public release still requires complete bilingual content and the applicable content review. The historical card's original runtime judgment is not silently rewritten.
If a future tool provides only a summary, complete source locations and intermediate numbers externally and identify that work as external. If results cannot be exported or reliably copied, retain the limitation. Seeing an answer once does not create a durable research archive.
Separate where a file is saved from where input is sent
Reading public SEC material in a browser and submitting research inputs to a model are different data flows. No task or attachment was submitted through CoTrading for this exercise. Actual application routing, logs, retention, and provider training policies were not checked. A model label and a local executable do not establish local-only processing.
For a generic AI exercise, copy only the synthetic A/B/C after inspecting the actual configuration and official information about what reaches model services. Record local files, browser sources, submitted content, and output destinations. “Saved locally” does not establish “never transmitted.”
A product update, changed model route, attachment format, or saving method calls for rechecking the relevant evidence. Reuse the same synthetic task to compare inputs, results, and handoff before updating the evidence card. Old screenshots support only their recorded version and scope. No brokerage account or real financial record is needed here.
Change one input and keep the handoff honest
Explain. Why is a cash increase of 11 not operating cash generation of 11? Why does an attachment entry not establish that material was read?
Reference answers · reveal
Answer. Cash changes include different activities, here operating cash, fixed-asset purchases, and borrowing. A visible control needs actual receipt and output evidence before it supports a reading claim. Neither conclusion can be inferred from the final impression alone.
Transfer. Change A so receivables decrease by 1 instead of increasing by 4. Keep every other input unchanged. Recalculate the three checkpoints without the original application, and identify records to update.
Answer. Collected receivables contribute +1. Operating cash is 12 + 3 + 1 + 2 = 18; defined free cash flow is 18 − 8 = 10; ending cash is 20 + 18 − 8 + 6 = 36. Register input v2, preserve v1's 13/5/31, and update calculations, explanations, and the handoff list. A correct 36 under changed inputs is not a tool error simply because the earlier expectation was 31.
Save research that can leave the application
Keep original inputs and versions, scope, formulas and units, concept sources, results, errors and gaps, human-check fields, and the status of which tool actually completed each step. Another reader should be able to recalculate both examples without your conversation history.
For handoff, ask the recipient to choose operating cash 13 or revised ending cash 36, locate inputs, and calculate before reading your explanation. Then ask which steps were actually performed by the application. Record the arithmetic result and the capability judgment separately. Unverified product fields stay unverified.
Fix the question and evidence before handing work between tools. Research value rests on traceable originals, calculations, and update conditions. If a different application can still pick up the explanation, the packet is more durable than a summary intelligible only inside its first conversation.
References
- SEC: Beginners' Guide to Financial Statement, February 4, 2007, cash-flow sections. Supports concepts, not this fictional company's figures.
- SEC: Non-GAAP Financial Measures, Question 102.07, May 17, 2016. Free-cash-flow definition, reconciliation, and discretionary-use limits; no individualized legal conclusion.
Editorial status
AI-assisted draft synchronized to the separately accepted website-reviewed Chinese source. Product scope is limited to existing interface-observation records; runtime validation has not been performed. See the author review. Independent AI full-text and bilingual review is complete. Human financial review and target-reader testing have not been performed and are not scheduled for this edition.