Two people finish the tutorial. One says, “I can now research independently,” and delivers an attractive report. The other says, “I still lack professional capabilities,” and delivers a failed experiment with clear gaps. Which is closer to a useful next step?
Inspect the evidence rather than the tone. Professional competence has no progress bar filled by pages read. Someone can reconcile financial figures while lacking capacity analysis, or produce an equity curve without knowing how to recover actual positions after a connection failure. Put those differences into T12, the capability-boundary record.
Methods and experience need different evidence
You will state what you can do under specified conditions, turn gaps into assessable deliverables, and choose one bounded next task while recording responsibilities outside its scope. Neither profit nor confidence supplies a percentage of “institutional competence.”
Methods tell you to inspect sources, assumptions, costs, and counterexamples. Experience also involves encountering changed definitions, faulty ledgers, industry turns, and execution constraints, then recognizing the consequential gap when materials are incomplete. Solving one exercise supports its particular task. Transfer to another dataset, market, or tool needs further evidence.
Professionals specialize too. Industry research, model validation, and trading operations are different responsibilities. G65 and G66 locate missing work; they do not require an individual to reproduce every position, buy the same equipment, or claim an occupational identity.
CFA Institute's diligence standard asks its applicable members and candidates to understand model assumptions and limitations and make appropriate checks of data and analytical tools.1 It helps describe work responsibilities here. Completing this course does not confer a credential or certification under that standard.
Inspect Jianning's six files before reading her self-assessment
T12-67 is a wholly fictional evidence archive. It does not establish that a real reader passed assessment or change Target's C1-B materials or C2-B's real-data status. Jianning works in the equipment industry, has four research hours per week, and chooses company research for depth. Quantitative tasks remain mechanism exercises.
The E03 ledger uses consistent currency units; no national currency is specified for it. The later annual-cost example separately uses dollars.
| File | Supplied contents | Supported scope |
|---|---|---|
| E01: company exercise | Supplied fictional source text for two periods; revenue 100 to 120; units and locations recorded; 20% growth recalculated | Extraction and arithmetic within provided materials, not independent acquisition of a complete company packet |
| E02: update | First report preserved before opening the second packet; a new cash gap changes the judgment; old version retained; no additional original public disclosure obtained | Updating under given inputs, with independent sourcing still missing |
| E03: small ledger | Initial cash 1,000; buy 10 shares at 20, sell 10 at 19; each order costs 1; no other changes; AI's 990 omits fees, corrected to 988 | Checking this trade and finding omitted costs, not long-history strategy validation |
| E04: experiment log | Three mechanism attempts, including negative results; evaluation history influenced revisions; no real long-term prices, events, or complete permissions | Preserving trials and contamination, not estimating persistent alpha |
| E05: review | Jianning checks E03 from original inputs with a calculator; a second AI agrees with the same summary; no other person takes the raw materials and repeats the task | A separate arithmetic path, without independent-reviewer handoff evidence |
| E06: operation | Only two month-end simulated observations after freezing; no orders; no partial-fill, permission, or fault-recovery evidence; changed household allowance untested | Narrow observation; real RUN and RISK not passed |
E03 is immediately checkable: 1,000 − 201 leaves 799 after purchase; sale adds 190 − 1, ending at 988. Investment loss is 12, or 1.2%. Losing money here provides positive evidence that Jianning can locate an omitted fee. A later stock gain would not repair an unchecked copy of 990.
She need not restart every financial-statement lesson, and these files do not qualify her to manage a quantitative portfolio. Her shortest useful next step is to locate an original company disclosure herself and ask another person to find the same field. E01 and E02 supply method evidence; applying it beyond a given packet is the missing step.
Write a deliverable in each T12 gap
Keep understanding, given-case work, independent research, funding, and operation separate. A task can contain a checked component and an unverified component without producing one overall grade.
| Capability task | Current evidence | Gap and next deliverable |
|---|---|---|
| Data acquisition | E01 locates supplied text; E04 lacks real long-history inputs | Company route: original disclosure with identity, publication time, period, location, and recalculated metric. Quant route: permissions, historical identity/events, gaps, and versions |
| Industry judgment | Equipment experience and a given-hypothesis update; no unique understanding of the target established | Test one product-adoption or collection mechanism using public material, competing explanations, scope, and disconfirming facts |
| Statistics and experiments | Three logged attempts and contamination; one reconciled ledger | With actual admitted data, freeze question, benchmark, divisions, costs, and trial budget; retain negative results, exposure, and sensitivity. One trade cannot estimate stable parameters |
| Engineering reproduction | Self-recalculation, without tool/settings/environment versions or complete exports | Original inputs, specification, ledger, outputs, and instructions; execution in another environment with differences explained, without the author rewriting code to do it for the recipient |
| Independent review | E05 lacks another reviewer's raw-input check | Another person checks a key judgment, calculation, and variant from originals; preserve disagreements and corrections, limited to the scope actually checked |
| Operation and responsibility | Two no-order simulations | Verify signals, permissions, active orders, holdings, cash, and fees for the chosen tool; test exceptions, pause, and recovery. Recalculate household allowance after changed conditions separately |
Jianning does not have to fill all six gaps immediately. She can restrict the task to ordinary nonfinancial companies and low-frequency public materials, leaving high-frequency work, derivatives, and multistrategy operation outside her responsibilities. Smaller scope reduces demands without removing known cash gaps or errors inside it.
Quantitative work likewise need not await every advanced mathematics topic. Explain returns, denominators, samples, dependence, leakage, selection, and costs actually used, then study model-specific requirements. Indicator names, purchased computing power, and AI-generated code cannot replace a ledger attachment.
FINRA's institutional algorithmic-trading guidance discusses testing, version records, monitoring, and problem handling as control work.2 An individual can simplify organization while retaining appropriate checks. Missing monitoring or recovery evidence leaves the work at research or simulation, rather than making the absence of colleagues an exemption.
What AI can help produce
AI can suggest original-source entry points, compare period text, extract numbers with locations, implement calculations, propose counterexamples, and organize a handoff. Saving preparation time can help Jianning use her four hours for important verification. Outputs still need their original sources and actual results checked.
It cannot turn missing prices into true history, an absent contract into an executed agreement, model agreement into another person's independent review, or “validated” into evidence that permissions and duplicate-order protection work. Without an actual recipient's work, leave the handoff outcome pending.
Self-review after an interval and an AI asked to find weaknesses can be useful. They remain different from organizational independence; the same person may retain a blind spot. When a consequential question exceeds current ability, narrow it, obtain suitable review, or stop. A more confident formulation does not complete the missing work.
Record an advantage with its cost and conditions
Small scale may ease some orders, conditional on quantity, quotes, and cost evidence. Fixed research costs can also weigh more heavily. In a separate example, an annual $600 research-data budget is 6% of $10,000 task capital, but 0.06% of $1,000,000. These are ratios of the same research budget, not investment fee rates, scale recommendations, or promised service benefits.
Freedom of choice permits skipping unfamiliar markets, declining fashionable opportunities, and choosing zero allocation. It also gives up coverage and opportunities. A scope statement must survive the next social-media surge before that freedom can be called an exercised advantage.
Time horizon can be more flexible than that of some short-term client mandates. Household payments, employment, and personal constraints still apply, and waiting has opportunity costs. Verify that funds will not be needed early while allowing new evidence to reject the hypothesis. A long horizon cannot protect a broken business explanation.
Industry experience can improve questions about equipment testing, substitutes, quality, and collections. Familiarity can also create blind spots and employment-linked concentration. A good product does not establish a cheap stock, and salary plus investments in one industry may share risk.
This advantage does not include sending employer-confidential information to AI or treating material nonpublic information as a trading route. Investor.gov describes illegal insider trading in breach of a duty or relationship of trust, including certain tipping and misappropriation.3 Unclear information boundaries need an applicable-requirements check, not an automatic pass because someone works in the industry. No individual legal determination is made here.
These four conditions can improve task selection and error control without automatically generating alpha. Persistent improvement over a feasible, comparable benchmark needs its own evidence. State which condition you will test in which limited task.
Give the next step a completion condition
Jianning chooses company research. In two sessions of two hours each, she will independently obtain a suitable public disclosure, verify identity, publication time, and metric scope, extract one number relevant to the question, write an opposing explanation, and ask another person to locate it. Time is her work budget, not an optimal learning cycle.
If the source is unavailable or the definition cannot be resolved, retain the blocker and next entry point instead of forcing a conclusion when time expires. Real quantitative data, operation, and household funding are not upgraded, and downloading an application is not completion evidence. One priority is enough for this round.
Afterward, T12's data row may say “specified original disclosure and one metric checked.” Statistics, engineering, and operation do not pass with it. Keep the old table and date the new evidence so capability expands through records, rather than stronger adjectives.
Change two pieces of evidence
Explain. Why can E03's 1.2% loss support a skill while a second AI's agreement does not establish independent research?
Transfer. Another reviewer now recalculates 988 from E03's original inputs and checks a variation. Meanwhile, Jianning changes the order from 10 shares to 60, claiming the earlier correct process permits the same funding approval.
Keep cash 1,000, purchase price 20, and fee 1 per successful order, with long-only trading and no borrowing. For this question, explicitly add: insufficient cash rejects the entire order without a fee. Update the independent-review row and calculate the proposed order's result before reading on.
Reference answers · reveal
Answer. The reviewer adds evidence for checking this given synthetic ledger and variant, without supplying real data, unknown-task transfer, or production operation. The new order requires 60 × 20 + 1 = 1,201, exceeding 1,000. Reject it, charge no fee, and retain cash 1,000 and zero shares. The earlier ten-share check cannot lend the missing 201.
At most 49 whole shares fit the stated cash-and-fee arithmetic, but silently replacing 60 with 49 is not approval of the original request. Partial resizing would require a permitted specification and separate authorization. Preserve the parameter version and rejected order.
Skill evidence comes from finding the omitted fees and reproducing the account, not grading profit. Model agreement alone supplies no independent sourcing or review process. Update the particular row, without grading the whole person.
Save a dated capability boundary
Fill T12 from work actually completed on C1-B or C2-B: task and scope, evidence location, method, prompts or assistance, independent checks, resource gaps, excluded domains, next deliverable, and review date. If no real project is complete, practice with this fictional archive and label it accordingly.
G62's Target packet and source demonstration are author-known historical reconstruction; a reader still needs additional original evidence, a preserved first report, and an update. G63's C2-B synthetic mechanism and issuer-identity check do not replace missing long-term real prices, events, permissions, or the reader's no-code execution. Record real independent RES, household RISK, and operational RUN separately, retaining unverified fields where evidence is missing.
Ask someone to choose a T12 row, explain your work from its attachments, and change a condition. Explaining why 1,000 cannot fund sixty shares at 20 establishes a checkable mechanism. An unknown real contract supporting revenue requires public evidence or a narrower conclusion. “Familiar with research” cannot merge those two tasks.
Turn a professional gap into the next inspectable piece of evidence. Data, experience, review, and operation cannot substitute for one another through a title. Personal freedom and scale also need testing under real constraints. Knowing the next deliverable advances the work more usefully than asking whether you are professional enough.
References
- CFA Institute: Standard V(A)—Diligence and Reasonable Basis, sources, third-party research, quantitative assumptions, and output validation. Work principles do not confer a professional credential.
- FINRA: Regulatory Notice 15-09, March 26, 2015, development, testing, monitoring, and controls for relevant firms; not a universal personal organization chart.
- Investor.gov: Insider Trading, material nonpublic information and related unlawful conduct. General boundaries, not a case-specific legal conclusion.
Editorial status
AI-assisted draft synchronized to the accepted second-round Chinese source. T12-67 is entirely fictional and certifies no real person's independent research or funding readiness. 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.