Every artifact in Juno's evidence inventory has a context where it builds the case and a context where it undermines it. A Carrier IQ screenshot anchors a making claim during a portfolio walkthrough. Pasted into a cold email, it reads as overreach — inspectable evidence delivered to someone who did not ask for it. A TinyFish production observation spoken in a live interview establishes daily fluency with AI systems. Written into a cover letter, it becomes a claim with nothing behind it the reader can verify.
The access-control boundary shapes what gets built and how. An artifact designed for public deployment carries different proof obligations than one designed for verbal context. This is a design constraint on the portfolio itself, and treating it as administrative overhead produces evidence that fails in the contexts where it actually gets used.
This framework is the single reference for all positioning documents, outreach, cover letters, and interview prep. It classifies artifacts into four disclosure classes and maps each class to the contexts where it can appear.
The access-control boundary determines the class, and the class determines the usage rules.
Disclosure Classes
Public. Intentionally published at junochen.com, visible to anyone, verifiable by a hiring manager who clicks the link. Permitted in cold outreach, cover letters, portfolio presentations, live interviews, verbal discussion. The reader can verify the claim by following the link.
Controlled-interview. Can be shown on screen during a live, permissioned session — portfolio review, panel interview, design exercise debrief. Cannot be linked in cold outreach, left behind as a PDF, or referenced in writing where the recipient could forward it without context. Authorization occurs before delivery. A browser overlay that hides content until someone clicks "I'm a recruiter" is presentation logic, not access control. If the HTML ships before the check, the artifact is reachable but not authorized.
Verbal-only. Can be discussed in conversation. Spoken, not shown, not linked, not screenshotted. Supplies context, fluency, and bridge narrative. Does not carry written portfolio claims, metrics, or visual evidence. The listener gets the insight but not the receipt.
Unavailable. Cannot be referenced in any external context — not named in a cover letter, not mentioned in an interview, not alluded to in outreach. Its existence in a build queue or on a homepage does not change its class if the authorization record does not support external use.
Context × Class Matrix
| Class | Cold Outreach | Cover Letter | Portfolio Presentation | Live Interview | Verbal Conversation |
|---|---|---|---|---|---|
| Public | Permitted | Permitted | Permitted | Permitted | Permitted |
| Controlled-interview | Prohibited | Prohibited | Permitted; authorize before delivery | Permitted; authorize before delivery | Permitted |
| Verbal-only | Prohibited | Prohibited | Prohibited | Discuss only; do not show or link | Permitted |
| Unavailable | Prohibited | Prohibited | Prohibited | Prohibited | Prohibited |
"Authorize before delivery" means a conversation with the recipient confirming the session is appropriate for non-public material. Not a legal release — a mutual understanding that what you are about to show is shared in context and not forwarded.
TinyFish Classification
The TinyFish rules — current-role context yes, technical AI currency yes, bridge narrative yes, portfolio proof no — are what this framework produces when you classify employer-specific artifacts honestly. They follow from the classification logic rather than being added as exceptions.
- The fact that Juno works at TinyFish: Public. It appears on LinkedIn. It contextualizes the career narrative. It answers "what are you doing now."
- Production observations (how agents behave in deployment, what operators do with approval gates, what traces reveal about model behavior): Verbal-only. Juno can discuss these in live conversation to establish that she works with production AI systems daily. There is no authorization to show TinyFish screens, share internal traces, or present TinyFish product work as inspectable evidence.
- TinyFish product screenshots, internal traces, product definitions, process documentation: Unavailable. No disclosure authorization exists.
Several items in the build queue name TinyFish-derived screenshots and traces as build targets. Those targets conflict with the standing boundary and remain undeliverable until the boundary changes. The framework does not prevent building them internally. It prevents deploying them externally. Preparing an artifact internally and deploying it externally follow different rules.
Artifact Inventory
The Portfolio Build Queue organizes evidence into five domains. Classification of the current inventory within each:
Trust Patterns
| Artifact | Class | Notes |
|---|---|---|
| Published Trust essay (junochen.com/trust-is-the-new-interface) | Public | Core positioning piece. Original Watch-Verify-Delegate ladder is retired; the essay still carries the five handoffs and design principles. |
| Thermo Fisher QA release gate (as described in the essay) | Public | Published example of a Decision Gate in a regulated context. |
| Carrier IQ longitudinal control record | Unavailable | The correction event, next-run change, and revised release criterion do not yet exist as published evidence. |
| Visual Trust Pattern Library | Unavailable | Partially drafted, not published at junochen.com. |
| TinyFish correction interface (proposed) | Unavailable | No disclosure authorization. |
| Two-axis revision field note (companion piece below) | Public upon publication | Classifiable as public once intentionally published at junochen.com. |
Design Iterations
| Artifact | Class | Notes |
|---|---|---|
| Figma → Claude Code → production → trace → revision loop | Unavailable | Named as build target; not yet published. |
| Process essay ("How I Design AI Products") | Unavailable | Draft material; not published at junochen.com. |
| Annotated trace-to-design decision | Unavailable | If derived from TinyFish traces, remains unavailable regardless of completion. If derived from Agentic Labs, becomes public upon publication. |
Design Patterns
| Artifact | Class | Notes |
|---|---|---|
| Carrier IQ, Retail Velocity, Brand Pulse case teasers | Public | Visible on homepage. |
| Full case studies (if built for controlled sharing) | Controlled-interview | Authorization before delivery required. Current browser-only checks do not provide real access control. |
| Seven-state agent state machine | Unavailable | Pending. |
| Agentic Labs decision-moment annotation | Unavailable | Draft, not published. |
AI-Native Execution
| Artifact | Class | Notes |
|---|---|---|
| Agent product definition (problem, success criteria, "done" for a probabilistic system) | Unavailable | Source determines eventual class. |
| Eval-forward design framework | Unavailable | Pending. |
| Design-system constraint proof (unconstrained → encoded → constrained → prevented drift) | Unavailable | Pending. |
Presentation and Credibility
| Artifact | Class | Notes |
|---|---|---|
| Failure-and-recovery documentation (Labs) | Unavailable | Pending per Lab. |
| TinyFish failure example | Verbal-only | Can be spoken; cannot be shown or written. |
| Live-demo scripts | Controlled-interview | Useful in live sessions only. |
| Career-narrative visual | Unavailable | Pending. |
| First-person failure narrative | Unavailable | Pending. |
| "The Confidence Rule" portable-judgment draft | Unavailable | Published as editorial but not at junochen.com. |
UAT Sentinel
UAT Sentinel appears on the homepage alongside the three canonical Labs (Carrier IQ, Retail Velocity, Brand Pulse). The boundary documentation prohibits its use in outreach, cover letters, or portfolio-match annotations. No record establishes its origin, independent disclosure authorization, or whether its homepage presence was an intentional boundary expansion or configuration drift.
Visibility on the homepage is a delivery fact, not a permission fact. Do not reference UAT Sentinel in any external context until provenance is established.
Using This Framework
When preparing any external-facing document — outreach email, cover letter, portfolio walkthrough script, interview prep — check each artifact you plan to reference against the inventory above. If it is not listed, classify it by asking three questions in order:
- Is it intentionally published at junochen.com? → Public.
- Is there authorization to show it in a live, permissioned session? → Controlled-interview.
- Can it be discussed verbally without showing or linking? → Verbal-only.
If none of the three applies, the artifact is unavailable.
If the answer to question 1 is "it's on the site but I'm not sure it was intentionally published for external use," treat it as controlled-interview until confirmed. Being reachable does not make an artifact authorized for external use.