Piece 1 showed where the portfolio breaks its own rules. This piece assesses whether the surfaces themselves do their trust-ramp jobs when the architecture is fixed.
The Trust essay's calibration ladder runs Watch → Verify → Delegate. Users watch the system, compare its output to their own judgment. They verify by spot-checking, building a reliability model. They delegate after enough evidence accumulates.
Your portfolio has the same ladder. Teaser = Watch. Full case = Verify. Interview = Delegate. Each stage has one job: reduce the specific uncertainty that blocks the evaluator from moving to the next stage.
Watch: Do the Teasers Earn the Next Click?
A hiring manager scanning teasers is in Watch mode. Comparing your output to their mental model of what a strong candidate looks like. The teaser's only job: give them enough signal to form a hypothesis worth testing. Not enough to evaluate you. Enough to want to.
Alibaba. Mandate-forward. Opens with "The signals before anyone touched a screen," names the procurement gap, delivers metrics at scale ($50B+ GMV, +7% DAU, +20% transactions, −47% security concerns). A CPO reading this forms a clear hypothesis: this person found the problem, built the case, shipped the solution. Earns the click.
Thermo Fisher. Mandate-forward. Opens with "The supply chain six partners never had," names the cost of late exception discovery (3–5× higher after commit dates), delivers the outcome ($20M margin opportunity, 100% partner adoption, 83% IRR). The regulatory visibility language is the strongest differentiator on the entire site. No other candidate's teaser will mention QA evidence and FEMA compliance in the same portfolio. Earns the click.
Red Cross. Mandate-forward. Displaced families, six disconnected tools, a federal clock. The constraint density is the signal: national deployment, untrained volunteers, surge conditions, 1,689 cases in two weeks. Best line on the site: "Complexity lives in the system, not the interface." Eight words that function as a design-leadership thesis statement. Earns the click.
Allē. Mandate-forward. Opens with "When the product is someone's face, points aren't enough." Dual-surface structure (consumer iOS + provider backoffice) visible in the teaser itself. Metrics land: 3.2× redemption, 47% reactivation, CAC from $92 to $42. Proves system thinking across consumer and enterprise surfaces simultaneously. Earns the click.
Equinox+. Fails the Watch function.
The title tag reads "Equinox+ · Teaser · Juno Chen." Every other teaser has a mandate-forward title. This one has a label. A CPO scanning browser tabs sees four cases that announce what they prove and one that announces what it is.
The body compounds the problem. Alibaba names the procurement gap. Thermo Fisher names late exception discovery. Red Cross names six disconnected tools under a federal clock. Allē names the gap between loyalty points and someone's face. Equinox+ names a platform consolidation. "Five brands. One platform. No compromise" is a tagline, not a design problem. "0→MVP in three months" is an execution speed claim that doesn't tell the CPO what design decision made the difference. The teaser says 75 hours of research produced four member archetypes and five architecture decisions, and that instructor followership became the primary navigation object. That last detail is the hypothesis worth leading with. But a CPO in Watch mode who can't form a testable hypothesis from the opening won't read deep enough to find it.
The case-number mismatch makes it worse. The body says "Case Study 03" while the homepage labels it Case Study 04. Small error. Also exactly the kind of error a trust-calibration framework says compounds.
Fix the title. Make it mandate-forward. Name the design problem, not the brand. Fix the case number. Pick one. Make it consistent everywhere. Restructure the opening. Lead with the design constraint (five brands with incompatible identity systems sharing one platform) and the decision that resolved it (instructor followership as primary navigation object). Fifteen minutes for the metadata. One focused hour for the content reframe.
Verify: Does the Gating Architecture Support the Next Stage?
The full case is the Verify surface. The hiring manager formed a hypothesis from the teaser and now wants to test it. They request access. You grant it. They read the full case and build a reliability model.
The current gating architecture eliminates this stage. All five cases deliver complete HTML before the localStorage check fires. A reviewer who views source receives the full case without ever requesting access. The Verify stage requires a relationship. A touchpoint where the hiring manager asks and you grant. The current architecture requires nothing. That touchpoint vanishes.
Uncontrolled access also kills your ability to sequence. In a properly gated architecture, you control when each case becomes available and match the case to the buyer's anxiety. A CPO worried about speed gets Thermo Fisher first. A CPO evaluating function-building gets Red Cross. When every case is available to anyone who can right-click, you lose the ability to lead with the right story for the right audience. You lose the positioning advantage the portfolio was built to create.
Delegate: The Interview
If Watch and Verify worked, the interview starts from earned credibility. The CPO arrives having formed a hypothesis from the teaser, tested it against the full case, built enough of a reliability model to engage on specifics. Their first question is shaped by the narrative you chose to present.
If the Verify stage leaked, the interview starts from uncontrolled ground. A CPO who read the full Alibaba case without requesting access arrives with pre-formed judgments you didn't get to frame. Their first question won't be "tell me about Alibaba." It will be a specific challenge anchored to a detail they fixated on, and you don't know which detail that was. You can't sequence the narrative because they already consumed it out of sequence. You can't lead with the strongest evidence because they may have already discounted it based on a reading you weren't there to guide.
Delegate requires that the evaluator's trust was calibrated through designed interactions. Leaked Verify means uncalibrated Delegate. You walk in blind to what they already think they know.
Architectural Prescription
Server-side gating. Do this first. Move the access check from localStorage to the server. The HTML response for protected pages should not contain case content unless the request includes a valid token. This single change fixes the Decision Gate violation on the private essay and the Intent-Setting violation on the five full cases. Client-side overlays are UX. Server-side checks are trust architecture. Your essay knows the difference. Build accordingly.
Remove the private essay link from the homepage. The essay is internal-only evidence. It should not be discoverable from the public loop. When a hiring manager needs it, send the link directly, with the access token, at the right moment in the conversation. That is a designed feedback path. A homepage link is an undesigned one.
Sequence access by buyer. Once server-side gating is in place, grant access to specific cases based on the buyer's profile. A CPO evaluating AI fluency does not need all five cases on day one. They need the Trust essay and one case that proves system-level thinking (Red Cross or Allē), with the rest unlocked as the conversation progresses. This turns the Verify stage from passive reading into active trust calibration, which is what the essay argues every handoff should be.
The raw materials are strong. Four of five teasers do their Watch-stage job. The full cases contain real depth. The Trust essay is the best positioning asset on the site. The architecture connecting these surfaces is where the framework breaks its own rules. Fix the architecture and the portfolio practices what the essay preaches. Leave it, and every panel that includes someone who thinks about systems will notice the gap before you walk into the room.