junochen.com Homepage Audit — The Trust Surface That Contradicts Its Own Thesis

Verified state, July 24, 2026

The direct HTTP source fetch now returns Juno Chen — Design Leadership for High-Stakes Systems as the document title. Hero headline: "Design leader for platforms where trust is the product." Meta description names Head of Design availability. Hero body cites $50B+ GMV scale, a pharma supply chain shipped zero to live in twelve months, agentic AI products built concept to production, current work at TinyFish.

Since the prior homepage audit on July 9 and the CPO-lens audit later that month: source-level positioning has moved from Product Designer to design leader. Five public case studies (Alibaba, Thermo Fisher, Red Cross, Equinox+, Allē) with substantive descriptions. Creator Scout is gone from the Agentic Labs carousel.

What remains open. A separate tool-based render of the same URL still returns "Product Designer" as the page title. I cannot confirm whether this is a stale CDN cache, a client-side rendering race condition, or a pre-rendered meta tag that was never updated. Different cause, different fix. What I can confirm: at least one non-source access path still serves the old positioning. A CPO who arrives through a link preview, a cached Google result, or certain browser conditions may see a version of you that contradicts the version you built.

The private essay still delivers its full body to any unauthenticated fetch. UAT Sentinel is still in the carousel. The Equinox+ teaser leaks internal vocabulary into buyer-facing metadata.

Your thesis, articulated in the Trust essay and embedded in every case study, is that design's job is to calibrate trust at the moment a human decides whether to act on what a system produced. A technically sophisticated buyer visiting junochen.com is doing exactly that. They are deciding whether to trust what your portfolio produced. Four live issues on the site create trust-calibration failures at the exact moment your thesis says those failures matter most.

Your site is a trust surface. Audit it like one.

P0 — what-do-you-count.html delivers private content before any gate

Verified state. The homepage links the "What Do You Count When the Playbook Doesn't Exist?" card to what-do-you-count.html. A HEAD request returns a 308 redirect to /what-do-you-count, which returns HTTP 200 with a 34,860-byte HTML document. That document contains a client-side password overlay using crypto.subtle.digest and a pf_access_v1 localStorage key. The full essay body sits in the page source before any password entry. TinyFish launch context. Pricing strategy. Cost-model reasoning. Agent-step economics. Outcome-billing logic. PAYG structure. Tier-naming decisions. All delivered to any bot, crawler, or technically literate visitor who opens DevTools.

No server-side denial. No 401. No 403. The content is delivered, then hidden with JavaScript.

The trust failure. This was the P0 in the access-gate piece two weeks ago. The argument has not changed because the state has not changed.

A CPO or engineering leader evaluating a candidate whose thesis is "design the decision gate where a human decides whether to trust what the system produced" will view-source. Not out of paranoia. Out of professional habit. Anyone who has shipped production software does this the way a chef tastes the food at a new restaurant. When they do, they find that the decision gate on your own site is cosmetic. The overlay looks like a gate. It functions as a curtain.

Asymmetric risk. If a buyer never inspects the source, this costs nothing. If they do, the cost is not a credibility ding. It is a direct contradiction of the thesis that makes you distinctive. Your most commercially sensitive strategic thinking is protected by the exact mechanism your Trust essay argues against: a surface that looks like control but provides none.

Minimum viable fix. Remove the what-do-you-count card from the homepage. Delete the link. The essay can live at its URL for anyone who has the password, but it should not be discoverable from the public portfolio surface. Fifteen minutes.

Correct fix. Move access control server-side. The page returns a 401 or 403 until authentication succeeds. The essay body does not exist in the response until the gate is passed. If your hosting stack does not support server-side auth natively, move the essay to a platform that does: Notion with restricted access, a password-protected Substack post, a serverless function that checks credentials before returning HTML. Retire the client-side pf_access_v1 pattern for any content that is genuinely private.

P1 — TinyFish case card appears in at least one render path

Verified state. Direct HTTP source fetch on July 24 does not contain a visible CS-05 TinyFish case card. A separate tool-based render of the same URL does return one. I cannot confirm which render path produces the card. Could be a cached build artifact, a conditional rendering flag evaluated differently by the extraction tool, a CDN edge state that was never purged. What I can confirm: the card is present in at least one non-source access method, and a July 19 check also surfaced it.

The trust failure. The published-portfolio boundary is clear. TinyFish is valid as current-role context and verbal interview proof. Currency, not collateral. Not a written case study. Not an external portfolio piece. If a rendered version of your homepage shows a TinyFish case study card, it violates that boundary in the one place where the violation is most visible.

A CPO follows a link from your outreach email. The rendered page shows a TinyFish case card. They click it. If it leads to a case study page, you have published work that should be internal. If it leads to nothing, you have a broken link on a portfolio that argues for precision. Both outcomes damage you. The broken link might be worse.

Asymmetric risk. If a buyer's browser renders the source-level HTML and no TinyFish card appears, this costs nothing. If their access path hits the rendered state that includes the card, you have a boundary violation visible to the exact person you need to trust your judgment about what should and should not be public. At the Head of Design level, that judgment is the signal.

Fix. Confirm no render path, no cached state, no build artifact produces a CS-05 TinyFish card. If the card exists in your source code behind a conditional flag, remove it entirely. Do not rely on the flag. Conditional rendering is the same category of problem as client-side access gating: it depends on execution context, and you do not control every execution context. Purge the CDN cache after removing the card. Verify from an incognito window on a different network.

P2 — Agentic Labs carousel still includes UAT Sentinel

Verified state. The direct source carousel contains four entries in its CARD_DATA array: Carrier IQ, UAT Sentinel, Retail Velocity, Brand Pulse. The homepage embeds an iframe for uat-sentinel-demo.html and links to uat-sentinel.junochen.com. Creator Scout is gone from the carousel, though the subdomain remains live at creator-scout.junochen.com.

The canonical external-use set remains three: Brand Pulse, Retail Velocity, Carrier IQ.

The trust failure. UAT Sentinel is an autonomous regression audit workspace with parallel browser agents, evidence capture, and release gating. Technically substantive. Genuine agentic design thinking. Quality is not the problem. Boundary is.

Every outreach email, every cover letter, every positioning document references three Agentic Labs. UAT Sentinel's CSP headers reveal a connection to agent.tinyfish.ai. A buyer inspecting your site with technical eyes sees that connection and has a reasonable question: personal project work or employer IP? You may have a clean answer. The question itself is friction you do not need at the moment they should be reaching for the calendar link.

Asymmetric risk. If a buyer does not count the Labs, this costs nothing. If they do, the mismatch between outreach saying "three" and the site showing four creates a calibration question at exactly the wrong moment. The tinyfish.ai CSP connection compounds it. Now they are pausing to wonder about IP boundaries instead of reading the next case study.

Fix. Remove UAT Sentinel from the CARD_DATA array. Remove the embedded iframe. Three Labs, three cards: Brand Pulse, Retail Velocity, Carrier IQ. If you want to expand the boundary to include UAT Sentinel, do that deliberately. Update all outreach materials to match. Prepare a clean explanation for the tinyfish.ai connection. Do not let the site make a boundary claim your outreach contradicts.

Creator Scout's subdomain being live is lower severity. Not linked from the homepage, not discoverable through normal navigation. Leave it.

P3 — Equinox+ teaser metadata says "Teaser" to buyers

Verified state. The Equinox+ teaser page has the document title Equinox+ · Teaser · Juno Chen and no meta description tag. The visible content already includes strong proof: Product Design Director, 600K+ premium members, 200+ clubs, 0→1 in 90 days, 4.8★ App Store.

The trust failure. "Teaser" is internal vocabulary on a buyer-facing surface. A CPO with six browser tabs open sees Equinox+ · Teaser · Juno Chen and processes "teaser" as "incomplete." The absent meta description means link previews in Slack, email, or LinkedIn pull nothing useful or pull arbitrary page text.

Asymmetric risk. If a buyer opens the Equinox+ page and never reads the tab title, this costs nothing. If they do, "Teaser" signals that you consider this page provisional. You want them to see enough proof that they ask for more. "Provisional" works against that.

Fix. Replace the document title:

Add a meta description grounded in the published case evidence:

Led product design for Equinox+: 0→MVP in 3 months, 5 brands on one platform,
600K+ premium members, 4.8★ App Store. Instructor-as-navigation architecture
drove nearly 2× retention for members who followed within 30 days.

Every number from the published case page. Nothing invented.

The source/render drift underneath all of this

One issue cuts across P0–P3. Direct HTTP source returns leadership-forward copy. A tool-based render returns "Product Designer." I cannot confirm from available evidence whether this divergence stems from a stale CDN cache, a client-side rendering sequence that overwrites server-delivered HTML, or pre-rendered meta tags that were never updated. The fix depends on the cause, so check all three:

• CDN cache: Purge all edge nodes. Verify from multiple regions.
• Client-side rendering overwriting server HTML: Move the hero copy to server-side or static generation so the leadership framing is present in the initial HTML response regardless of JavaScript execution.
• Pre-rendered meta tags (for social previews, link unfurling): Update them explicitly. These are often set once during initial deployment and forgotten. They are the version of you that shows up in Slack when someone pastes your URL.

The version a buyer sees depends on how they access the site. You do not control that. Control what you can: make sure every access path delivers the same positioning.

Copy blocks for the Head of Design version of this site

The source-level hero copy is already leadership-forward. What follows are refinements, not rewrites.

Revised headline and tagline

Current source headline: "Design leader for platforms where trust is the product."

Strong. Names the thesis and the level. Two options for a supporting line:

Full version: Design leader for high-stakes platforms. I design the moment where a human decides whether to trust what the system produced.

Compressed version: Design leader for high-stakes platforms. I design the human decision point in AI-native systems.

The full version is more precise and more distinctive. The compressed version is ten words and holds at any display size. If the hero renders the second line at body-text scale, use the full version. If it renders at headline scale, use the compressed.

Two-sentence professional statement

Leads with scope and consequence, not craft. The second sentence gives the buyer vocabulary for the problem you solve. Maps directly to the Trust essay's argument without requiring them to have read it.

Case card positioning lines leading with leadership proof

CS-01 Alibaba.com

"Led the redesign" instead of "procurement redesign." The verb carries the level signal.

CS-02 Thermo Fisher

"Built" as the lead verb. Then the organizational constraint that makes this a leadership story.

CS-03 American Red Cross

A CPO who has managed cross-functional complexity recognizes "could not be simplified away" as lived experience. Résumé language would say "managed multiple stakeholders." This says something a person who has actually done it would say.

Agentic Labs section reframing

Current framing: "four live agents built solo" with the question "what does a person need to see when an agent acts on their behalf?"

The question is excellent. Keep it. It is the buyer-facing version of the Trust essay's thesis applied to agentic systems.

"Built solo" is the problem. As I noted in the CPO-lens audit, "built solo" reads as maker evidence. A Head of Design candidate who emphasizes solo building raises a question for a CPO evaluating cognitive load reduction: does she want to lead, or does she want to make? Wrong question to invite.

Revised section framing:

Count drops to three. Matches the canonical boundary. "Designed and shipped" replaces "built solo." The Labs are positioned as thesis-testing rather than side projects.

For each Lab card, lead with the trust-calibration problem it tests:

• Brand Pulse. Tests whether a brand team trusts sentiment signals they did not generate.
• Retail Velocity. Tests whether a merchandiser trusts demand forecasts that contradict their instinct.
• Carrier IQ. Tests whether a logistics operator trusts a routing recommendation when the cost of a wrong route is a missed SLA.

Each one is a version of the same problem your case studies solve at enterprise scale. The Labs prove you are working on the next version of that problem. The case studies prove you have shipped the current one. That sequence matters.

Priority sequence

• Today. Remove the what-do-you-count card from the homepage. Fifteen minutes. Eliminates the highest-severity trust failure.
• Today. Remove UAT Sentinel from the Agentic carousel. Drop the count to three. Purge CDN cache.
• This week. Verify no render path produces a TinyFish case card. Purge all cached states. Test from incognito on a different network.
• This week. Update Equinox+ teaser document title and meta description.
• This week. Move what-do-you-count to server-side auth or an external platform with real access control.
• This week. Audit the source/render drift. Identify the cause. Fix it at the layer where it originates.
• Next week. Revise case card copy and Agentic Labs framing per the blocks above.

Source-level positioning is leadership-forward. The thesis is clear. The case inventory is strong. What remains is boundary hygiene: making sure the site's behavior matches its argument. For a candidate whose argument is that design's job is to calibrate trust at the decision point, that match is the proof.