Standing issue, cleared. "Product Designer" is not on the live page. The evidence says it wasn't there on July 24 either.
Three strings are in play. Separate them.
- Document title. Browser tab, first line of a search result. Currently reads
Juno Chen — Design Leadership for High-Stakes Systems. - Social metadata. Builds the preview card when someone drops your URL into Slack or LinkedIn. Repeats that line, closes on Head of Design availability.
- On-page headline. The sentence a human actually reads. "Design leader for platforms where trust is the product."
Not one of them says Product Designer. I pulled the page twice more, once identifying as LinkedIn's crawler, once as Slack's. Both were served the leadership metadata.
The old title survives in exactly one place I can verify: a cached extraction from roughly a month ago. It predates the July 17 modification stamp on your current page, which is why it disagrees with the July 24 audit and with today's live site. It also carries a headline you retired, "The stakes change. The craft doesn't.", five Labs including Creator Scout, and a TinyFish case card you pulled. That capture is a snapshot of a portfolio you no longer run.
So you have a re-crawl problem, not a copy problem. Platforms cache unfurls on their own side and can lag your origin by days or longer, which means the version of you circulating in DMs and comment threads may not be the version you shipped. Two actions. Ten minutes.
- Run junochen.com through LinkedIn's Post Inspector. It forces a fresh scrape and overwrites the stored unfurl.
- Paste the URL into a private Slack channel and look at what renders.
If either one comes back with the old title, then somebody who received your link in the last few weeks read a label you retired before mid-July.
That was the good news.
Nothing has shipped since July 17
Your server reports last modified July 17, 2026. A week before the July 24 audit. Every recommendation in that piece is still sitting in the document.
Confidence note, because it changes how you weigh the rest of this. A server-reported modification date is a claim made by the server, not an archive, and I could not retrieve an independent snapshot to corroborate it. But the July 24 inventory and today's live page match field for field. Same hero, same "built solo" line under the Labs, same Trust essay parked at the bottom with nothing connecting it to the four products that prove its argument. Treat the page as unchanged.
Which retires the audit I planned. Did your edits move the needle is not a question with a subject this month.
Here is the finding that does have one. The title swap solved altitude. It did not solve aperture, and those two things fail differently. "Design leader for platforms where trust is the product" reads at the right level for a Director+ search, and it names a category rather than a job. Every piece of evidence that would substantiate the claim lives one click below, behind a timed animation, or behind a login screen. Your homepage is a table of contents for a portfolio that outperforms it.
Three evaluators land on that page. Each gives you about ten seconds before deciding whether to spend a minute.
The craft evaluator asks a question the page never answers
Design manager, or the senior IC peer on your loop. Reading for how you think inside an interface. Your tags do that work: Behavioral Transparency, Operator Trust, Human Oversight, Signal to Action. Unusual vocabulary, and correct.
The prose around them is not. The agentic section opens by asking what a person has to see in order to trust an agent acting on their behalf. Right question. Also a question that reader asked themselves a long time ago, and your page never answers it. Below that, the human-work section says the work involves decisions with real costs, and that you led the function and shipped the work. Both sentences are true of every Director-level portfolio in the stack that person is reviewing this afternoon. A carousel doesn't lose that reader. A sentence that could sit under any of the other portfolios does.
Two mechanics finish the job. The agentic section rotates: one card surface, four records, JavaScript cycling. Carrier IQ is what a ten-second visitor sees, and the other three sit on a timer nobody waits out. Make it a static four-up grid. Cheapest high-value fix on the site. Second, "View case" ends at a password field, because every full case ships its HTML and then bounces an unauthenticated browser to /login.html. The one visitor who came specifically for depth gets a login form. What remains for them is scale lines and outcome rows, and those are not craft.
The system evaluator's numbers are on the wrong page
VP Product, or the cross-functional partner sitting on the hiring committee. Scanning for one thing: did this work move a business number at scale.
You own the best material on the site for that reader and none of it is on the homepage. The Thermo Fisher card says a digital supply chain was built from zero inside an organization where every new workflow had to displace an established one. Good sentence about a hard problem, carrying no number. Alibaba runs the same play: 40 million buyers, $50 billion GMV. That is terrain, and terrain is table stakes at this level.
One click down, this is what you are sitting on.
| Case | Homepage card | Teaser, one click down |
|---|---|---|
| Thermo Fisher | "Built from zero" — no figure | $20M margin, 83% IRR, 42% overhead reduction, six pharmaceutical partners, nine sites, twelve months |
| Alibaba | 40M buyers, $50B GMV | +7% DAU, +20% daily transactions, +2.2 NPS, 47% fewer buyer-reported security concerns |
The card tells them where you stood. The teaser tells them what changed while you were standing there. Only one of those gets read in ten seconds, and you put the wrong one there.
Your Thermo Fisher teaser calls the $20M a "margin opportunity." The full case calls it "margin recovered annually." A system evaluator who reads both sees the drift and quietly discounts the figure.
Use "recovered annually" in both places, assuming the business case supports it. Recovered is sourceable; opportunity is a projection and reads like one.
The leadership evaluator cannot find the mandate
Biggest loss on the page.
A CPO or a retained recruiter is not checking whether you deliver work. They are checking whether you have run a function. Your professional paragraph asserts that you combine design-function leadership with hands-on craft. At Director+, everyone asserts that. Assertion is the weakest evidence you own.
The proof is on your teaser pages, unused. Alibaba lists your role as Head of Design & Research, North America. Red Cross lists Product Design Director / GM · 0→1 Build. A general manager title on a national deployment is a mandate signal, and it appears nowhere a leadership evaluator will encounter it. You are the only person authorized to put those titles on your own homepage, and you have chosen not to.
"Built solo" compounds it. I flagged it in July. To a CPO, solo reads as capability without leverage.
And the Trust essay, the most executive-grade artifact you have published, sits at the bottom under Field Studies with no stated relationship to the four products that test its argument. It names five handoffs where trust in an agent is won or lost. Four of your Labs demonstrate four of them. No visitor could know that.
One negative, stated once
Every full case study returns its complete HTML to an unauthenticated request, and then a browser script redirects to the login page. The private pricing essay behaves the same way beneath an in-page overlay. The gate controls nothing, because by the time it fires the outcome it appears to govern has already happened.
This belongs in a bug queue. I am putting it in an altitude audit for one reason. You are positioning as the person who designs the moment a human authorizes a consequential action. A technically curious evaluator who opens view-source finds a decision gate that renders, and then loses.
The copy
Drop-in. No alternates.
Headline
Replaces the on-page H1. Same altitude, tighter aperture. The current line names a category; this one names a job.
Design leadership for the decision gate: the moment a person accepts, overrides, or reverses what a system just did.
Professional statement
Two sentences. Every clause backed by a live page on your site.
I build and lead design functions for systems where the wrong call costs margin, compliance standing, or a disaster survivor's payment: Head of Design & Research for Alibaba.com's $50B+ GMV procurement platform, Product Design Director on a pharmaceutical supply-chain platform taken from zero to live in twelve months. I now work the same problem in agentic software, where four live products test what a person has to see before they verify, override, or act on what an agent produced.
TinyFish stays on its own line underneath: Currently building agentic platforms at TinyFish. Employer context. Never portfolio proof.
Case card lines
Each one opens with the title, which carries the leadership read whatever tier the card serves, then closes on the outcome for its primary evaluator.
CS-01 Alibaba, system tier.
Head of Design & Research, North America. Desktop carried 25% of traffic and 80% of transaction value, and nobody had named the gap: +7% DAU, +20% daily transactions, +2.2 NPS, 47% fewer buyer-reported security concerns.
CS-02 Thermo Fisher, system tier.
Product Design Director. Zero to live in twelve months across six pharmaceutical partners and nine sites in three regions: $20M+ margin recovered annually, 83% business-case IRR, 42% overhead reduction, all six partners committed.
CS-03 American Red Cross, craft tier. Craft despite the GM title, because a design manager can hold the whole problem in one sentence: a single case record that has to serve every role in the chain, operated by people with no training, during a surge.
Product Design Director / GM, 0→1. One case record serving all three roles in the chain, under surge conditions and federal oversight from day one: national deployment in six months, six systems collapsed to one, $847K disbursed and 1,689 cases in the first two weeks.
If you name those three roles on the card, pull the exact wording out of the full case before you publish. Do not let the homepage assert a triad the case study does not.
Agentic Labs reframe
Retitle the section Agentic Systems: the thesis, in production. Introduce it with the essay linked:
In Trust Is the New Interface I argued that trust in an agent is won or lost at five handoffs: Intent-Setting, In-Progress, Output Review, Decision Gate, and Loop Feedback. These four systems are that argument running against real inputs. Each is live. Each is built around one handoff you can test yourself.
Then label each card with the handoff it owns.
- Carrier IQ, Decision Gate. Five carrier quotes in ninety seconds, with operator verification that does not slow the loop.
- UAT Sentinel, In-Progress. A seven-phase QA pipeline that keeps the engineer in oversight while the agent runs long.
- Retail Velocity, Output Review. One agent auditing 247 venues. The design problem is the evidence a human acts on, not the audit.
- Brand Pulse, Intent-Setting. Seventy-six live signals on one decision surface, for when the agent sees more than the user can.
Delete "built solo" everywhere it appears. Replace with "designed, built, and shipped end to end."
Two cautions.
- Loop Feedback is the handoff none of the four cleanly demonstrates, and UAT Sentinel's regression detection comes closest. Name the gap on the page rather than hoping nobody counts. The evaluator who reads the essay and then works down the cards will count, and being the person who flagged the hole is worth more than the hole costs.
- Retail Velocity's live app displays TinyFish Search, Fetch, and Agents as its scan technology. Strip the employer branding out of the demo, or move that one behind an access request. An evaluator who reaches that screen stops reading a question about design and starts reading a question about judgment.
Where each evaluator lands. Headline carries leadership. Sentence one of the statement plus the CS-01 and CS-02 cards carry system, because that is where the scale numbers live and the headline deliberately withholds them. Labs carries craft. All three sit above the fold only if the four-up grid replaces the carousel, which is what makes that twenty-minute fix the first thing you touch.
The aperture argument from issue four still holds. The same evidence reads differently depending on who is holding it. Today none of the three gets a landing point in the first ten seconds, a month-old capture of the retired title is still retrievable, the platform caches are unverified in either direction, and the page is a week past an audit nobody acted on. All of it is one afternoon.
- Altana's design-system language: The live Head of Product Design posting at Altana describes the design system as a machine-readable grammar that product managers, engineers, and agents compose, which is the most explicit statement yet of what your Labs section is actually demonstrating.
- HackerOne reports to the CPO: The Director, AI Product Design role names agent-to-agent interaction, human-in-the-loop control, transparency, and risk visualization in posting copy that maps almost line for line onto Carrier IQ and the Decision Gate handoff.
- The trust ladder is two axes, not one: Anthropic's analysis of agent autonomy in production found experienced users both auto-approving more often and interrupting more often, which suggests Watch → Verify → Delegate hides a second variable your Labs cards could make visible.
- Attribution and reversibility have shipped: Linear's agent-assisted editing release distinguishes agent-written changes visually, preserves the agent as author, and lets users restore earlier versions — a competitor to your thesis in production, and a useful benchmark for what Loop Feedback should look like when you build it.