Juno, I fetched every page. Here's where you stand.
The structural fixes from the last two months are real. "Product Designer" is gone from the origin. The document title reads "Design Leadership for High-Stakes Systems." The eyebrow signals availability, domain, and location correctly. Good.
The copy fixes did not ship. The homepage still runs a four-sentence professional statement with a TinyFish line. Case study cards lead with terrain instead of outcomes. The agentic section includes UAT Sentinel, putting four agents on the page instead of three. And the private essay serves full content to anyone who fetches the URL.
Most of the paste-ready copy from issue #6 was not adopted. This audit replaces those blocks with updated versions grounded in the site's current state. Everything below uses your published portfolio metrics only. TinyFish does not appear in any recommended copy.
Headline
Verdict: Improved but not sufficient.
Current H1:
"Design leader for platforms where trust is the product."
This tells a hiring manager what you care about. It does not tell them what you've led, at what scale, or why procurement and pharma and disaster response belong to the same designer. A thesis without evidence gets skipped at scan speed.
Replacement H1:
Design leadership for high-consequence systems — $50B+ procurement, pharma supply chains, disaster response, agentic AI.
"High-consequence systems" is the connective frame. It tells the hiring manager why these four domains are one candidacy: in each case, what the user decides based on what the interface shows them has serious downstream effects. The scale figure and domain list provide the specificity the current H1 lacks. The eyebrow line ("Available · Design Leadership · AI-native platforms · San Francisco Bay Area") works as-is. Leave it.
Professional Statement
Verdict: Improved but not sufficient.
Current statement (four sentences plus a current-role line):
"I've led design at $50B+ GMV scale, shipped a pharma supply chain from zero to live in twelve months, and built agentic AI products from concept to production. Across all of it: high stakes, tight constraints, and outcomes that had to hold. I lead design functions and stay close to the work. That combination — strategic scope, hands-on craft — is what makes the difference at the complexity level I operate at.
Currently building agentic platforms at TinyFish."
First sentence is strong. Sentences two through four dilute it with self-description. "I lead design functions and stay close to the work" is a claim the portfolio should prove. The statement's job is to create the reason to look at the portfolio. The TinyFish line violates the boundary I've drawn in every prior piece: TinyFish is your current employer, not your published portfolio evidence.
Replacement statement (three sentences):
Led design and research for Alibaba.com's $50B+ procurement platform — 200K suppliers, 40M buyers, +20% daily transactions. Built 0→1 systems in pharma supply chain ($20M+ margin, 100% partner adoption) and national disaster response (1,689 cases in two weeks). Then designed the trust layer for production agentic AI.
Three sentences instead of the two I'd normally prescribe. The alternative was cramming the 0→1 builds and the agentic work into one sentence with three parenthetical metric clusters, and at scan speed that collapses into noise. Three sentences, each with a clear job: enterprise scale, 0→1 proof, current domain. Seniority is established by "led." The density of outcomes gives a hiring manager a reason to scroll down and see the proof.
Case Study Cards
Verdict: Unchanged.
All three cards lead with terrain or stakes descriptions. The strongest metrics from each project appear on the teaser pages but not on the homepage cards where the scan happens.
The homepage card is the only copy a hiring manager reads before deciding whether to click. If the outcome isn't on the card, it doesn't exist at scan speed.
CS-01 Alibaba — replacement card copy:
+20% daily transactions · +2.2pt NPS · −47% security concerns. Led design and research for Alibaba.com's $50B+ B2B procurement platform — 40M buyers, 200K+ suppliers, regulatory variance across markets.
CS-02 Thermo Fisher — replacement card copy:
$20M+ margin opportunity · 100% partner adoption · 83% IRR. Zero-to-one pharma supply chain — six partners, nine manufacturing sites, twelve months from nothing to live.
CS-03 American Red Cross — replacement card copy:
$847K disbursed · 1,689 cases in two weeks. Six legacy systems replaced by one national disaster-response platform, deployed inside a single response cycle under federal oversight.
A hiring manager scanning these three cards in sequence sees transaction impact at $50B scale, a 0→1 build with financial proof, and a mission-critical system that worked under pressure. That's range and seniority without requiring a click.
Agentic Work Section
Verdict: Improved but not sufficient.
The rename from "Agentic Labs" to "Agentic Work" was the right move. The framing question ("when an agent acts on your behalf, what do you need to see to trust it?") is strong as a section thesis. Two problems remain.
Problem 1: UAT Sentinel. Four agents on the page, but UAT Sentinel is a developer-facing QA tool. It doesn't involve the trust-handoff question your framing promises. Remove it from the homepage. If you want it discoverable, put it on a secondary page.
Problem 2: The cards frame open design problems instead of demonstrating solved ones. Carrier IQ asks "how does an operator verify a result without slowing the loop?" Brand Pulse describes "Designing trust when the agent sees more than you can." A hiring manager reading these sees someone who identified interesting problems. They need to see someone who shipped answers. The fix is small: shift from question to past tense.
Replacement section intro:
Agentic Systems — Production AI Design
Three agents, three verticals, one design thesis: when an autonomous system acts on someone's behalf, what does the human need to see — and when — to trust the output enough to act on it? Each agent runs on live APIs against real data.
Replacement card copy:
Brand Pulse:
76 live signals consolidated into one decision surface. The agent processes more than any operator can verify manually. Designed the information layer so the operator trusts the output without auditing every signal — surfacing what matters for the decision, suppressing what doesn't.
Retail Velocity:
247 venues, one agent. The automation runs itself. Designed the output layer so a field team acts on the result — structured around their decision, not the agent's data model.
Carrier IQ:
5 carriers, 90-second quote cycle. Designed the verification step so an operator confirms the automated result without breaking the speed that justified automating it.
Private Essay Access Control
Verdict: Unchanged.
I fetched /what-do-you-count.html. The server returned HTTP 200 with the complete essay — byline, all six sections, conclusion. The password overlay is a client-side JavaScript check using a SHA-256 hash stored in the page itself. Anyone who views source, disables JavaScript, or fetches the URL programmatically reads the full essay without entering a password.
The password check must happen on the server, not in the browser. A successful fix returns a 401 or 403 HTTP response before any page content loads. If your current hosting doesn't support server-side authentication, move the essay behind a platform that does.
Vercel, Netlify, and Cloudflare Pages all support server-side authentication or edge middleware.
The homepage also still links directly to the essay with a "Read →" label. If the essay is meant to be private, the link should not be on the public homepage. Gate the link behind the same mechanism that gates the content.
Priority Order
- Case study card copy. Highest impact per minute of effort. You already have the metrics. Paste the replacement lines. This is the single change most likely to affect whether a hiring manager at an Act-tier company clicks through or bounces.
- Professional statement. Replace the four-sentence version with the three-sentence version. Remove the TinyFish line.
- Agentic section. Remove UAT Sentinel. Replace the intro copy and the card descriptions. Same editing session as items 1 and 2.
- Headline. Replace the H1. Lower priority than the cards and statement because the eyebrow and document title already carry some of the seniority signal.
- Private essay access control. This requires a hosting-level change, not a copy change, so it's a different kind of task. Close it.
Items 1 through 4 are copy pastes. Ship all four tonight.
After these five changes, the homepage clears the Act-tier bar. The document title, eyebrow, H1, and professional statement all signal Director+ altitude. The case cards lead with outcomes at the specificity Anthropic, Stripe, Brex, and Ramp hiring managers scan for. The agentic section reads as production design practice with a coherent thesis. The only remaining exposure is the essay, and that's a security fix, not a positioning one.
- OpenAI posted two roles on August 6 that map directly to your published evidence — Payments connects Alibaba's transactional trust and Allē's consumer/provider economics, while Engineering Acceleration names the exact instrument-observe-evaluate-rollback loop your Carrier IQ case can demonstrate.
- Gusto's Head of Design posting uses your Trust essay vocabulary almost verbatim — the role asks candidates to define when AI output is ready to ship, needs human review, or should not be sent, making Carrier IQ's wrong-output recovery and Thermo Fisher's regulatory release gate the most literal published evidence you have.
- Giga's Staff Product Designer role describes enterprise agents as non-deterministic systems and asks for zero-to-one through scale — the posting is the clearest test of whether your portfolio can show hands-on correction UX for agent outputs that vary between runs.
- OWASP classifies browser-only authorization as presentation rather than access control — their guidance recommends enforcing authorization server-side and returning only the data the caller is entitled to receive, which confirms that your current password overlay does not protect the essay content.