Ten roles survived verification this morning. Ten, not a hundred. There is no honest arithmetic that gets that number higher. So this edition changes the organizing variable and leaves the count where it is.
The previous board sorted 22 roles by design problem. Correct cut for the question it was answering, which was which of these is worth wanting. That is no longer the constraint. You are not short on good roles; you are short on a rule that tells you which one to touch on a Saturday.
So this board sorts on entry route: sponsor, application, recon, verification. Four doors, and four different ways to lose an afternoon.
Ten days old, the freshest thing in the pool, one of only two still inside the 21-day hot window, and the only one where published evidence changes what you should say.
Corrections that change the scores
Four posting dates in the prior pool were wrong.
- Affirm is 19 days old, not 10. The July 22 stamp was a refresh, not a publish.
- Fetch downgraded. The posting decayed into the extended band at 31 days, not 25. It restores on a fresh Director-level requisition or a recruiter who confirms the search is live.
- Maven is 40 days. Brex is 24.
Fetch reclassified takes the verified pool to four listings in the extended band, not three.
Three compensation bands also came into view on employer pages this pass. One of them moves a score. Fieldguide posts $180K–$210K, which caps the cash line and takes that role from a presumed 15/15 to 14/15. Vanta's two bands score at the top.
How to read a verdict line
Two fields now, not one. Running them together has been costing you weeks.
Fit verdict is the rubric's answer. Act, Watch, Ignore-until-trigger. It tells you whether the opportunity deserves your next four years.
Today-action is the routing answer. Sponsor-first, Apply, Recon, Verify. It tells you what your hands do before dinner.
A role can score 15/15 on fit and still be the wrong file to open today. Ambience and Maven are both exactly that.
Each entry also splits its unknowns. A proof gap is yours: a case you cannot show, a scale you have not run. You close it with positioning. A decision gap is theirs: undisclosed compensation, an unnamed hiring manager, ambiguous scope. You close it with a question. Different work, different day. Mixing the two up is how a strong candidate burns three hours on a cover letter that answers a question nobody asked.
Cluster 1 — Sponsor-accessible, 3 roles
Three roles where a real person is publicly identifiable and the application form is the wrong first move. At Head-of-Design level the application does not screen you in; it files you into a stack where the thing that differentiates you is structurally invisible. This cluster runs first even though cluster two matches it on Act density, because access decays faster than postings do. A named executive is reachable while the search is still taking shape, and that window shuts weeks before the requisition does. Gusto published its own design operating manual, which hands you a hook that needs no throat-clearing. Vanta and Amplitude each have a product executive who owns design on the record, with nothing public tying them to the requisition itself.
1. Gusto. Payroll, benefits and HR administration for small and mid-sized employers. The Unified Service Platform is the internal system its support organization works in to answer customer questions.
- Role: Head of Design, Unified Service Platform. Posted July 22. 10 days. Greenhouse
- Stage / capital: Late-stage private, multi-thousand headcount. Raise figures not verified this pass.
- Offices: San Francisco, Denver, New York.
- Value chain: Vertically integrated. Gusto carries the payroll-tax compliance obligation end to end, which drags design scope into the operational surfaces and not only the marketed ones.
- Best at: Swallowing payroll-tax complexity so the employer never encounters it.
- Signals: A second design role, Product Designer for HR Experiences, went up July 30. Two design postings in eight days, and the only sub-48-hour listing anywhere in this pool.
- Tier: Growth-stage platform. Read the Growth-stage platform playbook before outreach.
- Rubric: Company 11/12 (AI centrality 3, Stage 3, Design influence 3, Trajectory 2). Role 15/15 (Comp 3, Scope 3, Craft 3, AI exposure 3, Portfolio 3).
- Portfolio match: The prior battlecard read the posting's negation language correctly and then made one error: it treated "AI-native" as a single requirement. It is two, and they take different answers. Gate one is the service decision. The posting wants judgment about uncertainty, graceful failure, human override, and when a generated response should not go out at all. Answer that with the American Red Cross platform. Six legacy systems collapsed into one, three roles (field volunteers, finance officers, program directors) working a shared record for the first time, FEMA audit trails built into the flow, $847K disbursed. Support platforms and disaster relief share the structure that matters here: an internal operator hands over an answer with a compliance consequence attached to it. Bring the argument from "Trust Is the New Interface" that users anchor on the worst outcome, not the average one. Gate two is harder. The posting names Claude Code and Cursor and expects the leader to prototype the ugliest flows personally. Gusto's own account of its design org describes retiring static Figma files, teaching designers the pull-request lifecycle, one designer shipping roughly 150 pull requests in eleven weeks, and AI expectations written into interview plans. Your published portfolio does not answer gate two. Your current job does: Head of Product at TinyFish, an enterprise web agent platform, shipping to production daily. Context, never design proof.
- Verdict: Act. Today-action: Sponsor-first. Proof gap (yours): nothing published shows code-based prototyping. Do not stall to build one. Name the gap and set the operating reality of your current role against it. Downgrade trigger: 22 days of silence on the sponsor path, at which point this converts to Apply.
2. Vanta. Continuous security-compliance automation. It watches a company's systems, gathers evidence against frameworks like SOC 2, and runs third-party vendor risk.
- Role: Head of Design. Posted July 6. 26 days. Ashby
- Stage / capital: Late-stage private. Headcount and total raised not verified this pass.
- Offices: Remote, US. No geographic gate. Posted cash range $365K–$513K plus equity.
- Value chain: Midstream platform, sitting between the auditor and the audited. Unusually good for design influence, because the product is the workflow itself rather than a report squeezed out of one.
- Best at: Turning a discrete annual audit event into a continuous state.
- Signals: A Staff Product Designer for Design Systems went up June 26, ten days ahead of this one. Two design roles in ten days at a company with no publicly visible design leader is a function being stood up, not a seat being backfilled.
- Tier: Enterprise platform. Read the Enterprise platform playbook before outreach.
- Rubric: Company 11/12 (AI 3, Stage 3, Design influence 3, Trajectory 2). Role 14/15 (Comp 3, Scope 3, Craft 2, AI exposure 3, Portfolio 3).
- Portfolio match: Vanta's vendor-assessment flow runs like this. A reviewer requests evidence, logs findings, records a residual-risk recommendation, then picks one of three outcomes: approved, conditionally approved, or not approved. The middle one is the hard one. "Conditionally approved" is a promise about the future, and the question worth putting to Epling is what inside the product forces that condition to closure. Confidence in the whole system leaks out of that middle state. Map Alibaba onto it. You named a structural gap nobody had named, that procurement buyers on a $50B GMV platform were running on consumer-grade UX, built the research case, took the executive mandate, ran three cross-functional sprints, moved daily transactions +20% and NPS +2.2 points. The remit here spans GRC, Trust, Platform, self-serve and AI, which makes it a mandate-building job before it is a craft job. Mandate-building on a problem nobody has named yet is the specific thing on your record. Differentiator: the five-handoffs framework in your published essay predates this requisition, which makes it a thesis rather than a pitch reverse-engineered from a job description.
- Access vector: Jeremy Epling, Chief Product Officer, who oversees engineering, product and design per the company's leadership page and said as much himself in March. The posting names no hiring manager. No public source connects him to this specific requisition. Buyer-adjacent, not confirmed. Treat it that way.
- Verdict: Act. Today-action: Sponsor-first. This corrects the earlier Vanta piece, which reported posting age and compensation as unavailable. Both resolved: July 6, band posted. Proof gap (yours): you have not run a design org at Vanta's current headcount, and a five-surface remit will put that question on the table early.
3. Amplitude. Product analytics. Teams instrument their software, then query behavioral data to decide what to build next. Now spans Core Amplitude, Statsig for experimentation, and Wave.
- Role: Head of Product Design. Posted July 7. 25 days. Greenhouse
- Stage / capital: Public. Growth trajectory is mixed, which is the entire reason Stage scores 2 and not 3.
- Offices: San Francisco.
- Value chain: Midstream tooling. Design influence is real, and bounded by a category that commoditized fast.
- Best at: Making behavioral cohorting legible to people who are not analysts.
- Signals: The posting names a 15-person team across three product lines. No second design posting found this pass.
- Tier: Growth-stage platform. Read the Growth-stage platform playbook before outreach.
- Rubric: Company 9/12 (AI 2, Stage 2, Design influence 3, Trajectory 2). Role 15/15 (Comp 3, Scope 3, Craft 3, AI exposure 3, Portfolio 3).
- Portfolio match: The hard surface at Amplitude sits past the dashboard: the moment a product manager reads a generated insight and decides whether to ship against it. The failure mode is a confidently narrated correlation that turns out to be noise, and that does more damage than a wrong number, because narrative form suppresses scrutiny. Lead with Brand Pulse. You built a live system that scores every mention as it lands and produces a weekly narrative written by the agent, which means you have already made the calls about how much interpretive authority to hand an automated narrator and where to leave the seams showing for the human reading it. Nobody else in this search shipped that surface solo and can open it live in the room.
- Access vector: Gab Menachem, Chief Product Officer, appointed April 14, running product management, design and growth, reporting to the CEO. He is 109 days in seat, past the 90-day executive-hire window. The trigger that reopens that window is a follow-on design leadership posting inside 90 days of the hire. This one landed on day 84. It reopened itself.
- Verdict: Watch (company 9). Today-action: Sponsor-first. Decision gap (theirs): the requisition still names no manager and no reporting line, and the earlier readiness brief flagged unresolved ambiguity in the publicly visible design leadership. Nothing has changed. Ask about it in the first ten minutes.
Cluster 2 — Application-optimized, 3 roles
Three Staff-level roles where the posting language sits close enough to published work that the application is the strongest single move available. No organizational-scale proof gate, nothing to establish about whether you can run a function. All three can go out today. The trap is that they scan as one role: agentic AI in a regulated-ish workflow. That will tempt you into three variants of the same letter. They are three different surfaces with three different failure modes, and each needs its own evidence.
4. Fieldguide. AI for audit and advisory firms. Its Field Orchestrator runs substantive audit testing across population analysis, sampling, evidence collection, extraction and workpaper generation.
- Role: Staff Product Designer. Posted July 10. 22 days. Ashby
- Stage / capital: Venture-backed private, sub-500 headcount. Raise total not verified this pass.
- Offices: Remote, US.
- Value chain: Downstream application sitting directly on the professional's decision. Highest design-influence position on this board.
- Best at: Generating a testing plan the auditor reviews, then stopping at major steps and leaving exception review and final sign-off with the human.
- Signals: Posted range $180K–$210K plus equity, one nationwide listing, no state variants. No second design posting.
- Tier: Enterprise platform, regulated. Read the Enterprise platform playbook before outreach.
- Rubric: Company 12/12 (AI 3, Stage 3, Design influence 3, Trajectory 3). Role 14/15 (Comp 2, Scope 3, Craft 3, AI exposure 3, Portfolio 3).
- Portfolio match: The decision moment is a senior auditor reviewing linked evidence and exceptions before sign-off. The failure mode is volume-induced rubber-stamping: the agent produces something that looks finished, so the reviewer approves what they did not read. Thermo Fisher mySupply is the structural analogue, and it is exact. 197 orders in the queue, surface the three that matter. Five failure modes turned into five modules, out of 32 interviews conducted standing on warehouse floors. Six pharma partners, 100% adoption inside twelve months, $20M in margin unlocked. Differentiator: most strong AI design candidates have designed the agent's output. You designed the reviewer's queue. And you are shipping enterprise agents into production every week right now, working traces, attribution and reversibility as operating problems rather than case-study vocabulary. Each half is rare on its own; the pair is what they cannot hire around.
- Verdict: Act. Today-action: Apply. Decision gap (theirs): the posted band tops out under your floor. Negotiation, not disqualification, but put it on the table in the first call instead of at offer. Say it plainly: "I want to flag early that I'm targeting above the posted band on cash, and I'd rather know now whether there's room at the top of the range or in the equity component than find out at offer." Downgrade trigger: 43 days with no recruiter contact, or the requisition leaves the Ashby feed.
5. Affirm. Consumer point-of-sale financing. A purchase gets split into scheduled payments, underwritten one transaction at a time.
- Role: Staff Product Designer, Agentic Experiences. Posted July 13. 19 days. Greenhouse
- Stage / capital: Public. Post-IPO, which is why Stage scores 2 and equity upside is a moderate line rather than a strong one.
- Offices: Remote, US.
- Value chain: Downstream consumer application with an embedded merchant surface.
- Best at: Underwriting the individual transaction rather than the person, at checkout speed.
- Signals: Cash range present on the requisition. No design cluster in the current feed.
- Tier: AI-native. Read the AI-native playbook before outreach.
- Rubric: Company 10/12 (AI 2, Stage 2, Design influence 3, Trajectory 3). Role 14/15 (Comp 3, Scope 3, Craft 3, AI exposure 3, Portfolio 2).
- Portfolio match: What separates this from every other agentic role on the board is that the person doing the delegating cannot be trained. A consumer authorizing an agent to act on credit picks up a repayment obligation they may never have modeled, so the failure mode is an obligation entered without comprehension, not an error in output. Lead with Allē: 30M members, dual-surface across provider and consumer, 3.2× redemption lift at $42 acquisition cost. Then bring the trust ladder from the essay, Watch to Verify to Delegate, as a staging model for how much authority a first-time consumer should be able to hand over and how each rung gets earned instead of granted by default. Differentiator: agentic design candidates come almost entirely out of enterprise, where you get to onboard the user. Consumer volume at that scale is what they cannot hire around.
- Verdict: Act on rubric. The requisition names consumer agentic initiatives, search and discovery, and a design language for trust, clarity and predictable outcomes in financial products. That settles the surface. This is consumer delegation, not merchant tooling, so the Allē opening holds. Today-action: Apply.
6. Brex. Corporate cards and spend management. Issue the cards, enforce the policy, close the books.
- Role: Staff Product Designer, AI. Posted July 8. 24 days. Employer page
- Stage / capital: Capital One subsidiary since the acquisition closed April 7, 2026, at roughly $4.5B. The pre-IPO equity thesis that put Brex on your primary list is gone.
- Offices: San Francisco, New York.
- Value chain: Downstream application over a financial data substrate the company owns.
- Best at: Compressing expense reconciliation toward zero human touch.
- Signals: Range posted on the requisition. No qualifying design cluster in the current feed.
- Tier: AI-native. Read the AI-native playbook before outreach.
- Rubric: Company 8/12 (AI 2, Stage 1, Design influence 2, Trajectory 3). Role 13/15 (Comp 3, Scope 2, Craft 3, AI exposure 3, Portfolio 2).
- Portfolio match: The surface is mid-flight correction. The agent categorizes a transaction, flags a policy exception, routes an approval. The damage comes from the irreversible classification rather than the wrong one: the action completed and no affordance exists to take it back. Carrier IQ is the evidence. You designed and shipped an agent that completes forms and selects coverage in a regulated line, which meant deciding in code where the human interrupt sits and what it costs to put it there. The prior Brex brief already built the delegation and in-flight supervision argument. Do not rebuild it. Use it as written and hold the cover letter to reversibility alone.
- Verdict: Watch on stage, not on craft. The mandate is still the closest agentic-supervision fit on the board. Today-action: Apply. Decision gap (theirs): Staff scope at Brex may be a defined lane rather than an expandable one. The posting does not say. Ask the ceiling question in screen. Downgrade trigger: day 43, August 20, unless a recruiter conversation has opened by then.
Cluster 3 — Recon-targets, 2 roles
Two roles that look good and cannot be scored cleanly, because something the company controls is left unsaid. Do not apply. Do not write outreach. Each needs exactly one question answered, and in Vanta's case the question fits inside a conversation you are already trying to start.
7. Vanta — Staff Product Designer, Design Systems. Posted June 26. 36 days. Extended: may be filled or deprioritized. Ashby
- Offices: Remote, US. Posted cash range $236K–$278K plus equity.
- Tier: Enterprise platform. Read the Enterprise platform playbook before outreach.
- Rubric: Company 11/12. Role 13–15/15, depending on the unresolved variable. Posted band scores 3/3.
- The gap (theirs): The posting already puts agent runtime interfaces, chat and canvas surfaces inside the mandate, alongside Alpaca components, tokens and governance. What it does not say is whether the systems designer sets the patterns for those surfaces or inherits them from the product teams. That single answer is the difference between a 15 and a 13.
- Portfolio match: If the mandate does set those patterns, the artifact to bring is the five-handoffs structure from the trust essay, applied as a system primitive rather than a per-screen decision: how a design system encodes confidence, provenance and reversibility once, so that five product areas do not each invent a private vocabulary for uncertainty. A systems argument, not a case study. This is the one role on the board where that is the stronger opening.
- Recon action: Read the two Vanta requisitions side by side and settle whether the systems role sets or inherits the patterns for the five surfaces named in the leadership posting. Do not apply to both. Two applications from one candidate ten days apart reads as untargeted, and the leadership role is the one worth the slot.
- Verdict: Watch. Today-action: Recon. Trigger to downgrade: day 43, August 8. Trigger to restore: confirmation from the Head of Design conversation that the systems mandate owns pattern-setting across the AI surfaces.
8. Fetch. Consumer rewards app. Scan receipts, earn points, redeem. The other half of the business is an outcomes-based advertising platform where receipt-verified purchase data closes the attribution loop for brands.
- Role: Director, Product Design. Posted July 1. 31 days. Extended. Gem
- Stage / capital: Late-stage private, consumer scale. Raise total not verified this pass.
- Offices: Remote, US, per the requisition. Geographic gate cleared. Fetch is not a Bay Area company and does not need you to be sitting in one.
- Value chain: Two-sided. The consumer app sits upstream of the advertiser platform, and value concentrates on the advertiser side, which is probably not where this design mandate lives.
- Signals: Senior Product Designer, Growth went up July 13, twelve days after the Director role. A Director paired with a growth IC hire tilts the org toward acquisition and retention work.
- Tier: Growth-stage platform. Read the Growth-stage platform playbook before outreach.
- Rubric: Company 8/12 (AI 2, Stage 2, Design influence 2, Trajectory 2). Role unscoreable. Compensation is null on the employer record and no state variant surfaced anywhere.
- Portfolio match: Equinox+ is the closer read than anything agentic here. Zero to MVP in three months across five brands on iOS and Android, which is a multi-brand consumer velocity problem, not a trust problem. That is the finding in itself. This is the only role on the board where your agentic work is not the lead, and if you write to it as an AI role you will sound like someone who did not read it.
- Recon action: One question. Does Director cover the consumer app or both sides? Consumer-only means the ceiling is already built.
- Verdict: Watch. Today-action: Recon. Decision gap (theirs): compensation and scope. Cap this at one email. Trigger to Ignore: no scope answer by day 43, August 13. Trigger to restore: a mandate that spans the advertiser surface.
Cluster 4 — Hold-until-verified, 2 roles
One of these clears Act on the rubric and is held back only by the decay rule. The other is a Watch either way, which settles which one you verify first. Both are past four weeks. The fit verdict stays where the rubric puts it so neither drops off the board, but the today-action is verification, because applying into a 38-day-old search already in final rounds costs you a slot and teaches you nothing. Verification means a direct message to a recruiter or a design contact asking whether the search is still open. Reloading the posting page is not verification.
9. Ambience Healthcare. Clinical AI. It turns the clinician-patient conversation into structured notes, orders and patient instructions during the visit, then suggests ICD-10 and E/M codes and flags documentation that lacks the support it needs.
- Role: Staff Product Designer. Posted June 24. 38 days. Extended. Ashby
- Stage / capital: Venture-backed private, clinical AI. Raise total not verified this pass.
- Offices: San Francisco. Range posted.
- Value chain: Downstream, sitting on the highest-consequence decision moment anywhere on this board.
- Best at: Coding-aware documentation across ambulatory, emergency and inpatient settings inside one product.
- Signals: No recent signal. Last confirmed milestone is this posting's first publication on June 24. No second design role showed up in the Ashby feed this pass, which on a 38-day-old requisition reads as a single-seat search rather than a function build. Reason to verify before you write, not after.
- Tier: Healthcare/Regulated. Read the Healthcare/Regulated playbook before outreach.
- Rubric: Company 12/12. Role 15/15.
- Portfolio match: A clinician accepts, edits or rejects a suggested order or code before it enters the medical and billing record, under time pressure, with a license attached to the outcome. The failure mode is acceptance-by-default under load. Thermo Fisher's exception-first architecture gives you the triage logic, but the sharper opening is the core claim of "Trust Is the New Interface": one confident wrong answer does more damage than ten correct ones repair. In a clinical setting that stops being a rhetorical line and becomes the product's actual risk model. Differentiator: you have designed for expert reviewers working under audit conditions, and you are building the agents that generate what gets reviewed, in production, this quarter. Very few candidates have stood on both sides of that handoff, and fewer still are standing there right now.
- Verdict: Watch. The rubric clears Act at 12/12 and 15/15, and the 22–42 day band caps it. Justification for keeping it live: specialized clinical AI domain, genuinely thin candidate pool, plausible long fill cycle. Today-action: Verify. Verify this one first. Higher-scoring of the two, and it hits the 43-day cliff on August 6. Trigger to Ignore: no confirmation of an active search by then. Trigger to restore: recruiter confirmation, or a repost.
10. Maven Clinic. Virtual care for women's and family health, delivered through employer and health-plan benefits.
- Role: Senior Staff Product Designer, Care Delivery. Posted June 22. 40 days. Extended. Greenhouse
- Stage / capital: Late-stage private, benefits-distributed care. Raise total not verified this pass.
- Offices: Remote, US hubs. Range posted.
- Tier: Healthcare/Regulated. Read the Healthcare/Regulated playbook before outreach.
- Rubric: Company 8/12 (AI 1, Stage 2, Design influence 2, Trajectory 3). Role 12/15 (Comp 2, Scope 2, Craft 3, AI exposure 2, Portfolio 3).
- Portfolio match: Care Delivery is the coordination layer between patient, provider and care advocate. Different facet of the Red Cross work than the one Gusto wants: not the systems consolidation, the handoff itself. Three roles working a single shared record for the first time, 1,689 cases in two weeks, surge architecture that absorbed 10× volume with zero retraining. Care coordination breaks at handoff and at volume spikes. Same pair of failure conditions. Maven Intelligence is a real AI orchestration layer and the posting names uncertainty, human deferral and confidence indicators outright, so the AI work exists. What is thin is centrality. This is care coordination with AI inside it, not an AI company. That is why company lands at 8 and role at 12, under the numbers this role would carry on centrality alone.
- Verdict: Watch. Today-action: Verify. Read this one against the retirement below. Maven's VP of Design posting comes off the board this pass at six weeks, which means the reporting line above this role is either newly filled or abandoned, and which one it is changes what Senior Staff actually means here. Put that in the same message. The July 30 refresh is a system timestamp: Greenhouse tracks publish date and refresh date as separate fields and does not treat a refresh as hiring activity. Read it neutral, not as renewed intent. Trigger to Ignore: August 4, day 43. Trigger to restore: confirmed VP of Design hire plus an active search on this requisition.
Board health
Sub-48-hour listings: one, and it sits below your level. Gusto's Product Designer, HR Experiences, July 30. It counts anyway, as cluster evidence.
Design cluster check, and the answer turns on how strictly you read the rule. Loose definition, two or more design roles first published inside 14 days: four companies qualify.
| Company | Roles in cluster | Gap |
|---|---|---|
| Gusto | Head of Design (Jul 22), Product Designer (Jul 30) | 8 days |
| Vanta | Design Systems (Jun 26), Head of Design (Jul 6) | 10 days |
| Fetch | Director (Jul 1), Senior Growth (Jul 13) | 12 days |
| Maven | VP of Design (Jun 16), Design Lead (Jun 17), Senior Staff (Jun 22) | 6 days — now stale |
Strict reading, two IC product-design roles: null. Every one of those four pairs a leadership role with a single IC role.
Take the strict reading here. It is the more useful one. A leadership-plus-IC pair means the leadership search is already live rather than forming, and that is a different signal entirely. Gusto, Vanta and Fetch are hiring the leader now. The window is open, not approaching.
Retired this pass: Ramp, Capital One (two roles), Maven VP of Design. Six weeks or older, no counter-evidence.
No live employer posting recovered: Stripe Risk, HackerOne, Jasper. Absences, not closures. They come back if a direct requisition surfaces.
All ten URLs loaded this morning and all ten records returned as published. Dead links got killed before publication instead of flagged after. If one 404s by Monday, that is decay rather than an error, and you will not have spent a Saturday finding out.
- What a refreshed date proves: Greenhouse exposes
first_publishedandupdated_atas separate job-post properties, and its public documentation never defines an update as candidate review or renewed hiring intent — which is why four dates in this pool needed correcting. - Gusto's builder mandate: The company's account of turning designers into builders describes retiring static Figma files, teaching the pull-request lifecycle, and writing AI expectations into interview plans and career frameworks — read it before you write anything to gate two.
- Where the auditor still decides: Fieldguide's Field Orchestrator post spells out exactly where the agent pauses and where exception review and sign-off stay with the human, which is the design surface this Staff role owns.
- The Vanta buyer in his own words: Jeremy Epling's First Round interview from March is the closest thing to a public brief on how he thinks about product, design and engineering as one organization.