Design Context File v10 — Juno Chen

Version: 10 Date: October 4, 2026 Supersedes: v9, September 27 Primary change this cycle: Domain 1 artifact registered (Commitment Boundary Surface). Four new role signals processed. Per-company positioning revised — D1 now leads where the company's frontier challenge involves mid-turn correction or consequential-action policy.

Identity and Background

Name: Juno Chen Current title: Independent — Agentic Labs (solo-built applications exploring agent trust, delegation, and human oversight) Prior roles: Head of Design & Research, Alibaba.com North America · Product Design Director / GM, BCG Digital Ventures (multiple ventures) · Design roles at earlier-stage companies Portfolio: junochen.com Portfolio last verified: September 27, 2026 (pre-v9). A live fetch was not confirmed this cycle. Claims below about homepage content, TinyFish presentation, and Labs discoverability reflect that check. Re-verify before generating outreach that depends on current site structure. Core positioning claim: Designs the surfaces where humans review, correct, and authorize agent-produced work in high-stakes, regulated, or trust-dependent systems.

TinyFish Context

Hard boundary: past-tense role and technical credibility only.

Juno worked at TinyFish, an AI company, in a design capacity. This establishes proximity to production AI systems — model behavior, traces, auditability, reversibility, system constraints. TinyFish is not a case study, not a source of portfolio visuals, not a source of outcome metrics. The homepage's current presentation of TinyFish as selected work does not override this boundary.

Prohibited: Internal screens, architecture, pricing, product outcomes. "First designer" claims without independent evidence. Any invented or expanded departure narrative.

Sequencing rule: Published design evidence first. TinyFish as bounded practitioner context second. Return to the target role's design mandate.

Forward-Looking Artifact Inventory

Handoff name note: The Trust essay uses Intent-Setting, In-Progress, Output Review, Decision Gate, and Loop Feedback. This registry maps those to the canonical positioning terms: Delegation (← Intent-Setting), In-Progress, Decision Gate, Recovery (← Output Review), and Loop Feedback. When referencing the essay directly, use its own terminology.

D1 Artifact — What Changed

The Commitment Boundary Surface replaces the Inference-Aware Decision Surface as the D1 artifact. The prior concept addressed routing economics — model selection, cost, latency, verification burden (see Build Queue specs for specification history). The new concept addresses a different problem: the point where changing a user's instruction stops being cheap and starts requiring cancellation, compensation, or disclosure.

The organizing question: you changed the instruction — what changes now?

Before the boundary, the system absorbs the change at near-zero cost. After it, the system must cancel external commitments, disclose the change to counterparties, compensate affected parties, or accept irreversibility. The design problem is making this boundary visible in real time, before the user crosses it.

CarrierIQ connection: CarrierIQ's Approve Bind gate is a commitment boundary in its simplest form — before it, the user changes carriers, coverage, or terms freely; after it, the policy is bound and unwinding requires cancellation. The Commitment Boundary Surface extends this to systems where the boundary is a gradient: partial commitments, conditional holds, progressive lock-in that the user needs to see accumulating.

Commitment-state work previously lived in D4 or crossed all four domains. Registering it under D1 reflects a judgment that making the cost of correction visible is an interaction design problem first and a system-architecture problem second.

Agentic Labs Foundations

Approved for external use: Brand Pulse, Retail Velocity, Carrier IQ. Other apps (UAT Sentinel, Creator Scout) may be publicly reachable but are not approved for positioning.

Case Study Registry

All outcomes below are Juno-published portfolio claims, not independently verified. Numbering follows v9. The live site uses different identifiers for CS-04, CS-05, and CS-06 — use project names and URLs in external materials until the site and this file agree.

Cummins / ZED Connect: Supporting industrial-IoT evidence, Product Design Director at BCG Digital Ventures. No public case page or approved metrics. Use only when industrial, fleet, logistics, or connected-vehicle relevance is high.

Metric Cautions

• Alibaba: Use case-page figure 25M+ desktop sessions, not homepage 40M buyers (different denominators).
• Thermo Fisher: Use $20M margin opportunity, not $20M+ margin recovered annually.
• Alle: The figure 47% appears for both lapsed-member reactivation and expiry-notification open rate. Do not present as unambiguous. The 68% figure (members citing expiry notification as return reason) is separately recorded.
• Equinox+: 4.8 is launch-state, not current.
• Carrier IQ: Timing claims ("five carriers in 90 seconds," "eight portals in four minutes") are internally inconsistent. Do not use.

Trust Essay and Five Handoffs

Published at junochen.com/trust-is-the-new-interface. Confirmed reachable October 3, 2026. Re-verify before outreach.

The essay defines five handoffs in human-agent delegation using the labels Intent-Setting, In-Progress, Output Review, Decision Gate, and Loop Feedback. The table below maps these to canonical positioning terms used elsewhere in this file.

Per-Company Positioning

Act Tier

Act 1: Stripe — Senior Staff Product Designer, Data & AI

Lead with Commitment Boundary Surface. Stripe's Data & AI surfaces involve routing decisions with financial consequences — fraud detection, risk scoring, merchant review. The commitment boundary question applies directly: when does changing a parameter stop being a settings adjustment and start requiring disclosure or compensation?

Support with CarrierIQ (staged review before irreversible action), Alibaba (B2B platform scale, procurement trust), Thermo Fisher (BCG DV, multi-organization operations with human release gates).

Do not assume the Data & AI role owns inference routing because of the OpenRouter acquisition (established in v9 company signal; confirm current status before referencing in outreach). Read the posting before mapping the company story.

Act 2: Anthropic — Product Designer, Safeguards

Lead with CarrierIQ's five handoffs, approval gate, and regulated review structure. The Commitment Boundary Surface is relevant but secondary — Safeguards is about preventing harm, which maps to Decision Gate and Recovery more than to cost-of-correction visibility.

Support with Red Cross (BCG DV, mission-critical authority), Thermo Fisher (BCG DV, regulated exception handling), Trust essay.

Qualification note: Production-quality Python is an entry qualification at Evals, not at Safeguards. Front-end coding at Safeguards is a preference, not a gate.

Act 3: Anthropic — Product Designer, Design Systems

Lead with Commitment Boundary Surface. A design system for AI-native products has to encode commitment states — which components signal reversible actions, which signal irreversible ones, which show the boundary approaching. This is D1 work applied at the systems level.

Support with Equinox+ (BCG DV, multi-brand platform architecture, design-system-adjacent unification), Alle (BCG DV, dual-surface consistency), CarrierIQ (the interaction patterns a commitment-aware design system would need to support).

Act 4: Anthropic — Product Designer, Developer Platform (Claude)

Lead with Commitment Boundary Surface. The developer platform is where external builders encounter agent delegation, and where the cost of mid-turn correction becomes a developer-experience problem. The gap between infrastructure capability and the user-facing surface's ability to explain it is the core design challenge.

Support with CarrierIQ (agent-readable instruction provenance), Trust essay (delegation framework), Alibaba (enterprise-scale platform design).

Act 5: OpenAI — Product Design Lead, Growth — Codex

Growth role. No frontier artifact directly addresses adoption and conversion. Alle (BCG DV) leads on growth outcome proof (3.2x redemption, CAC reduction). The prior Codex positioning led with Alle and used CarrierIQ as the agent-trust bridge; that structure holds.

D1 bridge: Codex adoption depends on users trusting agent-produced code enough to commit it. The commitment boundary — when does accepting a suggestion stop being reversible? — is the trust problem that underlies the growth problem.

Support with CarrierIQ (agent-trust proof), Trust essay, Equinox+ (BCG DV, consumer product craft, launch velocity).

Watch Tier

Watch 1: Figma — Director, Product Design, AI & Context

Why Watch: Director-level role requires org leadership and team-building positioning alongside craft. Posting scope and reporting structure not yet confirmed via Figma's own careers page; the detailed description comes from an investor-operated job mirror.

D1 relevance is high. Figma's AI features involve context-aware suggestions that accumulate into design decisions. The commitment boundary maps here: when does accepting an AI suggestion stop being a nudge and start being a structural choice that's expensive to reverse?

Promote to Act if the posting confirms a craft-heavy mandate with limited team-building scope, or if adjacent postings reveal an IC-heavy org structure.

Watch 2: OpenAI — Product Designer, Identity

Carried from v9 Act 3. Moved to Watch: no fresh posting signal since v9.

When activated, lead with Claims Ledger concept mapped to evidence binding and the Authorized/Committed/Settled distinction. CarrierIQ's Approve Bind gate as inspectable proof. Red Cross (BCG DV) for role-specific authority and payment controls.

Watch 3: Anthropic — Product Designer, Evals & Prompts

Carried from v9. Remains Watch: production-quality Python is an entry qualification, not a portfolio gap. Do not invest positioning effort unless that qualification boundary is cleared.

Positioning Rules

1. TinyFish: Past tense, role context, technical credibility. Not portfolio proof. The homepage's presentation does not override this boundary.
2. BCG Digital Ventures: CS-02 through CS-05 and Cummins/ZED Connect carry mandatory BCG DV attribution in every reference.
3. Artifact status: The Commitment Boundary Surface, Inference-Aware Decision Surface, Claims Ledger, Correction Lineage, and Delegation Contract must not be described as built, in progress, shipped, or published.
4. Retail Velocity: Visibly credits TinyFish components. Do not describe as independent of TinyFish.
5. Portfolio discoverability: Labs and individual case studies may not be findable from the homepage without direct links. Every outreach must include direct application URLs.
6. Agentic Labs approved set: Brand Pulse, Retail Velocity, Carrier IQ only.
7. Trust essay: Link directly to junochen.com/trust-is-the-new-interface. Re-verify URL before outreach.
8. General: Do not fabricate hiring managers, reporting lines, or internal company decisions. Do not describe any target company as "design-led." Do not treat a role signal as proof of exercised authority, reporting structure, interview weighting, or accessible slate.

Anticipated Interview Questions This Cycle

Q: How would you design for mid-turn correction in an agent system?

Ground in CarrierIQ. The user reviews parallel carrier quotes before authorizing a bind. Before the bind, correction is free — change carriers, adjust coverage, restart. After the bind, correction requires cancellation and may involve fees or coverage gaps. The design problem is making that boundary visible before the user crosses it. Reference the Commitment Boundary Surface concept and state its status accurately: specified, not shipped.

Q: What's different about designing for AI systems versus traditional product design?

Do not answer abstractly. Ground in production. Alibaba's procurement trust required making verification status visible across 200K+ suppliers — not just whether a supplier was verified, but what that verification actually covered. Thermo Fisher's exception-first triage (BCG DV) required designing for the failure case before the happy path. CarrierIQ requires showing the user what the agent did and why before asking for authorization. The thread across all three: the system's work product needs to be inspectable, not just its output.

Q: How do you think about inference costs affecting UX decisions?

The prior D1 concept addressed this directly (see Build Queue specs): the user specifies consequences — cost ceiling, acceptable delay, minimum reliability — while the system handles model parameters. The Commitment Boundary Surface extends this. Inference cost is one dimension of the commitment boundary, alongside external obligation, disclosure requirement, and reversibility window.

Q: You have a lot of specification work that hasn't shipped. Why?

Acknowledge directly. Eight-plus specification cycles without a shipped artifact is a real cost. The production proof is in CS-01 through CS-06 and the three approved Agentic Labs apps. The specifications represent where the thinking is headed. The case studies and Labs prove the ability to ship.

Prohibited Content and Phrasing

• Forward-looking artifacts described as built, in progress, shipped, or published
• Carrier IQ timing claims ("five carriers in 90 seconds," "eight portals in four minutes")
• Alibaba's 40M buyers substituted for 25M+ desktop sessions
• Alle's 47% presented as unambiguous reactivation evidence
• Portfolio outcomes presented as independently verified
• Cummins/ZED Connect treated as a full case study or source of metrics
• Claims of recent production-code authorship, named prototyping-tool proficiency, LLM-evaluation-pipeline experience, production Python, or security-operations experience without new evidence
• TinyFish internal screens, architecture, pricing, traces, or product outcomes
• "First designer" at TinyFish without evidence
• Invented or expanded TinyFish departure narrative
• Retail Velocity described as independent of TinyFish
• Any target company described as "design-led"
• Implication that Labs are discoverable from the homepage
• Fabricated hiring managers, reporting lines, or internal company decisions
• Role signal treated as proof of exercised authority, reporting structure, interview weighting, or accessible slate