_private/qwestly-docs/Features/bite-sized-interview/gap-assessment.md
Table of Contents
Bite-Sized Interview — Gap Assessment
Last updated: June 5, 2026 15:30 ET
Gap assessment against the plan doc at candidate/docs/plan/feature/bite-sized-interview.md.
1. Architecture Shift: Agent vs. Scoped Prompts
The bite-sized interview will fold into qwestly-agent as a single conversation thread. System instructions swap per value prop rather than running five independent sessions. The scoped prompts themselves are absorbed as tools within the agent.
Resolved (Jun 5 sprint retro): Remove the collapsible sidebar. Use one conversation thread with swapped system instructions. A sticky top row (mobile) or full display (desktop) shows value prop navigation. Tapping a VP injects a transition message and swaps the system instruction. Vela will create a planning doc for the 5-prompt integration.
2. Right-Rail Artifacts Abandoned
The live-updating right-rail artifact panel is gone.
Impact: Epic 5 now over-scoped (includes artifact-panel.tsx). The plan needs updating to reflect chat-inline artifacts instead of a right-rail panel. The congratulatory page (Epic 7) still needs to show artifacts at the end.
3. Value Prop Transition UX
How does the user move between value props? Resolved at the sprint retro (Jun 5):
- Remove the collapsible sidebar entirely
- Single conversation thread, system instruction swaps per VP
- Sticky top row on mobile, full display on desktop for VP navigation
- User taps a value prop → injects "let's skip to X section" → system instruction swaps
- Tool call signals section completion to the frontend
David will implement the transition mechanics. Vela will build the evaluation system for conversation completion detection.
4. Artifact Spec Per VP — Partially Defined
Adam owns drafting concrete artifact shapes for each value prop. Current status per VP:
| VP | Artifact | Status |
|---|---|---|
| 1 — Your Story | LinkedIn about hook, resume rewrite, bio, Qwestly card narrative | LinkedIn about suggestion exists. Still needs clarification: are resume rewrite and bio in scope for v1? |
| 2 — Your Wins | Qwestly card narrative optimized for hiring manager opens | Qwestly card exists. Prompt produces card content. |
| 3 — Your Dream Job | Curated role feed signal, tighter matching, less noise | Undefined. What does "tighter matching, less noise" translate to in the UI? Needs definition. |
| 4 — Your Skills | AI readiness indicator on profile | No changes made to Qwestly card yet. Adam needs to confirm how the prompt / card should reflect AI readiness. |
| 5 — Your Interests | Peer intro pathways, topic-based matches | Scoped work exists for this. |
Reference: Adam's original value props & artifact table
Impact: VPs 1, 3, and 4 need clarification from Adam before artifact agent prompts can be finalized. VPs 2 and 5 are in better shape.
5. Post-Completion Dashboard
Adam prototyped a post-completion dashboard that the original bite-sized interview plan didn't account for. Achievable vs. net new breakdown:
Achievable with existing data
| Dashboard element | Status |
|---|---|
| Drafts (ready for review) | Have data — VP-generated documents exist |
| Intros queued this week | Have data — networking opportunities exist |
| Questions remaining / preferences | Have data — onboarding session tracks this |
| Empty/partial/complete states per row | Have data — per-VP completion status exists |
Achievable but will show zeros
| Dashboard element | Why zero |
|---|---|
| Roles serviced this month | Job matching is far from production-ready |
| Hiring manager intros | No real hiring managers on the platform yet |
Net new — requires separate scoping
| Dashboard element | What's needed |
|---|---|
| Interest-based peer matching | Unstructured matching capability — match candidates by industries/interests, not structured company data. Needs new service. Dominick plans to hand to Vela. |
| Document editor page | Draft artifacts need a document-first UI (Prose Mirror) with chat-assisted editing. Not a chat-first page. Look at ChatGPT Canvas mode patterns. David to explore UX. |
What Adam's prototype implies
- The post-completion flow now has a defined destination: a dashboard, not a simple congratulatory page
- The "drafts" column shows three categories (Your Opportunities, Hiring Manager Intros, Companies You're Interested In) — this is a richer view than the plan's single artifact-per-VP model
- Document editor is a distinct sub-feature that needs its own planning doc, separate from bite-sized interview
Impact: The congratulatory page in the plan (Epic 7: congratulations.tsx) is underspecified compared to what Adam prototyped. David should not build the simple version if the dashboard is the real target. Dashboard and document editor need separate planning docs — don't let them delay bite-sized interview, but align the handoff now so there's no rework.
6. Document Editor Page — Net New Feature
Owner: Dominick (initial planning), David (implementation)
Adam's prototype introduced a document editor page for reviewing draft artifacts. When a candidate clicks a "ready for review" draft on the dashboard, they land on a document-first page (not a chat page) where they can edit inline and use AI chat to refine the document.
Key design references: ChatGPT Canvas mode, Prose Mirror for rich text editing.
This feature was not accounted for in the bite-sized interview plan. It needs:
- A separate planning doc scoping the UI, the chat integration, and the backend for persisting edits
- Decision on whether it's part of v1 or post-launch
Dominick will do initial planning and scoping, then hand off to David for implementation.
7. Profile Grader — Decoupled from Bite-Sized Interview
The profile grader/scorer appeared in Adam's dashboard prototype but was always out of scope for bite-sized interview. It's been decoupled entirely — it will be built as a standalone feature and absorbed into qwestly-agent as a tool.
Tied to the unauthenticated onboarding initiative. No impact on the bite-sized interview timeline.
8. Public/Marketing Site Update
The public-site marketing page needs to be updated to reflect the new onboarding flow and value props. The current site was built before the bite-sized interview redesign — copy, screenshots, and the signup CTA flow all need alignment with the new experience.
Not accounted for in the bite-sized interview plan. The public site is a separate Next.js app (public-site) under the monorepo.
9. Eval/Completeness Model — Thresholds vs. Binary
Adam described a three-tier interview pyramid:
- Level 1: Minimum info for candidate summary
- Level 2: Sufficient for making matches
- Level 3: Enough data for consulting/advice
The plan's eval model is a simpler binary criteria_met: boolean per VP (Section 4, Section 6). The conversation reveals a more nuanced model that's not reflected in the plan's types:
// Plan: binary
criteria_met: boolean;
// Conversation: tiered
threshold: "minimal_summary" | "match_ready" | "consulting_ready";
Impact: The eval service (eval.service.ts) and EvalState type may need to model tiers, not just pass/fail. This affects when the system decides to move on vs. continue asking.
10. Push-to-Talk
The plan specifies push-to-talk in packages/ui as a reusable @qwestly/ui component. David built it there — it's in the qwestly-ui repo under packages/ui as a submodule. The component is done and working with ephemeral keys.
Impact: No gap — delivered as planned.
11. Mobile UX Gaps
Several mobile-specific issues emerged that the plan doesn't address:
- Sidebar VP navigation doesn't work on mobile (limited vertical space)
- Drag-to-rank on the ranking screen needs visual affordance on mobile
- The sticky header/question takes too much vertical space on mobile
- Max-width on mobile chat feels wide
The approach: remove the title/description header, integrate the first question into the chat stream.
Impact: The plan assumes a desktop-first design with a sidebar. Mobile needs a different navigation model (bottom sheet? inline transitions?). This should be designed before David builds the full sidebar.
12. Open Questions from the Plan
Five open questions from Section 10, with current status:
| # | Question | Status |
|---|---|---|
| Q1 | HM onboarding — what happens after "I am hiring"? | Resolved — "I am hiring" checkbox removed entirely; no HM branching flow in v1 |
| Q2 | What replaces the Qwestly card as the destination? | Resolved — users land on the new dashboard; Qwestly card is a draft item inside a "Your Presence" card alongside LinkedIn about and experience summary |
| Q3 | Card timing — immediate vs. background-build? | Resolved — generates in background; dashboard shows generating/ready status |
| Q4 | Skip mechanics — queue at end, drop, or stay clickable? | Unresolved |
| Q5 | Networking opt-in placement — where in the flow? | Leaning resolved — likely through dashboard integration with multiple entry points (settings, dashboard cards) |
Impact: Q4 affects the chat flow (skip behavior). Q1–Q3 resolved. Q5 leaning resolved through dashboard.
📋 Q4 Asana ticket · Q5 Asana ticket
13. Epic Ownership & Velocity Drift
The plan assigns clear epic ownership, but the conversations show drift:
| Epic | Plan Owner | Actual Work |
|---|---|---|
| E2 (prompts) | Vela | Delivered — prompt chat endpoint shipped Jun 4; David integrated same day |
| E4 (ranking) | David | Done (mobile fixes merged) |
| E5 (chat UI) | David + Dominick | David building most of it; Dominick on STT proxy |
| E7 (completion) | Dominick + David | Not started — blocked on Adam's dashboard/artifact decisions |
David is doing chat UI work (originally Dominick's in E5) and push-to-talk UI (originally split). Dominick is doing more cross-architecture decisions (agent vs. prompts, dashboard scoping) than planned.
Impact: Vela's prompt endpoint is now delivered and integrated. The prompt pipeline is partially unblocked — one VP is wired end-to-end. The remaining four VPs and artifact agent prompts still need completion, but the frontend-backend contract is now validated.
14. Sprint Retro Outcomes (Jun 5)
Key decisions and next steps from the full-team sprint retro:
Architecture & UX resolved
- Single chat thread with swapped system instructions per VP (see Section 1)
- Sticky top-row VP navigation on mobile, full display on desktop (see Section 3)
- Collapsible sidebar removed entirely
Scope prioritization
- Gap assessment covers all initiatives, not just bite-sized — not all in one sprint
- Dominick + Adam will do a first pass on V1 requirements and re-scope
- Some features may scale back in the V1 prioritization pass
- Working backwards from a 1000-candidate target
Current sprint focus
- Vela: Complete 5 value prop prompts; create integration planning doc for 5-prompt system; build evaluation system for completion detection
- David: Resolve UX transition and mobile gaps; continue chat interface polish
- Dominick: Work with Adam on scope prioritization; profile grader scoping (needs unauthenticated chat sessions + deterministic scoring rubric)
- Team: Considering daily standups for next few weeks to improve coordination
Summary: Top 7 Items
- Agent vs. scoped prompts — Resolved — single chat, swapped system instructions, sticky top-row nav (see Section 1)
- Artifact spec per VP — VPs 2 and 5 defined; VPs 1, 3, 4 need Adam's clarification before artifact prompts can be finalized
- Value prop transition UX — Resolved — sticky top-row nav, injects transition message, tool call signals completion (see Section 3)
- Open questions — Q4 (skip mechanics) unresolved; Q5 (networking opt-in) leaning resolved through dashboard
- Remaining VP prompts — one is done and integrated; four more to go, plus artifact agent prompts. Vela creating 5-prompt integration plan.
- Document editor page — net-new feature from Adam's prototype; needs its own planning doc. Dominick to scope, David to implement.
- Public/marketing site update — copy, screenshots, and signup CTA need alignment with the new onboarding flow