diff --git a/blume.config.ts b/blume.config.ts
new file mode 100644
index 0000000..dde7be7
--- /dev/null
+++ b/blume.config.ts
@@ -0,0 +1,6 @@
+import { defineConfig } from "blume";
+
+export default defineConfig({
+ title: "Jobyssey",
+ description: "Documentation powered by Blume.",
+});
diff --git a/docs/hanover-park/index.mdx b/docs/hanover-park/index.mdx
new file mode 100644
index 0000000..0ce57ce
--- /dev/null
+++ b/docs/hanover-park/index.mdx
@@ -0,0 +1,77 @@
+---
+title: Hanover Park
+description: Software Engineer (Fullstack). Where the loop stands, who is in it, and what happens Wednesday.
+sidebar:
+ label: Loop overview
+ icon: calendar
+---
+
+Inbound loop. Chris Hladczuk (co-founder/CEO) cold-emailed on 2026-08-20; every step since has been driven by their side. Next step is the in-person system design with Nick Puljic (co-founder/CTO).
+
+:::warning[Wednesday 2026-09-09, 3:30pm, in person]
+115 East 23rd St, 8th floor. Arrive 3:20 to 3:25, no earlier. Tell Gabe (office manager) you are there for an interview with Nick; Nick collects you. Dave is remote that day; La Rosa is onsite. About one hour: small talk, then one system design problem. Notepad available; whiteboarding optional.
+:::
+
+
+
+ Who he is, what he built, how he grades.
+
+
+ Where his path and mine overlap: YC, cofounding, fintech 0 to 1, leaving a company.
+
+
+ Founder-to-founder and technical questions, with what each one is for.
+
+
+ What Dave and John said the interview is, and what passes.
+
+
+
+## Process
+
+
+
+ Done. Comp positioning set: my ask 300 to 350k total, their stated floor "north of 200k" driven by leveling.
+
+
+ Done. La Rosa scorecard: extremely strong technical ability; product sense assumed from background but not demonstrated in the room.
+
+
+ Done. Logistics, rubric, and the read on Nick. See the system design page.
+
+
+ Next. Full-stack assessment, creativity weighted above correctness.
+
+
+ Following week (company WFH Thu/Fri for the office move; new office from Monday). Real tickets in a demo environment under NDA; ship to prod if possible, short demo at end of day. Graded on communication and the final demo.
+
+
+ No LeetCode at any stage.
+
+
+
+## People in the loop
+
+| Person | Role | Notes |
+| --- | --- | --- |
+| Chris Hladczuk | Co-founder, CEO | Public voice. Sold everything organic until the first two AEs signed this month. Hands-on in recruiting. |
+| Nick Puljic | Co-founder, CTO | Wednesday's interviewer. Low public surface area, deeply technical. |
+| David "Dave" May | Founding Talent Lead | 18 years recruiting (Amex, Bloomberg, ZocDoc, Chainalysis, Coalition). Six months in; hire #25. |
+| John | Engineer since ~5 people | Barcelona; runs the fund-admin migration pipeline (4 to 6 months of onboarding down to 4 to 6 hours). |
+| La Rosa | Engineer | Ran the technical screen; onsite Wednesday. |
+| Gabe | Office manager | First person you meet on the 8th floor. |
+
+## Company snapshot
+
+- AI-native fund administration for VC and PE: general ledger, waterfall engine, investor portal, portfolio management, rebuilt from scratch rather than wrapping QuickBooks or Carta.
+- Operating model: agents prepare (parse emails and documents, propose journal entries, categorize ~90% of cash transactions), CPAs approve. Zero-fault-tolerance domain.
+- 27M Series A led by Emergence with Lux and Susa (Mar 2026). AUA figures vary by source: 15B in the March press, 25B+ in the May JD, 30B+ on the careers page. Say "north of 25 to 30B" and let him correct.
+- Pricing is basis points on AUM, not seats. They refuse funds under ~100M and refuse to unbundle.
+- Engineering: ~20 engineers, 35 to 40% ex-founders, no PMs, no designers, no QA. Five days in office, Gramercy. Scaling 50 to 100 people by year end.
+- 409A at 175M, which Dave flagged as a discount against actual growth.
+
+## Watch items
+
+- Relocation from Virginia to NYC; Dave: "cross that bridge when we get there."
+- Comp gap between my 300k+ ask and their stated 200k+ floor. Final number is leveling plus interview performance, so Wednesday moves the number.
+- Work-trial day is a real time cost; treat the demo as release quality.
diff --git a/docs/hanover-park/meta.ts b/docs/hanover-park/meta.ts
new file mode 100644
index 0000000..eed9d43
--- /dev/null
+++ b/docs/hanover-park/meta.ts
@@ -0,0 +1,7 @@
+import { defineMeta } from "blume";
+
+export default defineMeta({
+ title: "Hanover Park",
+ icon: "landmark",
+ pages: ["index", "nick-puljic", "shared-ground", "questions", "system-design"],
+});
diff --git a/docs/hanover-park/nick-puljic.mdx b/docs/hanover-park/nick-puljic.mdx
new file mode 100644
index 0000000..ef466ba
--- /dev/null
+++ b/docs/hanover-park/nick-puljic.mdx
@@ -0,0 +1,94 @@
+---
+title: Nick Puljic
+description: Essential brief on Nick Puljic, co-founder and CTO of Hanover Park. Identity, career, what he built, and how he is likely to grade.
+sidebar:
+ icon: user
+---
+
+Two-time fintech founder, low public surface area, personally built a double-entry general ledger. Treat him as a peer who happens to be interviewing you, not as a gatekeeper.
+
+## Identity
+
+| | |
+| --- | --- |
+| Full name | Nicholas "Nick" Puljic |
+| Role | Co-founder and CTO, Hanover Park (Feb 2024 to present), NYC |
+| Education | Columbia University, BS Computer Science, 2015 to 2019 |
+| LinkedIn | https://www.linkedin.com/in/nicholas-puljic (~2.4k followers) |
+| GitHub | github.com/NickPuljic: account since 2016, zero public repos, zero gists, no bio |
+| Press | Forbes 30 Under 30 with Chris Hladczuk |
+
+:::danger[Do not reference open-source work]
+He has a GitHub with nothing on it. Dave's guess ("I don't even imagine he has a GitHub") was right in substance. No blog, no podcast appearances, no talks found. Chris is the public voice; Chris's June 2026 post says it took a film crew showing up at the office to get Nick on camera. Flatter the work, not the personal brand.
+:::
+
+## Career
+
+
+
+ Big-company baseline. Dave described it as an internship.
+
+
+ Regulated healthcare at scale. Likely where he met Jake Ruth, his future co-founder.
+
+
+ His 0 to 1. Retail-investor education and equity analysis web app. Co-founders Jake Ruth (CEO, ex-Oscar senior SWE) and Daniel Pronk (CFO, finance YouTuber, ~192k subs, brought the Discord audience). Raised ~1.3M seed including 500k from YC. Priced 6.99/month or 74.99/year, 7-day trial, no credit card. Launched 2021-12-11 after a four-month ~200-user beta from Pronk's Discord. Originally "Stonk Reporter." Stack, from Jake's HN launch thread: fully serverless AWS, Cognito auth, API Gateway plus Lambda (Python), DynamoDB plus Postgres, S3/CloudFront/Route 53, Finnhub market data. Nick exited June 2023; the company continues under Jake with iOS/Android apps and 170k+ tickers. He is no longer listed among YC's active founders for it.
+
+
+ See below.
+
+
+
+Aggregators also list Remedy Partners and Nannytucket earlier; unverified, do not cite.
+
+## What he built at Hanover Park
+
+His stated thesis: fund administration was "ripe for the future of Services as a Software." Per Emergence: "Co-founders Chris Hladczuk and Nick Puljic built the general ledger, the waterfall engine, the investor portal, and the portfolio management layer from scratch... you're designing the plumbing around intelligence."
+
+- Traction: 1B to 15B AUA in under 12 months; 25B+ per the May JD; 30B+ per the careers page. 27M Series A led by Emergence with Lux and Susa; 31M total raised. Angels are fund CFOs at Bain Capital, Clearlake, CD&R, Hellman and Friedman.
+- Shipped notables: first MCP server in the space (fund data queried from Claude or ChatGPT), zero-fee automated payments, Ramp and Brex expense sync, the Asylum Ventures migration (30+ entities and a decade of history in ~100 hours).
+- Positioning line: "Stripe for payments. Ramp for expenses. Hanover Park for investments."
+
+### Stack for the role
+
+| Layer | Choice | Source |
+| --- | --- | --- |
+| Frontend and API | TypeScript, Next.js on Vercel, tRPC between browser and app | JD, John |
+| Database and auth | Supabase / Postgres | JD, John |
+| Background jobs and AI inference | Trigger.dev (wrapper over EC2) | JD, John |
+| Long-running agent jobs | Vercel Sandbox via the Vercel Harness Agent (wraps Claude Code / Codex / Anthropic, normalizes messages for storage and chat display) | John |
+| Serverless backend (earlier posting) | SST on AWS | MTS JD |
+| Experiments | Eve (John likes the DX), jcode | John |
+| AI budget | Effectively unlimited; ~800/month company card plus more on request | John |
+
+### Culture, per the JD
+
+"Engineering-first." "No bureaucracy. No hierarchy." No QA team, no PM writing specs, no design handoff. You ship it, you own it, and make "architecture and technical decisions alongside the CTO." Required: strong full-stack fundamentals, production LLM/AI workflow experience, high ownership. Values: Egoless Execution, Maniacal Urgency, Own the Details. The Design Engineer hire "will work directly with Nick."
+
+## How he will grade
+
+
+
+ Stock Unlock was Lambda plus Dynamo plus Cognito. Hanover Park is Supabase plus SST/Trigger.dev on AWS. Hand-rolled Kubernetes, Kafka, or custom orchestration will read as ceremony. Pick boring managed primitives, defend them on shipping speed and operational headcount, then name the point where you would outgrow them.
+
+
+ Correctness primitives are home turf and transfer to almost any "rebuild X" prompt: immutable append-only journal, derived read models, idempotency keys on ingestion, exactly-once effects on money movement, reconciliation loops, full audit trail, replay and backfill for corrections. "We never mutate history, we post a reversing entry" is his own system reflected back.
+
+
+ Stock Unlock was a dense consumer data-viz web app; Hanover Park hired a design engineer for "interaction patterns for workflows where humans and AI agents work side by side." Design the review surface: how a human sees agent output, its confidence, the diff, and corrects it. Dense tables, optimistic UI, perceived latency, empty and error states.
+
+
+ Extraction from unstructured input, model proposes, human approves, system commits. Confidence thresholds routing to review queues. Evals on real edge cases. Cost and latency per document. This is his product thesis and the cheapest route to the "creativity, not generic" bar.
+
+
+ Trade-off narration is the signal: what you cut in v1, what you would only build at 10x, what you buy instead of build, what ships in a week versus a quarter. Ship then iterate is house doctrine, paired with Own the Details on anything touching money.
+
+
+
+## Temperament, per Dave
+
+Passionate, deeply technical, gets animated when the problem is interesting. Engage him, do not lecture. He appreciates good questions and dislikes softballs. A correct-but-generic answer from someone with my background reads as underperformance ("with his background, I expected a lot more").
+
+## Sources and caveats
+
+Compiled 2026-09-03 from LinkedIn search-index snapshots (the profile itself returns HTTP 999), YC, the Ashby job board, the GitHub API, Emergence Capital, Euclid/Verticalist, and press. Employment history and Columbia CS corroborated across three independent sources. Dave's prep-call account (2026-09-03) and John's screen (2026-09-02) supply the stack and temperament details.
diff --git a/docs/hanover-park/questions.mdx b/docs/hanover-park/questions.mdx
new file mode 100644
index 0000000..1380a57
--- /dev/null
+++ b/docs/hanover-park/questions.mdx
@@ -0,0 +1,100 @@
+---
+title: Questions to ask
+description: Questions for Nick, grouped by when to use them, with what each one is for and what a good answer sounds like.
+sidebar:
+ icon: message-circle
+---
+
+Dave's standard: no softballs. Every question below is either something only Nick can answer or something that makes him think. Pick four to six; the small-talk ones go first, the closer goes last.
+
+:::info[The one Dave loved]
+"What are you continuously looking to preserve, and what are you totally open to changing?" Dave's own answer: preserve velocity plus accuracy without trading quality, the talent bar, the product; open to changing everything, per Egoless Execution. Ask Nick the same and compare answers.
+:::
+
+## Small talk (first five minutes)
+
+
+
+ **For:** founder-to-founder opener. My side in one line: interviewed on the pre-Sonr prototype, did not get in, raised anyway.
+ **Good answer:** specifics about office hours, the Demo Day forcing function, or investor access, rather than "the network."
+
+
+ **For:** shows I read the launch history. Invites a real opinion about forcing functions.
+ **Follow-up:** whether he and Chris considered YC for Hanover Park or went straight to Emergence.
+
+
+ **For:** the cofounding conversation. My contrast: solo for five years, and what I want now is the engineering seat, not the fundraising one.
+ **Good answer:** how he and Chris divide decisions in a company with no PMs.
+
+
+ **For:** signals self-awareness about my own transition and gives me his checklist.
+ **Good answer:** concrete failure modes (over-scoping, reluctance to be told no, building for a customer that is not theirs).
+
+
+
+## Technical (during or after the design)
+
+
+
+ **For:** his home turf. Sets up the reversing-entry conversation.
+ **Bridge:** on-chain state is append-only by construction; the interesting part is the derived read models and how corrections are posted.
+
+
+ **For:** the human-in-the-loop thesis, which is also the differentiator in the design.
+ **Good answer:** thresholds, review queues, evals on real edge cases, cost per document.
+
+
+ **For:** shows I know the stack from two independent sources (JD and John).
+ **Bridge:** I ran Durable Objects for the same class of problem; curious where the wrapper-over-EC2 model bites.
+
+
+ **For:** the Asylum migration is the achievement to compliment. John owns the pipeline; Nick owns the correctness proof.
+ **Bridge:** the Cosmos to Base migration was the same problem with a smaller ledger and no CPA.
+
+
+ **For:** the exit-ramp question. I agree with the managed default; I want his outgrow point.
+ **Watch:** do not propose Kubernetes. Let him say where the seam is.
+
+
+ **For:** authorization is my domain (UCAN, capability-based auth). One question, not a lecture.
+
+
+ **For:** the founder-hindsight question. Every CTO has an answer and enjoys giving it.
+
+
+
+## Role and team
+
+
+
+ **For:** ownership model in practice. Also tells me what the work trial will grade.
+
+
+ **For:** references a specific conversation with his engineer; asks for the roadmap without asking for the roadmap.
+
+
+ **For:** signals I read the other JD and care about the front-end half, which La Rosa said I under-sold.
+
+
+
+## Closer
+
+"What are you continuously looking to preserve, and what are you totally open to changing?"
+
+Then, if there is time: "What would make you say no to me after today?" Direct, and it gives him the opening to name the gap while I can still address it.
+
+## The question he will ask me
+
+"Why are we still interviewing you?" The real question is why Hanover Park. Not "Sonr is sunsetting and you are a rocket ship." The answer, in order:
+
+
+
+ Money movement that has to be right, with an AI layer that prepares and a human that approves. I spent five years on the version of that where the human is a wallet owner and the mistake is irreversible.
+
+
+ Chris cold-inbounded me. Dave, La Rosa, and John each described the same company: no PMs, engineers own a metric back to front, 40% ex-founders. Every conversation confirmed the last one.
+
+
+ Rebuilt ledger, waterfall engine, the Asylum migration, first MCP server in the space. I want to build the next one of those, not fundraise for it.
+
+
diff --git a/docs/hanover-park/shared-ground.mdx b/docs/hanover-park/shared-ground.mdx
new file mode 100644
index 0000000..c341d05
--- /dev/null
+++ b/docs/hanover-park/shared-ground.mdx
@@ -0,0 +1,83 @@
+---
+title: Shared ground
+description: Where Nick's path and mine overlap, and how to use each overlap in the room without turning it into a comparison.
+sidebar:
+ icon: users
+---
+
+The rapport plan. Every row is something both of us have lived through, with the honest version of my side and the angle that turns it into a conversation rather than a pitch.
+
+:::tip[The rule]
+Peer, not applicant. He raised 1.3M, I raised 4.7M; he got into YC, I did not. Neither number is the point. Trade on the shared experience, and never lead with the comparison.
+:::
+
+## Side by side
+
+| | Nick | Me |
+| --- | --- | --- |
+| Degree | Columbia CS, 2019 | VCU Applied Math, 2020 |
+| First real job | Amazon, then Oscar Health in hypergrowth | Leidos, CBP Mobile Passport (federal, regulated) |
+| Founder solves own problem | Retail-investor tooling from Pronk's Discord audience | Wallet hacked, 15k lost, built a seedphrase-less wallet |
+| Y Combinator | W22 with Stock Unlock, 500k from YC | Interviewed for the pre-incorporation Sonr prototype (2020/21), not accepted |
+| Founding structure | Co-founder/CTO twice (Jake Ruth and Daniel Pronk; Chris Hladczuk) | Solo founder, CEO and de facto CTO for five years |
+| Team | Small team, 3 co-founders at Stock Unlock; ~20 engineers now | 7 FTE plus two agencies; 20+ engineers at peak, all reporting to me |
+| Money movement | General ledger, waterfall engine, zero-fee payments | USDC settlement, MPC custody, passkey wallets, on-chain fee handling |
+| Migration under correctness constraints | Asylum Ventures: 30+ entities, a decade of books, ~100 hours | Cosmos L1 to Base L2 migration of live state; 120+ developers onboarded |
+| Infra posture | Serverless by default (Lambda/Dynamo, then Supabase/Trigger.dev) | Both: Cloudflare Workers and Durable Objects, AWS Lambda, and Kubernetes validator ops at 99% uptime |
+| Leaving a company | Exited Stock Unlock June 2023; company continued under Jake | Sunsetting Sonr now, investors informed, wrap-up in weeks |
+| Public surface | Almost none; Chris is the voice | Fast Company byline, MIT Sloan guest lecture, top-5 global GitHub contributions in 2021/22 |
+
+## Y Combinator
+
+**His side.** Stock Unlock was W22. YC put in 500k of a ~1.3M seed. The product launched December 2021 off a four-month beta, so YC overlapped with the first months post-launch. He left 18 months after Demo Day.
+
+**My side, said plainly.** "We interviewed with YC on the file-sharing prototype before Sonr was incorporated. Didn't get in. Raised 4.7M anyway from Torch, SHMA, Bag Capital, and Builder." Do not soften it and do not dwell on it. The interview-but-no-invite is a normal founder story and he will have friends with the same one.
+
+**Why it transfers.** He has seen the inside of the batch; I have seen the outside. Ask what YC actually changed for a subscription consumer product, whether the Demo Day forcing function helped or hurt, and whether any of it carried into Hanover Park's raise. Hanover Park is not listed as a YC company in the notes; the ~40% ex-founder engineering team includes ex-YC founders. Ask rather than assume.
+
+## Cofounding versus solo
+
+**His side.** Twice a CTO with a business-side co-founder who owns the outside: Jake and Daniel at Stock Unlock, Chris at Hanover Park. He has never been the public face and appears to prefer it.
+
+**My side.** Solo for five years: fundraising, customer discovery, team direction, and architecture all on one desk. The honest lesson from Dave's screen: I want a seat at the engineering table, not the fundraising one.
+
+**Why it transfers.** He chose his structure; I am choosing mine now. The interesting question is how he and Chris split decisions in an org with no PMs, and what a founder-turned-IC gets right and wrong in their first six months on his team. Forty percent of his engineers have made that exact transition.
+
+## Leaving well
+
+**His side.** Left Stock Unlock in June 2023 with the company still operating. Whatever the reason, he handed off a live product with paying subscribers.
+
+**My side.** Sonr wind-down: investors informed last month, landing page plus ops cleanup, two to three weeks. Sunsetting on good terms; never found product-market fit.
+
+**Why it transfers.** Both are stories about the end of a company told without bitterness. If it comes up, one sentence each and move on. The framing that landed with Dave: "not that I didn't want to stop, it was at the point where I had to move on."
+
+## Money that must be right
+
+**His side.** Double-entry ledger, waterfall engine, capital calls and distributions for multi-billion-dollar funds; one hallucinated journal entry can cost a 1B client.
+
+**My side.** Irreversible on-chain operations, USDC settlement, MPC key shares where a bug loses funds permanently, a wallet-generation incident fixed the same week (40 seconds down to ~400ms by ripping out protobufs).
+
+**Why it transfers.** Same shape of problem: adversarial or irreversible money movement, idempotency, audit trails, reconciliation. This is the strongest bridge from my background into the system design, and the ground for the ledger questions.
+
+## Serverless pragmatism, with one difference
+
+**His side.** Lambda plus Dynamo plus Cognito at Stock Unlock; Supabase plus Trigger.dev plus SST now. Managed primitives, small team, ship fast.
+
+**My side.** Cloudflare Workers and Durable Objects for the Motr enclave, Lambda at Unissant, and the part he has not done: running four geographically distributed validator nodes on Kubernetes with KMS-secured keys.
+
+**Why it transfers.** I can agree with the default and speak credibly to the exit ramp. In the design, pick managed primitives, then name the specific point where Supabase or Trigger.dev stops scaling and what replaces it. Do not propose Kubernetes up front.
+
+## Builder who lets the product speak
+
+**His side.** No talks, no blog, empty GitHub, avoids cameras. Forbes 30 Under 30 is the only self-promotion on record.
+
+**My side.** The opposite public record, but the thing I am proudest of is the same kind of thing: a 700-commit chain repo and a sole-author crypto library, not the byline.
+
+**Why it transfers.** Compliment the ledger and the Asylum migration by name, not the press. Do not mention my GitHub ranking unless asked.
+
+## What not to do
+
+- No comparison of raise sizes, batch status, or press.
+- No "I could have done YC" framing; the honest version is shorter and better.
+- No lecturing on blockchain; translate every Sonr example into ledger, settlement, or custody terms.
+- No softballs about culture. The questions page has the real ones.
diff --git a/docs/hanover-park/system-design.mdx b/docs/hanover-park/system-design.mdx
new file mode 100644
index 0000000..bc244a4
--- /dev/null
+++ b/docs/hanover-park/system-design.mdx
@@ -0,0 +1,85 @@
+---
+title: System design rubric
+description: What the Wednesday interview is, what Nick weights, the La Rosa feedback to correct, and a pacing plan for the hour.
+sidebar:
+ icon: pencil-ruler
+---
+
+Assembled from Dave's prep call (2026-09-03) and John's screen (2026-09-02). Nothing here is guessed.
+
+## Format
+
+- One hour, in person, one problem: build or rebuild a system.
+- Dave would not name it, but: "most human beings know what it is," not necessarily something engineers have built. No curveballs, not Twitter or Spotify. Expect a lightbulb moment on hearing it.
+- John: product-oriented, conversational, focused on user flows rather than fine-grained architecture. Not traditional system design.
+- Notepad available; may be pure verbal articulation.
+
+:::danger[Full-stack assessment]
+An exquisite scalable backend with no front-end thinking is a fail. Dave: "I can guarantee you." Design the screen someone actually uses.
+:::
+
+## What Nick weights
+
+
+
+ A correct-but-generic solution reads as underperformance for my background. Push past the safe design in at least one place, then explain why.
+
+
+ Who are the users. What are we building with. Why: teardown or rebuild. What are we improving. What are we optimizing for: speed, cost, reliability. Ask until genuinely ready.
+
+
+ Time versus polish. Ship-and-iterate versus wait-for-perfect. What I cut in v1, what I would only build at 10x, what I would buy instead of build. Junior candidates skip this entirely.
+
+
+ Say what I am doing and why, out loud, continuously. Communication is the deciding variable, not raw ability.
+
+
+ Five years building Sonr, raised the money, fell on my face, re-approached. Use it explicitly, as design intuition rather than biography.
+
+
+
+## Pacing
+
+Do not rush to finish; do not rabbit-hole on one creative area. 98% of the way with strong signal beats a rushed complete answer.
+
+| Minutes | Do |
+| --- | --- |
+| 0 to 5 | Small talk. One founder-to-founder question from the questions page. |
+| 5 to 15 | Scoping. Users, constraints, what "better" means, what is in and out of v1. |
+| 15 to 35 | Core design. Data model and correctness primitives first, then the main user flow end to end. |
+| 35 to 50 | The creative move, the front-end review surface, and the AI layer if the prompt admits it. Name the trade-offs. |
+| 50 to 60 | What I would cut, what breaks at 10x, what I would build next. Then my questions. |
+
+## The default toolkit
+
+Managed primitives, then the exit ramp.
+
+
+
+ Postgres as system of record. Append-only journal or event log where anything is money or history. Derived read models. Idempotency keys on every ingestion path. Reconciliation loop. Background jobs on a managed queue (Trigger.dev shape). Name the point where Postgres or the queue stops scaling and what replaces it.
+
+
+ The review surface: what a human sees, what the system is confident about, the diff, and the correction path. Dense tables, optimistic UI, perceived latency, empty and error states. Say who the user is at each screen.
+
+
+ Model proposes, human approves, system commits. Confidence thresholds route to review queues. Evals on real edge cases. Cost and latency per item. Only if the prompt admits it; do not force it in.
+
+
+
+## Correcting the La Rosa feedback
+
+Scorecard: extremely strong technical ability; "you could see this guy knows how to build." Gap: likely strong product sense but did not demonstrate it. Credit came from the resume, not the room. My own diagnosis: with John I over-indexed on proving engineering chops because the resume already said founder.
+
+Fix for Wednesday:
+
+- Lead with the user and the flow, then the system. Product sense is the thing to prove; engineering is already believed.
+- Rehearse out loud beforehand. Dave's literal advice: in front of a mirror.
+- Translate every Sonr example into the prompt's domain. No blockchain vocabulary unless he asks.
+
+## Bar for passing, asked directly
+
+Creativity and curiosity in problem solving. Visible thought process and clear communication. Pushing boundaries rather than the safe generic design. Owning technical design and its trade-offs. Leverage the five-year founder advantage: build, ship, please customers, design for scale and reliability.
+
+## After
+
+If it goes well: onsite work trial the following week in the new office. Real tickets in a demo environment under NDA, ship to prod if possible, short demo at end of day, meet the team. Graded on communication and the final demo. Explicitly two-way fit.
diff --git a/docs/index.mdx b/docs/index.mdx
new file mode 100644
index 0000000..74e33f6
--- /dev/null
+++ b/docs/index.mdx
@@ -0,0 +1,12 @@
+---
+title: Introduction
+description: Interview briefs and loop context for the job search, one section per active company.
+---
+
+Briefs assembled from the Twenty CRM (meeting notes, role context) and the candidate profile under `.claude/skills/job-application-assistant/`. Each company gets a section: loop overview, interviewer briefs, shared-ground mapping, questions, and the rubric for the next round.
+
+
+
+ Software Engineer (Fullstack). System design with Nick Puljic, Wednesday 2026-09-09, 3:30pm.
+
+
diff --git a/package.json b/package.json
new file mode 100644
index 0000000..667db8c
--- /dev/null
+++ b/package.json
@@ -0,0 +1,13 @@
+{
+ "name": "jobyssey",
+ "private": true,
+ "type": "module",
+ "scripts": {
+ "dev": "blume dev",
+ "build": "blume build",
+ "doctor": "blume doctor"
+ },
+ "dependencies": {
+ "blume": "^1.6.3"
+ }
+}