Lab Mode — Crisis Library
Explore any crisis in isolation. No debt accrual. No journal. Pure exploration.
11 of 11 crises
The shape of the ledger
The first contract obligates Ledgerline to deliver an architecture description within 30 days. The data model picked here is the architecture description. There is no deferring this — the next 35 months of crises all reference the shape of the transfer record.
Trade-off
The ledger as a data model: document vs. double-entry table vs. event log. Atomicity of paired writes vs. ease of evolution vs. queryability of history.
The bank timed out mid-transfer
A treasurer is on the phone. Resolution is required within the call. The transfer is $128,000. Doing nothing means either a duplicate transfer or a stuck transfer; both are escalatable. There is no third-party safe answer.
Trade-off
Idempotency keys; at-least-once vs. exactly-once; the gap between "request succeeded" and "money moved".
The reconciliation report takes six hours
The customer has issued a written deadline. Helen does not write twice; the next message will be from her general counsel. Doing nothing means losing the anchor customer.
Trade-off
OLTP vs. OLAP separation. Read replicas, CDC, pre-aggregated rollups, and whether reporting reads are allowed to touch the primary.
Priya signed a multi-currency customer
The contract is signed. Priya cannot un-sign it without a board conversation. The choice is not whether to ship multi-currency — it is what shape the multi-currency data model takes, and Priya needs an answer by end of week to write the SOW.
Trade-off
Domain modeling for money: base-currency-with-FX vs. native-currency-per-row, and the question of which FX rate timestamp the ledger commits to.
Black Friday at the payroll customer
Customer-facing SLA is breached. Halberd's CFO is on the phone. The partner bank's rate limit is a real ceiling; doubling our worker pool will make it worse, not better.
Trade-off
Queue theory and back-pressure: throughput vs. latency, the difference between scaling workers and respecting the downstream's rate.
The auditor wants every transaction touched by employee X
Regulatory deadline. The bank charter we operate under depends on responding. Doing nothing is not an option. This crisis is the M5 callback: if the user picked harder-primary at M5, audit-gap matured here, and the overlay surfaces the M5 journal entry verbatim before any option is shown.
Trade-off
Audit logging as a first-class data model: immutability, provenance, the difference between having logs and having queryable evidence.
The $4,300 hole
Largest customer; 48-hour clock; the diff is between two systems' state, neither of which has its own history. If the user picked the mutable-ledger option at M0 and never paid it down, there is no native answer.
Trade-off
Event sourcing as a debugging tool: the difference between current state and the story of how the state was reached. Why "what does the row say now" is not enough.
The partner bank's 90 days
Ninety-day clock. The partner bank handles ~60% of transfer volume. Failing to migrate means failing the SLA on every contract by Month 27. This crisis is the M8 + M12 callback — the literal-replay overlay surfaces the M8 USD-display-convert journal entry and/or the M12 more-workers journal entry as the cause.
Trade-off
Vendor abstraction, the strangler pattern, dual-write. Why "we'll write a thin wrapper later" is debt waiting to happen.
The junior engineer's migration
Customer-facing outage during business hours. CEO is on customer calls. Devon is twelve months into her first job. The technical fix is bounded; the people decision is not. This crisis is the M24 callback — the literal-replay overlay surfaces the M24 rewrite-integration journal entry, which is the reason no one had time to build migration tooling.
Trade-off
Change management and deploy safety. Procedural fixes vs. tooling fixes. Whose work it actually is to make the next migration safe.
GDPR and PSD2 on the same day
Board has approved expansion; legal has confirmed both regulations apply; cancellation now means returning the Series B. The choice is architectural shape, not whether to expand.
Trade-off
Data residency vs. latency. Regional architecture as compliance, not as performance optimization. The regulatory map.
Acquisition due diligence
Finale. Communication-shaped, not architectural. The user's accumulated debt ledger and decision pattern determines which coda fires (see finale.json). The crisis itself is the writing exercise; the coda is what the user produces.
Trade-off
Communication. Narrative. Owning your history. This is not a technical exercise; it is the ability to defend three years of trade-offs in writing, in front of someone who will not be persuaded by enthusiasm.