data/opus-8.jsonl

{"prompt": "we need a plan for the storage engine rewrite. current one is a naive append log with a full scan on startup and startup is now 40 seconds. i want options — lsm, b-tree, or keep the log but add an index — with what each costs in write amplification and what the migration looks like for existing data files", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.9, "slice": "core", "lang": "en"}
{"prompt": "implement the wal with crc per record, torn-write detection on recovery, and fsync batching with a configurable interval", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.9, "slice": "core", "lang": "en"}
{"prompt": "the checkout page needs an order summary that stays visible while you scroll the form on desktop, and collapses to a sticky total bar on mobile", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "the free shiping threshold in the banner says 50 but the actual rule is 75", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "we hae two Arc<Mutex<HashMap>> caches that do the same thing in different modules, merge them into one type and use it in both places, no behavior change", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "the compactor thread deadlocks maybe once a week under load. lock order looks fine to me but obviously it isnt", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.9, "slice": "core", "lang": "en"}
{"prompt": "explain what guarantees our current flush path gives on power loss, i need to know before i promise anything to a customer", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "write the on-disk format spec — header layout, record framing, endianness, version negotiation. someone should be able to write a reader from this doc alone", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "boundary", "lang": "en"}
{"prompt": "- add size variants to the ProductCard (sm/md/lg)\n- price should show strikethrough when there's a sale\n- badge slot top-left for \"new\" / \"low stock\"\n- image aspect 4:5, no layout shift\n- the whole card is a link but the wishlist heart shouldn't trigger navigation", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "inventory reservation service — reserve on add-to-cart with a 20 min hold, release on abandon, and don't oversell under concurrent checkout", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "roadmap for internationalizing the store. currency, tax, addresses, and the copy. we have 4 markets planned and one of them is japan", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "the sort dropdown resets to default when you paginate", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "our cart total is computed on the client for display and on the server for charging, and they've drifted. single implementation shared by both, and the totals must match to the cent on the test fixtures", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "carts randomly empty for logged-out users. we store them in a cookie plus localstorage and i think theyre fighting each other but i cant reproduce reliably", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "does our discount stacking logic allow combining a percentage code with a fixed-amount code? read the code and tell me, marketing is asking", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "product copy guidelines doc for the merch team — voice, length limits per field, and what our templates do with line breaks", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "spark job to backfill the orders fact table from the raw event archive, 3 years, idempotent per day partition", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "plan how we go from batch-daily to near-real-time on the orders pipeline without maintaining two implementations of the same logic", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.9, "slice": "core", "lang": "en"}
{"prompt": "the airflow retries is set to 0 on the ingest dag, make it 3 with a 5 minute delay", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.1, "slice": "core", "lang": "en"}
{"prompt": "our dbt models are one giant folder. reorganize into staging/intermediate/marts with the naming convention, same compiled sql", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "row counts in the mart are 3% lower than the source and it moves every run. i've checked the joins twice", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "review our partitioning and clustering choices on the events table in bigquery, we're spending too much on scans", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "the pipeline docs describe a `raw.orders_v2` table that doesn't exist, and omits the two tables we actually use. get it accurate", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "boundary", "lang": "en"}
{"prompt": "chart page for the merchandising team: revenue by category over time, stacked area, with a category picker and csv export", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "idk just look at it", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "vague-eval", "lang": "en"}
{"prompt": "pasting the profile output because i dont know what to make of it. this is the hot path in our tokenizer:\n\n   Overhead  Command   Shared Object        Symbol\n     31.44%  lexer     lexer                [.] acme_lex_next\n     18.02%  lexer     libc.so.6            [.] __memmove_avx_unaligned_erms\n     12.71%  lexer     lexer                [.] acme_intern_symbol\n      9.88%  lexer     libc.so.6            [.] malloc\n      7.10%  lexer     libc.so.6            [.] free\n      4.02%  lexer     lexer                [.] acme_buf_reserve\n\nso 17% in malloc/free and 18% in memmove. we tokenize about 40MB/s and i want 200. where do i start", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "pasted-context", "lang": "en"}
{"prompt": "arena allocator for the parser so we stop malloc'ing per node, and free the whole arena at the end of a parse", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "syntax highlighting in the code editor component, incremental, and it can't block typing on a 5000 line file", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "the -O2 in the makefile should be -O3 for the release target", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.1, "slice": "core", "lang": "en"}
{"prompt": "our error handling is a mix of return codes and a global errno-alike. move to consistent return codes, same failure behavior at every call site", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "what does the memory ordering on that atomic actually buy us? relaxed vs acquire here, i genuinely dont know if the current code is correct", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "developer docs for embedding the library — build flags, the two init modes, threading rules, and please be explicit about what's not thread safe", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "boundary", "lang": "en"}
{"prompt": "decide the approach for our plugin abi. c abi with opaque handles, or wasm, or a subprocess with ipc. write the comparison and recommend", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.9, "slice": "core", "lang": "en"}
{"prompt": "segfault in the destructor path but only in the release build with lto on. debug and reldebug are clean", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.9, "slice": "core", "lang": "en"}
{"prompt": "cross compile the lib for aarch64 in CI and publish both artifacts", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "the promo code input should show inline validation and the applied discount as a row in the summary, not a toast", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "core", "lang": "en"}
{"prompt": "our currency formatting helper is duplicated in the email templates, the web app and the pdf renderer. one shared function, identical output on the golden files", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "tax is calculated before shipping in one flow and after in another, and i'm pretty sure one of them is wrong but i dont know which", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "boundary", "lang": "en"}
{"prompt": "compare our two approaches to search on the storefront — the postgres tsvector one on the branch and the meilisearch spike — and tell me which to take forward", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "boundary", "lang": "en"}
{"prompt": "checkout error copy is all 'something went wrong'. write real messages for the 8 cases we can actually distinguish", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "boundary", "lang": "en"}
{"prompt": "gift cards. design the whole thing — issuance, partial redemption, expiry rules per market, and what happens on a refund to a gift card order", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "address autocomplete integration with a fallback to manual entry, and normalize what we store so the shipping label doesn't break", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "the size picker should disable out-of-stock sizes instead of letting you select them and fail at checkout", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "core", "lang": "en"}
{"prompt": "order confirmation email has the wrong support address", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.1, "slice": "core", "lang": "en"}
{"prompt": "figure out the plan for headless — decoupling our storefront from the platform templates — then do the product listing page as the pilot", "purpose": "planning", "secondary": "frontendImpl", "mixed": true, "difficulty": 0.8, "slice": "mixed", "lang": "en"}
{"prompt": "work out why the inventory numbers drift and then write the note for the ops team", "purpose": "debugging", "secondary": "writing", "mixed": true, "difficulty": 0.7, "slice": "mixed", "lang": "en"}
{"prompt": "go on then", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "vague-eval", "lang": "en"}
{"prompt": "ok brain dump: our observability is three tools that dont talk to each other, logs in cloudwatch, metrics in datadog, traces in a self hosted jaeger that half the services dont report to. every incident starts with 10 minutes of tab switching. i want a plan to consolidate, including what we accept losing and roughly what it costs", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "otel collector deployment with tail sampling — keep all errors, keep anything over 1s, sample the rest at 5%", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "trace waterfall view in our internal tool, collapsible spans, and highlight the critical path", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "log retention is 400 days, we only need 30", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.1, "slice": "core", "lang": "en"}
{"prompt": "our log statements use 5 different key names for the request id. pick request_id and sweep the whole codebase", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "boundary", "lang": "en"}
{"prompt": "some traces have a 900ms span called 'unknown' with no attributes and no children. what generates that", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "how much are we actually spending on logs per service, based on the volume in the config and the retention", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "write our logging conventions doc: levels and when to use them, what must never be logged, and the required fields", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "boundary", "lang": "en"}
{"prompt": "slo definitions for the 6 tier-1 services, with the error budget policy and what happens when we burn it. i want the doc, not the dashboards", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "boundary", "lang": "en"}
{"prompt": "synthetic checks hitting our critical flows every minute from 3 regions, alerting only when 2 of 3 fail", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "status page component that reads from our incidents api, per-service history strip for the last 90 days", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "pagerduty urgency for the disk alert is high, should be low", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.1, "slice": "core", "lang": "en"}
{"prompt": "our alert rules are copy pasted per service with the service name changed. template them, exact same rules firing at the exact same thresholds", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "the disk alert fires and resolves every 20 minutes all night. disk is at 71% and stable", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "read our runbooks and tell me which ones reference tools or hosts that dont exist anymore", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "incident review doc for the 3 hour degradation last thursday. i'll give you the timeline, you write the narrative and the action items", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "esto es lo que sale en los logs y no entiendo qué lo provoca:\n\n2026-07-28 22:14:02 WARN  [pool-3-thread-7] c.a.sync.Reconciler - remote version 41 < local version 43, skipping\n2026-07-28 22:14:02 INFO  [pool-3-thread-7] c.a.sync.Reconciler - marking local dirty entity=doc_8812\n2026-07-28 22:14:03 WARN  [pool-3-thread-2] c.a.sync.Reconciler - remote version 41 < local version 43, skipping\n2026-07-28 22:14:03 INFO  [pool-3-thread-2] c.a.sync.Reconciler - marking local dirty entity=doc_8812\n2026-07-28 22:14:04 ERROR [pool-3-thread-9] c.a.sync.Reconciler - version conflict unresolvable entity=doc_8812 attempts=32\n\ndos hilos procesan la misma entidad. debería haber un lock por entidad", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "pasted-context", "lang": "es"}
{"prompt": "make the deploy notification message include the diff link and the author", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "so uh i think our feature flag system is the problem. 200 flags, no owners, no expiry, and the evaluation is a 400 line if-chain. plan the cleanup and the guardrails so it doesnt happen again", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "flag evaluation service with percentage rollouts, targeting rules, and a local cache with a streaming update channel", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "flags admin ui: list with owner and age, filter for stale, and a detail view showing the rules and current exposure", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "remove the `legacy_checkout` flag, its been 100% off for a year — delete both branches, keeping the off behavior", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "boundary", "lang": "en"}
{"prompt": "a flag flipped to on for 4% of users who shouldnt match the targeting rule. only in the eu pods", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "explain how a flag change propagates to a running pod and what the worst case delay is", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "write the guide for adding a flag: naming, owner, expiry date, and the cleanup checklist", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "just the usual", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "vague-eval", "lang": "en"}
{"prompt": "graphql federation across our 5 services, or a single gateway schema stitched manually, or just... not. write the analysis. our clients are one web app and one mobile app, and the mobile team wants fewer round trips", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.9, "slice": "core", "lang": "en"}
{"prompt": "persisted queries so the mobile client sends a hash instead of the document, with a build step that extracts them", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "the query result panel needs a json tree view with collapse-all and a copy-path action per node", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "introspection is enabled in prod, turn it off", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.1, "slice": "core", "lang": "en"}
{"prompt": "our resolvers do auth checks inline, inconsistently. move to a directive, same access decisions — enumerate them first so we can compare", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "one nested query takes 12 seconds and the dataloader batching looks like its not kicking in for that path only", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "is our depth limiting actually enforced, or just configured? i see the plugin registered but not in the request path", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "boundary", "lang": "en"}
{"prompt": "schema documentation for external consumers — the three main entry points, pagination conventions, and error extensions", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "chart out the migration from our rest api to graphql for the mobile client, screen by screen, with a way to run both during the transition", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "subscriptions over websockets with auth on connect, and per-subscription authorization not just connection-level", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "the loading skeleton flashes for cached queries, only show it after 200ms", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "boundary", "lang": "en"}
{"prompt": "our generated types are checked in and stale. regenerate and wire the generation into the build so it cant drift", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "mutations succeed but the ui doesn't update for one specific list. cache normalization thing probably, i dont know apollo well enough", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "review the schema for anything that will be painful to change later — non-nullable fields we're not sure about, enums that should be strings, that kind of thing", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "write the changelog for the schema changes this month, and mark which are additive vs breaking for clients", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "core", "lang": "en"}
{"prompt": "cron expression in the k8s cronjob is `0 0 * * 0` but it should run daily", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.1, "slice": "core", "lang": "en"}
{"prompt": "here's the failing test, i think its a small thing:\n\nFAIL src/lib/slug.test.ts > slugify > handles unicode\n\nAssertionError: expected 'cafe-latte' to be 'café-latte'\n\n- Expected\n+ Received\n\n- café-latte\n+ cafe-latte\n\n at src/lib/slug.test.ts:22:31\n\nTest Files  1 failed (14)\n     Tests  1 failed | 88 passed (89)\n\nwe changed slugify to strip diacritics last week for url safety, but this test wants them preserved. i think the TEST is what's wrong now, we decided ascii-only slugs. can you sort it out either way", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "pasted-context", "lang": "en"}
{"prompt": "design the multi-step form framework we keep needing — validation per step, back without losing data, resumable from a saved draft", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "draft saving for the long application form, autosave every 10s and on blur, with a 'last saved' indicator", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "file upload field with drag and drop, per-file progress, cancel, and thumbnails for images", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "the required asterisk is on the optional fields and missing on the required ones, which is impressively wrong", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "we have four form libraries in the repo. standardize on react-hook-form and port the others, same validation behavior and error copy", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "submitting the form twice fast creates two records. i thought the button disabled on submit", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "how does our validation handle a field that becomes required based on another field's value? point me at where that's implemented", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "help text for every field in the application form, short, and it needs to say WHY we're asking for the sensitive ones", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "core", "lang": "en"}
{"prompt": "plan the accessibility remediation and then fix the form labels as the first pass", "purpose": "planning", "secondary": "frontendImpl", "mixed": true, "difficulty": 0.6, "slice": "mixed", "lang": "en"}
{"prompt": "explain what the middleware chain does then write it up for the readme", "purpose": "review", "secondary": "writing", "mixed": true, "difficulty": 0.5, "slice": "mixed", "lang": "en"}
{"prompt": "ok next bit", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "vague-eval", "lang": "en"}
{"prompt": "i want the design for our email sending. currently three services each talk to sendgrid directly with their own templates and nobody can tell what we sent a user. i'm thinking one service, template registry, per-user send log, and preference enforcement in one place", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "email service with template rendering, a send log, and unsubscribe enforcement that no caller can bypass", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "email preferences screen — per-category toggles, a digest frequency picker, and an unsubscribe-from-all with a confirm", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "reply-to on the notification emails is noreply@ but should be support@", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.1, "slice": "core", "lang": "en"}
{"prompt": "our 22 email templates each have their own copy of the header and footer html. extract layouts, and the rendered output should be byte identical for the existing tests", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "emails render fine everywhere except outlook desktop where the layout collapses to one column", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "can a user who unsubscribed still get transactional emails? i want to know exactly which categories bypass the preference check", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "write the copy for the 6 lifecycle emails: welcome, day 3 tips, trial ending, trial ended, win-back, and the annual renewal reminder", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "figure out how we'd support customer-specific smtp for enterprise, including bounce handling when it's their server", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "bounce and complaint webhook handling — suppress the address, notify the owner, and dont retry into a hard bounce", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "the email preview in the admin renders raw html as text", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "template variables are referenced as {{user.name}} in some templates and {{name}} in others. normalize on the dotted form, rendered output unchanged", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "boundary", "lang": "en"}
{"prompt": "one customer gets every email twice, has done for months, and their record looks identical to everyone elses", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "read the send path and tell me where a template rendering error would surface — does the job retry forever or dead letter", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "deliverability doc: our spf/dkim/dmarc setup, why the subdomain split exists, and what breaks if someone changes it", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "sure whatever works", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "vague-eval", "lang": "en"}
{"prompt": "架构方案：把上传服务从单体里拆出来，需要支持断点续传和病毒扫描，先给方案不要写代码", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "zh"}
{"prompt": "virus scanning on upload via clamav in a sidecar, quarantine on detection, and the user sees a clear state not a generic error", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "file browser grid with type icons, multi select, and a details pane that slides in from the right", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "max upload size says 10MB in the ui and 8MB in the server config", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "the storage abstraction has s3 assumptions leaking through it — bucket names in the interface. clean the interface, both backends keep working", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "resumable uploads resume from the wrong offset when the client reconnects on a different pod", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "walk me through what happens if the scan sidecar is down when an upload arrives", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "docs for the upload api — the multipart protocol, resumption, size limits, and the scan states a client might see", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "plan for supporting folders and nested permissions in the file store. currently flat with per-file acls", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.9, "slice": "core", "lang": "en"}
{"prompt": "share links with optional password and expiry, and a view-only mode that doesnt allow download", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "the breadcrumb truncates the middle folders on narrow screens but hides the wrong end", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "boundary", "lang": "en"}
{"prompt": "our path handling uses string concat in 20 places, some with a leading slash some without. one helper, same resolved paths", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "downloads of files with non-ascii names get mangled filenames on windows but only through the browser, curl is fine", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "audit whether a share link can be used to enumerate other files in the same folder", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "user-facing help page about sharing — the link types, what each one allows, and how to revoke", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "propose the approach for versioning files. keep n versions or all, dedupe by content hash, and how the ui presents history", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "content-addressed dedupe for uploads with reference counting, and a gc job that's safe to run while writes are happening", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "version history panel: list with timestamps and authors, preview, restore, and a diff for text files", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "gc runs at 3am utc which is peak for our apac customers, move it", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.1, "slice": "core", "lang": "en"}
{"prompt": "the sha256 helper appears in 6 files with slightly different chunk sizes. one implementation, same digests obviously", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "boundary", "lang": "en"}
{"prompt": "restoring an old version sometimes restores the wrong one, off by one in the ordering when two versions share a timestamp", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "boundary", "lang": "en"}
{"prompt": "explain the reference counting in the dedupe store and whether a crash mid-gc can orphan or delete live data", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "write the internal design doc for the dedupe store from the code as it exists, we never wrote one", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "boundary", "lang": "en"}
{"prompt": "plan the whole thing and then do the first slice", "purpose": "planning", "secondary": "backendImpl", "mixed": true, "difficulty": 0.6, "slice": "mixed", "lang": "en"}
{"prompt": "check the pr for correctness and tidy up the naming while you're there", "purpose": "review", "secondary": "refactor", "mixed": true, "difficulty": 0.5, "slice": "mixed", "lang": "en"}
{"prompt": "resume", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "vague-eval", "lang": "en"}
{"prompt": "figure out how we do zero-downtime schema changes as a policy, not per-migration heroics. expand-contract, tooling, and how CI blocks a dangerous one", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "migration linter in CI that rejects a migration adding a not-null column without a default, or taking an access exclusive lock on a big table", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "migration status page for the admin: which migrations have run per environment, and flag drift", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "the migration timeout is 5s which fails every real migration, make it 15 minutes", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "our 340 migrations include 40 that were edited after being applied. squash the pre-2025 ones into a baseline schema, verified identical against a prod dump", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "a migration applied cleanly in staging and hung for 20 minutes in prod on the same table size", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "review the pending migration for lock risk and tell me if it needs a maintenance window", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "boundary", "lang": "en"}
{"prompt": "document our migration workflow: naming, the review checklist, and the rollback expectations per type of change", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "what's the story for testing against a real database in CI without 20 minutes of setup. template databases? snapshots? write the plan", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "testcontainers setup for the integration suite with a shared postgres per run and per-test schema isolation", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "the test db name is hardcoded so two parallel runs clobber each other", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "our tests share mutable fixture state through module level globals. make each test independent, same assertions, and it should pass with random ordering", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "tests pass individually and fail as a suite, and which one fails changes depending on the order", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "look at our test suite and tell me what's actually covered vs what just executes without asserting anything meaningful", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "testing guide for the repo: what layer to test what at, our fixture conventions, and when a test is allowed to hit the network", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "core", "lang": "en"}
{"prompt": "list view for the audit log with filters by actor, action and date, virtualized, and a permalink per entry", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "eh nvm do the other thing", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "vague-eval", "lang": "en"}