blakeslee-research/nucleic-purpose-classifier
3
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
c9d21bfc43ddf73a7fa2ba1d8f24004957a6eb49
nucleic-purpose-classifier/data/opus-23.jsonl
T

190 lines
54 KiB
JSON
Raw Normal View History

adding purpose-classifier data
2026-07-29 23:45:26 -07:00
{"prompt": "svelte 5: build the comment thread component. nested replies up to 3 deep, collapse long chains, optimistic insert when you post, runes not stores", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "airflow dag for the nightly Bramble refresh: extract from clickhouse, three transforms in sequence, publish to duckdb, alert on sla miss past 6am", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "docs for the /ingest websocket protocol. frame types, the handshake, backoff expectations on the client side. it's built, it's just tribal knowledge right now", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "training loss goes to nan around step 3k, every time, same step give or take 50. lr is 3e-4 with warmup, grad clip at 1.0, bf16. i've tried lowering the lr and it just moves the cliff later", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.85, "slice": "core", "lang": "en"}
{"prompt": "should we build the notification fan-out on temporal or just do it with pg + a poller? want the tradeoffs written down, we'll decide friday", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "drop the `deprecated_` prefix from the two helpers in lib/dates.py, nothing external imports them", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.15, "slice": "boundary", "lang": "en"}
{"prompt": "is our jwt verification actually checking `aud`? i see the audience in the config but i can't find where it's compared", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "the three env parsing helpers in cmd/ all do the same thing with slightly different error messages. one helper, keep the strictest behavior", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "core", "lang": "en"}
{"prompt": "ok so thinking out loud here. we have a ClickHouse cluster that's fine for the dashboards but the ad-hoc queries analysts run are killing it, and finance wants a semantic layer so nobody hand-writes revenue definitions again. options as i see them: cube.dev in front, or dbt metrics + a thin api, or just give them a read replica and stop worrying. i don't know enough about the tradeoffs. can you write this up properly with a recommendation, including what we'd have to own operationally in each case", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.85, "slice": "core", "lang": "en"}
{"prompt": "kill the `--legacy-json` flag, nothing uses it", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "phoenix liveview: the presence list flickers empty for a frame on reconnect and users notice", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "cassandra -> scylla, worth the migration for us? write me something i can circulate. we're at 12TB, read heavy, p99 matters more than throughput", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "boundary", "lang": "en"}
{"prompt": "flutter: bottom nav with 4 tabs, keeps its own navigator stack per tab, and the badge on the third tab reads from the unread count provider", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.55, "slice": "core", "lang": "en"}
{"prompt": "explain vs summarize, i want the summarize version: what does packages/foxglove actually do, three paragraphs, for the wiki page a PM will read", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.45, "slice": "boundary", "lang": "en"}
{"prompt": "nats jetstream consumer for the `orders.*` subjects with a durable name, max deliver 5, and a dlq subject for the poison messages", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.55, "slice": "core", "lang": "en"}
{"prompt": "read internal/halyard/lease.go and tell me if the ttl renewal can race with expiry. no changes, i just want to know if my worry is real", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "boundary", "lang": "en"}
{"prompt": "raise the alert threshold on QueueDepthHigh from 500 to 2000, it pages constantly and nobody acts on it", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.15, "slice": "core", "lang": "en"}
{"prompt": "hoist the duplicated pagination logic out of the six list resolvers into one helper. same page sizes, same cursor encoding, don't change the wire format", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.55, "slice": "core", "lang": "en"}
{"prompt": "angular: the reactive form marks itself dirty on init because we patchValue before subscribing. users get the unsaved-changes prompt on a page they never touched", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.55, "slice": "core", "lang": "en"}
{"prompt": "readme for the driftwood python package. what it does, install from our internal index, the one-liner usage, and a note that the 0.x api will break", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "bevy: sprite batching for the tilemap. 200x200 tiles, currently one entity per tile and it dies at 60fps", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "design the multi-region story for Larkspur. active-passive is fine, but i need the failover to be under 5 min and i don't want to lose writes. postgres, one primary in us-east", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.9, "slice": "core", "lang": "en"}
{"prompt": "prisma schema + migration for the new `webhook_endpoints` table — url, secret (encrypted at rest), event filter array, per-endpoint enabled flag, fk to workspace with cascade", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "pasted from the review comment thread, want your take on whether they're right:\n\n```\n@rmk commented on src/queue/worker.ts line 88\n\n> const jobs = await redis.lrange(key, 0, batchSize - 1)\n> await redis.ltrim(key, batchSize, -1)\n> for (const raw of jobs) {\n> await handle(JSON.parse(raw))\n> }\n\nthis isn't atomic — if the process dies between lrange and ltrim you\nreprocess, and if it dies after ltrim you lose the batch entirely.\nalso the sequential await in the loop means one slow job blocks the\nwhole batch, which defeats the point of batching.\n\nsuggest either a lua script for the pop, or move to a proper\nreliable-queue pattern with a processing list.\n\n@dtl replied\nwe're at-least-once anyway so the reprocess case is fine, and jobs are\nidempotent. the loss case i agree is bad.\n```\n\nis dtl's defense sound? and is the sequential loop actually a problem given our job mix", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "pasted-context", "lang": "en"}
{"prompt": "pull the tailwind config out into the shared preset so all three apps stop having slightly different spacing scales", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "fix the s in \"recieve\" — it's in the email template subject line", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.05, "slice": "core", "lang": "en"}
{"prompt": "map view for the delivery tracker. mapbox, cluster the pins over 50, draw the route polyline, recenter button that doesn't fight with user panning", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.65, "slice": "core", "lang": "en"}
{"prompt": "plan out the offline-first rewrite of the field app and then get the local sqlite schema + the sync queue skeleton in place so i can see it working", "purpose": "planning", "secondary": "backendImpl", "mixed": true, "difficulty": 0.85, "slice": "mixed", "lang": "en"}
{"prompt": "temporal workflow for the KYC pipeline: three external vendor calls, each can take days, human review activity in the middle, and the whole thing has to be resumable after a deploy", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "why so slow", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "vague-eval", "lang": "en"}
{"prompt": "release notes 4.1, and please mention the breaking change to the `--out` flag prominently, people are going to get bitten", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.35, "slice": "core", "lang": "en"}
{"prompt": "zig: replace the recursive descent in the config parser with an explicit stack. it stack-overflows on deeply nested files. same accepted grammar", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "boundary", "lang": "en"}
{"prompt": "the docker healthcheck says healthy while the app is still doing migrations, so traffic hits a 503 for ~20s every deploy", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "hook the export button to the new /v1/exports endpoint, show the progress toast, and download when it's done", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "audit our GitHub Actions for supply chain hygiene — unpinned actions, secrets in pull_request_target, missing permissions blocks. report only, i'll fix", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "duckdb query in the CLI that gives me p50/p95/p99 latency per endpoint per day from the parquet logs. add it as `latency` subcommand", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "ansible role name is `postgres_setup`, standard is verb-less nouns, make it `postgresql` and update the two playbooks that include it", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "boundary", "lang": "en"}
{"prompt": "milestones + owners for getting the design system to 1.0. 34 components, 19 done, and we need docs and a11y sign-off on each. quarter boundary is in 7 weeks", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.65, "slice": "core", "lang": "en"}
{"prompt": "one line: the retry count in workers/sync.ts should be 5 not 3", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.05, "slice": "core", "lang": "en"}
{"prompt": "an dieser stelle brauche ich eine erklärung: wie entscheidet der scheduler in services/thistle welche jobs zuerst laufen? ich sehe zwei sortierungen und weiß nicht welche gewinnt", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.55, "slice": "core", "lang": "de"}
{"prompt": "vue 3: the virtual list keeps its scroll position when you filter, so you end up staring at blank space. should reset to top on filter change but keep position on refetch", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.55, "slice": "boundary", "lang": "en"}
{"prompt": "changelog + a short migration note for consumers of the `signalbox` sdk, the event payload gained a required field", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "core", "lang": "en"}
{"prompt": "godot: inventory grid ui, drag items between slots, stack counts, right click splits the stack. 8x4 grid, gamepad navigable too", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "OTEL spans through the whole checkout path — worker at the edge, the two go services, and the python risk scorer. propagate traceparent, don't lose the link across the queue hop", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "someone left a mess in the terraform, here's terraform plan, and i genuinely can't tell if this is safe to apply:\n\n```\n # aws_db_instance.primary will be updated in-place\n ~ resource \"aws_db_instance\" \"primary\" {\n ~ allocated_storage = 200 -> 500\n ~ backup_retention_period = 7 -> 14\n ~ instance_class = \"db.r6g.xlarge\" -> \"db.r6g.2xlarge\"\n ~ parameter_group_name = \"pg16-tuned-v2\" -> \"pg16-tuned-v3\"\n id = \"harborview-prod\"\n }\n\n # aws_db_parameter_group.pg16_tuned_v3 will be created\n + resource \"aws_db_parameter_group\" \"pg16_tuned_v3\" {\n + family = \"postgres16\"\n + parameter {\n + apply_method = \"immediate\"\n + name = \"max_connections\"\n + value = \"800\"\n }\n + parameter {\n + apply_method = \"pending-reboot\"\n + name = \"shared_buffers\"\n + value = \"{DBInstanceClassMemory/16384}\"\n }\n }\n\n # aws_security_group_rule.db_from_bastion will be destroyed\n - resource \"aws_security_group_rule\" \"db_from_bastion\" {\n - cidr_blocks = [\"10.2.0.0/24\"] -> null\n - from_port = 5432 -> null\n }\n\nPlan: 1 to add, 1 to change, 1 to destroy.\n```\n\nwhich of these forces a reboot and is the sg rule removal going to lock us out mid-apply", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "pasted-context", "lang": "en"}
{"prompt": "make the settings sheet dismissable by swiping down, it only closes via the X right now", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "boundary", "lang": "en"}
{"prompt": "consolidate the four almost-identical Dockerfiles into one with build args. images should come out the same size and contents", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.55, "slice": "core", "lang": "en"}
{"prompt": "sketch out how we'd do incremental static regeneration for the 400k product pages without paying for 400k builds. next app router, on cloudflare", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "requests to the risk scorer time out at exactly 60s under load, but the scorer itself logs a 1.2s response. envoy is in the path", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "doc comments on every exported func in pkg/tanager. godoc style, and the package doc should explain the two-phase commit thing because nobody gets it from the code", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "add a `--json` output mode to every subcommand. one shared encoder, stable field names, and stderr stays human readable", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "the copy button on code blocks copies the line numbers too", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.25, "slice": "boundary", "lang": "en"}
{"prompt": "compare our current approach (polling the api every 30s) with SSE and with websockets for the live order board. i mostly want to know what breaks at 5k concurrent viewers", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.65, "slice": "boundary", "lang": "en"}
{"prompt": "rails: the counter cache on `comments_count` drifts. we do a raw update in one migration and a bulk delete in the moderation job, neither touches the cache. want it fixed properly and a rake task to recompute", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "astro: content collection for the docs, sidebar generated from frontmatter order, and a version switcher that swaps the whole tree", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.65, "slice": "core", "lang": "en"}
{"prompt": "necesitamos documentar el flujo de autenticación para el equipo nuevo — diagrama en mermaid, los cuatro endpoints, y qué pasa cuando el refresh token expira. el código no se toca", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "es"}
{"prompt": "sqlalchemy 1.x style queries all over the repo, move to 2.0 select() style. behavior and emitted sql should match, tests are the contract", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "cloudflare worker that signs image transform urls and proxies to r2, with a 1 year cache and a purge hook on upload", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.55, "slice": "core", "lang": "en"}
{"prompt": "roadmap-ish: what would it take to get cold start under 200ms for the lambda-backed api. current is 1.4s worst case. i want options ranked by effort", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.75, "slice": "core", "lang": "en"}
{"prompt": "the mobile menu doesn't close when you tap a link, you have to hit the X", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "boundary", "lang": "en"}
{"prompt": "look at the two implementations of ULID generation we somehow have (pkg/ids and web/src/lib/ulid.ts) and tell me if they'd ever produce colliding or out-of-order ids across processes", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "pytorch dataloader is the bottleneck, gpu util sits at 30%. num_workers is 8, pin_memory on, decoding jpegs on cpu. 4090, images are 1024px", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "keyboard-only pass on the checkout flow: tab order, focus trap in the modal, escape closes, and the error summary gets focus on submit failure", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "write the FAQ section for the self-hosting docs. i'll give you the questions support gets most: can i use my own s3, why does it need two containers, how do i rotate the encryption key, is there an air-gapped mode", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.45, "slice": "core", "lang": "en"}
{"prompt": "elixir: genserver that owns the rate limit buckets, ets backed, and survives restart without letting everyone through at once", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "same as the last cleanup — inline the single-use `buildQueryParams` wrapper and delete it. it's only called from one place and it obscures more than it helps", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.25, "slice": "boundary", "lang": "en"}
{"prompt": "plan the api versioning scheme (we have none) and then implement the routing + deprecation header middleware for it", "purpose": "planning", "secondary": "backendImpl", "mixed": true, "difficulty": 0.75, "slice": "mixed", "lang": "en"}
{"prompt": "playwright suite times out on CI about a third of the time. the failing test is never the same one. 4 workers, chromium only, the app is served from a preview build", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.75, "slice": "core", "lang": "en"}
{"prompt": "sort out the theme tokens", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "vague-eval", "lang": "en"}
{"prompt": "raspberry pi: mqtt bridge that reads the CAN bus, decodes the six PGNs we care about, and publishes to the broker with a 1hz rate cap per signal", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "second pair of eyes on the schema before i migrate:\n\n```sql\ncreate table membership (\n id bigserial primary key,\n workspace_id bigint not null references workspace(id),\n user_id bigint not null references app_user(id),\n role text not null check (role in ('owner','admin','member','guest')),\n invited_by bigint references app_user(id),\n invited_at timestamptz not null default now(),\n accepted_at timestamptz,\n removed_at timestamptz,\n unique (workspace_id, user_id)\n);\n\ncreate index on membership (user_id);\ncreate index on membership (workspace_id, role);\n```\n\nthings i'm unsure about: the unique constraint means a removed member can never be re-invited, right? and do i actually need both indexes given the unique already gives me one on (workspace_id, user_id)", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "pasted-context", "lang": "en"}
{"prompt": "bump the eslint config package to 9.2 and run the codemod for the flat config rename", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.25, "slice": "core", "lang": "en"}
{"prompt": "figure out the tab bar", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "vague-eval", "lang": "en"}
{"prompt": "we're going to open source the parser. plan the extraction: what stays private, license headers, how we keep the internal fork in sync, CI for external contributors, and the first release checklist", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.75, "slice": "core", "lang": "en"}
{"prompt": "kotlin: replace the nested when-expression state handling in PaymentReducer with a sealed interface + exhaustive dispatch. identical transitions, the tests shouldn't need edits", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "safari only: the sticky footer jumps up when the on-screen keyboard opens and covers the input we just focused", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.65, "slice": "core", "lang": "en"}
{"prompt": "docs page comparing our three storage backends (local, s3, r2) — capabilities table, config snippets, and which one we recommend for what", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.45, "slice": "boundary", "lang": "en"}
{"prompt": "settings page skeleton with the sections stubbed: profile, notifications, api keys, danger zone. just layout and nav for now, no data", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "core", "lang": "en"}
{"prompt": "server-side cursor pagination for the /activity feed. keyset on (created_at, id) desc, opaque base64 cursor, and it has to be stable when new events land mid-scroll", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "is there a reason `parseAmount` returns a float? asking before i change it to a decimal type and break something i can't see", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.45, "slice": "boundary", "lang": "en"}
{"prompt": "the sql in dashboards/queries.py is embedded in f-strings with `.format()`. move them to .sql files loaded at import, parameterized properly. output must be identical", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "npm audit noise aside, this is real, and i just want it patched:\n\n```\n$ npm run lint\n\n> [email protected] lint\n> eslint . --max-warnings=0\n\n/repo/src/upload/presign.ts\n 41:11 error 'crypto' is imported but never used @typescript-eslint/no-unused-vars\n 58:5 error Unsafe usage of optional chaining no-unsafe-optional-chaining\n 59:20 error Expected a `string` but received `any` @typescript-eslint/no-unsafe-argument\n\n/repo/src/upload/index.ts\n 12:1 error Import in body of module; reorder to top import/first\n\n✖ 4 problems (4 errors, 0 warnings)\nERROR: \"lint\" exited with 1.\n```\n\nthe optional chaining one is on `parts?.length - 1` which is genuinely wrong, the rest are mechanical", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "pasted-context", "lang": "en"}
{"prompt": "profile the tile renderer hot loop and tell me where the time goes. don't optimize yet, i want to understand it first", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "boundary", "lang": "en"}
{"prompt": "design the permissions model for Owlet. resources nest (org > project > dataset), roles are inheritable, and there's a share-by-link case that doesn't fit rbac cleanly. write it up before i code myself into a corner", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.85, "slice": "core", "lang": "en"}
{"prompt": "line 214 of ingest.go, the ctx should be the request ctx not context.Background()", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "boundary", "lang": "en"}
{"prompt": "onboarding checklist widget — 5 steps, checkmarks animate in, collapses to a progress pill once complete, and it should remember dismissal per user", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.55, "slice": "core", "lang": "en"}
{"prompt": "grafana dashboards as code with jsonnet. one dashboard per service from a shared library, and the alert rules alongside", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "memory grows unbounded in the websocket gateway over ~12h and then k8s OOMKills it. 8k connections steady state. i've got heap profiles but they mostly show []byte and i don't know who's holding them", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.85, "slice": "core", "lang": "en"}
{"prompt": "spell out what our SLOs should be for the public api and write the doc. we have no numbers today, just vibes and a status page", "purpose": "planning", "secondary": "writing", "mixed": true, "difficulty": 0.65, "slice": "mixed", "lang": "en"}
{"prompt": "make the diff view in the PR page collapse unchanged hunks with a \"show 42 more lines\" affordance", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.55, "slice": "core", "lang": "en"}
{"prompt": "何が原因でこのテストだけ CI で落ちるのか調べてほしい。ローカルでは通る。タイムゾーンかもしれないけど確証がない", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "ja"}
{"prompt": "dedupe the two http clients. one has retries and the other has tracing, i want one that has both, and every caller moved over", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.65, "slice": "core", "lang": "en"}
{"prompt": "swagger says the field is optional, the handler 400s without it. which one's the bug? i genuinely don't know what the intended contract is", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "boundary", "lang": "en"}
{"prompt": "nix flake for the dev env: go 1.23, node 22, protoc with the two plugins, and postgres 16 for the test harness. has to work on darwin and linux", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.65, "slice": "core", "lang": "en"}
{"prompt": "quick copy fix — the empty state says \"No results found for your search\" even when the user hasn't searched yet", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "boundary", "lang": "en"}
{"prompt": "PR description for the branch, plus i want a one-paragraph summary i can paste into the release notes later", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "svelte transitions on the sidebar are janky on low-end android. currently animating width, i assume that's the problem but i want it smooth without changing the layout behavior", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.55, "slice": "boundary", "lang": "en"}
{"prompt": "let's decide the caching strategy for the product api. i've written notes below, turn them into a real design with numbers:\n- catalog changes maybe 200 times a day, reads are ~4k rps\n- prices change more often, and staleness on price is a legal problem not just a ux one\n- we have redis already but no cdn in front of the api\n- multi-tenant, and tenant A must never see tenant B's cached page\n- there's a bulk endpoint that returns up to 500 products, which i suspect ruins any naive per-key cache\nwhat's the layering, what are the TTLs, and how do we invalidate on price change without a stampede", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.85, "slice": "core", "lang": "en"}
{"prompt": "rust: the borrow checker is fighting me in the arena allocator and i can't tell if my design is wrong or i just need a lifetime annotation. cannot borrow `*self` as mutable more than once", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "explain the sharding function in pkg/ferrite/shard.go and whether adding a 5th shard rebalances everything or just moves 1/5th", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "supabase edge function for stripe webhooks. verify the signature, handle the 6 event types we care about, upsert into subscriptions, idempotent on event id", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "our error codes are documented in three places and all three disagree. pick the code as truth, write one canonical table in docs/errors.md, and note which of the other two docs to delete", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "extract the retry/backoff bits from every service into a shared package. there are 5 slightly different implementations, keep the most conservative jitter", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "set log level to info in prod, someone left it on debug and we're paying for it", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.1, "slice": "core", "lang": "en"}
{"prompt": "storybook stories for the 9 form components, including the error and disabled states, and a controls panel for the props that matter", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "clickhouse: the materialized view stopped updating three weeks ago and nobody noticed. inserts to the source table still land, the mv is just empty for that period", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.75, "slice": "core", "lang": "en"}
{"prompt": "write up the tradeoff between doing tenant isolation with RLS vs separate schemas vs separate databases, for our size (2k tenants, biggest is 40% of the data). recommendation at the end", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "add ETag + If-None-Match support to the config endpoint so the agents stop pulling 2MB every 30 seconds", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "rename everything called `blob` in the storage layer to `object` — types, methods, table columns, the metrics names. it's a naming thing, nothing behavioral", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.55, "slice": "boundary", "lang": "en"}
{"prompt": "is it a problem that we call `time.Now()` inside the comparison instead of passing a clock? asking about testability more than correctness", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "boundary", "lang": "en"}
{"prompt": "changelog for the last 40 commits, grouped, skip the dependabot ones", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "dark mode looks fine everywhere except the pdf preview pane which is blinding white. invert or frame it, your call, but it needs to stop being a flashbang", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.45, "slice": "core", "lang": "en"}
{"prompt": "load test says we fall over at 900 rps, and i want to know why before i throw instances at it. attaching the k6 summary and the pod metrics from the run:\n\n```\n ✗ status is 200\n ↳ 91% — ✓ 41205 / ✗ 3897\n\n http_req_duration..............: avg=1.42s min=41ms med=812ms max=30.01s p(95)=4.88s p(99)=21.4s\n http_req_failed................: 8.63% 3897 out of 45102\n http_reqs......................: 45102 901.2/s\n iteration_duration.............: avg=2.21s\n vus............................: 400 min=400 max=400\n\n checks.........................: 91.36%\n```\n\n```\nNAME CPU(cores) MEMORY(bytes)\nquill-api-6b8f9c7d4-2vk9x 980m 412Mi\nquill-api-6b8f9c7d4-8jd2p 1010m 398Mi\npgbouncer-5f7c8b9d6-xk4mn 120m 88Mi\n```\n\ncpu limit is 1000m per pod. is that just \"add pods\" or is something pathological", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.75, "slice": "pasted-context", "lang": "en"}
{"prompt": "小さい修正: `retry_after` の単位が秒じゃなくてミリ秒になってる。api のレスポンス側を秒に直して", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "ja"}
{"prompt": "audit the js bundle. what's the biggest thing in there that we don't need, and is moment still being pulled in by something", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "dagster asset graph for the recommendations pipeline. four assets, partitioned daily, backfill-safe, and the last one writes to the feature table the api reads", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "flatten the inheritance in the exporter classes. BaseExporter -> AbstractFileExporter -> CsvExporter is two levels too many for what it does, composition instead", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "notification banner in the mac app — slides down from the titlebar, auto hides, click takes you to the relevant row. appkit not swiftui, the rest of that window is appkit", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "outline the plan for cutting our AWS bill 30% and then actually do the two easiest items on it", "purpose": "planning", "secondary": "quickFix", "mixed": true, "difficulty": 0.7, "slice": "mixed", "lang": "en"}
{"prompt": "the docs build is fine but every internal link in the generated site has /docs/docs/ doubled up", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "boundary", "lang": "en"}
{"prompt": "write the release blog post for 5.0. big items: the new query planner, plugin api, and we finally dropped node 18. tone should be confident but not breathless", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.45, "slice": "core", "lang": "en"}
{"prompt": "server side: soft delete with a 30 day grace period on documents. cascade to comments and attachments logically but not physically, and a restore endpoint", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.65, "slice": "core", "lang": "en"}
{"prompt": "gimme the rundown on lib/cinderblock, i inherited it and there's no readme", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "spacing on the invoice line items is inconsistent — some rows 8px, some 12px. make them all 10px", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.15, "slice": "boundary", "lang": "en"}
{"prompt": "reorganize the monorepo. currently apps/, packages/, and a `shared/` that's neither. i want a target layout and the moves done with git history preserved where possible", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.75, "slice": "core", "lang": "en"}
{"prompt": "our chart legend truncates series names with no tooltip so users see \"Enterprise – Nor…\" and \"Enterprise – Nort…\" and can't tell them apart", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "core", "lang": "en"}
{"prompt": "vou precisar de um plano de migração dos jobs do sidekiq pro solid_queue. temos 40 workers, alguns com retry customizado, e não podemos parar a fila", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.75, "slice": "core", "lang": "pt"}
{"prompt": "intermittent 401s from the mobile client, maybe 1 in 200 requests, and retrying works. clock skew? token cache? i can't reproduce it on demand", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "openapi -> typescript client generation as part of the build, and check the generated file in so diffs are visible in PRs", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "what's the difference between `sync` and `syncNow` in the swift package, and which one should i be calling from the background task", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "boundary", "lang": "en"}
{"prompt": "add the missing type hints to analytics/*.py and write proper docstrings while you're at it. mypy strict should pass after", "purpose": "writing", "secondary": "refactor", "mixed": true, "difficulty": 0.6, "slice": "mixed", "lang": "en"}
{"prompt": "just make the button blue", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.05, "slice": "vague-eval", "lang": "en"}
{"prompt": "command palette for the web app — fuzzy search over commands and recent docs, cmd+k, arrow nav, and it should feel instant with 3k items", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "spike doc: can we serve the ML model from the edge? 40MB onnx, 30ms budget, and we'd need the tokenizer there too. tell me if it's plausible and what the shape would be", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "rewrite the git history? no. but do move the four `utils.ts` files into one `lib/` with named modules, nothing else", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.45, "slice": "core", "lang": "en"}
{"prompt": "the CSV export writes \\r\\n on linux and \\n on mac because we use the platform default. pick \\r\\n always, excel users are the ones who care", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "grpc health checking + graceful shutdown in the go services. drain in-flight, stop accepting, deregister from consul, 15s max", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "the last 4 releases each broke something in the mobile client and i want to know if the api contract is the problem or the client's assumptions are. read both sides and tell me", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.75, "slice": "core", "lang": "en"}
{"prompt": "docs for the CLI's config file. every key, the type, the default, the env var equivalent, and precedence order. generate it from the struct tags if you can, otherwise by hand", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "app crashes on launch on iPad only, portrait only:\n\n```\n*** Terminating app due to uncaught exception 'NSInternalInconsistencyException',\nreason: 'Attempting to load the view of a view controller while it is deallocating\nis not allowed and may result in undefined behavior (<SPLSplitHost: 0x14e0a2f00>)'\n\n0 CoreFoundation __exceptionPreprocess + 172\n1 libobjc.A.dylib objc_exception_throw + 56\n2 Foundation -[NSAssertionHandler handleFailureInMethod:...] + 216\n3 UIKitCore -[UIViewController loadViewIfRequired] + 1244\n4 UIKitCore -[UIViewController view] + 28\n5 SwiftUI SplitViewHost.updateHostedViews() + 412\n6 Sandpiper SidebarContainer.body.getter + 88\n7 Sandpiper RootScene.body.getter + 204\n```\n\nnothing in RootScene changed this release, but we did move the sidebar selection into an @Observable model", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "pasted-context", "lang": "en"}
{"prompt": "long press on a message opens the reaction picker, six emoji + a plus, and the picker should flip above the bubble if it's near the bottom", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "figure out the right way to do zero-downtime column type changes on our biggest table (bigint -> uuid on 400M rows) and write the runbook for it", "purpose": "planning", "secondary": "writing", "mixed": true, "difficulty": 0.9, "slice": "mixed", "lang": "en"}
{"prompt": "the `Deploy` button is enabled while a deploy is in flight, so people double-click it and we get two deploys", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "boundary", "lang": "en"}
{"prompt": "batch insert path for the metrics writer. currently one insert per point, want batches of 500 or 200ms whichever first, with backpressure to the caller", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.65, "slice": "core", "lang": "en"}
{"prompt": "does the code in workers/reaper.ts actually handle the case where the row was already deleted by another worker, or is it relying on the delete being idempotent by luck", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.55, "slice": "boundary", "lang": "en"}
{"prompt": "clean up the imports across web/src — there's a mix of relative and alias paths, sometimes in the same file. standardize on the @/ alias", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.35, "slice": "boundary", "lang": "en"}
{"prompt": "summarize this incident channel into a timeline i can put in the postmortem:\n\n```\n14:02 robb pages went off, api error rate 40%\n14:03 robb looking\n14:05 ani grafana shows db connections maxed, 800/800\n14:06 robb nothing deployed since 11:40\n14:08 ani there's a query doing a seq scan on events, 40s, hundreds of them\n14:09 ani it's the new /reports/export endpoint. shipped at 11:40.\n14:11 robb killing the long runners\n14:12 robb connections down to 300, error rate falling\n14:14 ani error rate back to baseline\n14:16 robb we should feature flag that endpoint off\n14:19 ani flag off, deployed\n14:31 ani confirmed stable for 15m. the missing index is on events(workspace_id, created_at)\n14:52 robb index created concurrently, took 8 min\n15:05 ani re-enabled the endpoint at 10%, latency looks fine\n15:40 ani 100%, closing this out\n```", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "pasted-context", "lang": "en"}
{"prompt": "settings screen on watchOS. three rows, a toggle, and it has to sync with the phone app's defaults", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "zod schemas for all 22 api responses, generated from the openapi if possible, and used at the fetch boundary so bad payloads fail loudly", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "look at how we handle the case where a user is deleted while they have an active session. i suspect nothing handles it and i'd like to know before a customer finds out", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "everything's broken again", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "vague-eval", "lang": "en"}
{"prompt": "phased plan to get the frontend off jQuery. 60 pages, some of them are just forms, a few have real interactivity. we can't do a big bang and we ship weekly", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "shorten the truncation on the breadcrumb from 30 chars to 20", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.05, "slice": "core", "lang": "en"}
{"prompt": "swap our custom debounce for the one in the utils package, it's the same implementation and one of them has a bug with leading edge", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.35, "slice": "boundary", "lang": "en"}
{"prompt": "sse endpoint for build logs with heartbeat comments every 20s so proxies don't kill it, plus last-event-id resume", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "gantt-ish view for the deploy pipeline stages — horizontal bars, live durations ticking, failed stages in red with the error inline", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "help text for `roundhouse migrate --help` is a wall of text. rewrite it: one line summary, grouped flags, two examples at the bottom", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "boundary", "lang": "en"}
{"prompt": "restore from the nightly backup produced a db that's 40% smaller than prod and i can't figure out what's missing. pg_dump/pg_restore, custom format, and the restore reported no errors", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "review the migration in db/migrations/0142_split_addresses.sql for anything that'll lock the table longer than a second", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "picking up the search relevance work from last sprint. plan the eval harness — golden queries, judgments, the metric, and how it gates a model change in CI", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "the api key input shows the full key after save. mask it except the last 4", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.25, "slice": "boundary", "lang": "en"}
{"prompt": "gradle: split the `core` module's build.gradle.kts, it configures 6 things that belong in convention plugins. build outputs identical", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.65, "slice": "core", "lang": "en"}
{"prompt": "background upload with progress on android — workmanager, resumable, survives process death, and notification with a cancel action", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "tooltip on the disabled submit button explaining why it's disabled, because right now people just click it repeatedly", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.35, "slice": "boundary", "lang": "en"}
{"prompt": "readme is out of date in like six places and there's a section about a feature we removed. rewrite it top to bottom against what the code actually does now", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.45, "slice": "core", "lang": "en"}
{"prompt": "the pagination is off by a page somewhere between the api and the ui. api says total 240, page size 25, ui renders 9 page buttons and the last one is empty", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.45, "slice": "core", "lang": "en"}
{"prompt": "walk me through what happens on `kelpstack up` from the first line of main to a running stack. i want the sequence, not a code review", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.55, "slice": "boundary", "lang": "en"}
{"prompt": "worker pool with graceful resize — grow to 64 under load, shrink to 8 idle, no dropped jobs during a resize, and expose the current size as a gauge", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "design the schema and access patterns for storing 2 years of per-minute sensor readings from 30k devices. we're on postgres now and it's already groaning at 3 months", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.85, "slice": "core", "lang": "en"}
{"prompt": "reduce the motion on the page transitions, someone complained it makes them queasy and honestly they're right", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.35, "slice": "boundary", "lang": "en"}
{"prompt": "s'assurer que le lock est bien relâché en cas d'exception dans jobs/reaper.py — j'ai un doute sur le try/finally", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.45, "slice": "boundary", "lang": "fr"}
{"prompt": "collapse the 11 boolean flags on the User model into a `preferences` jsonb with a typed accessor layer. same defaults, same behavior, migration included", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "boundary", "lang": "en"}
{"prompt": "commit message. it's a fix for the double-charge thing plus a test", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "cron entry is `0 */4 * * *` but the job needs to run at 2am and 2pm only", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.1, "slice": "core", "lang": "en"}
{"prompt": "user reports: \"the app logs me out randomly.\" i have three reports, all android, all on wifi->cellular transitions as far as i can tell. no server side session expiry in the logs at those times", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "tree view for the file browser, lazy load children, keyboard nav, and multi-select with shift/cmd. it'll hold maybe 20k nodes at the extreme", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "webhook signature verification helper published as a tiny package for our customers, plus the docs page showing it in node, python and ruby", "purpose": "backendImpl", "secondary": "writing", "mixed": true, "difficulty": 0.6, "slice": "mixed", "lang": "en"}
{"prompt": "assess whether we can drop the redis dependency entirely. it's used for sessions, the rate limiter, and one cache. tell me what each would cost us to move", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "the thing with the flags", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "vague-eval", "lang": "en"}
{"prompt": "translate the error messages in i18n/en.json to de and fr. tone should be formal-ish for de, and keep the placeholders intact", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "core", "lang": "en"}
{"prompt": "cut the bundle: code split the admin routes, they're 300kb of the 480kb main chunk and 2% of users see them", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.55, "slice": "boundary", "lang": "en"}
{"prompt": "im starting the tenant offboarding feature. plan it first — data export, hard delete after 30 days, the legal hold exception, and how we prove deletion happened", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "set `strict: true` in tsconfig and fix the fallout. there'll be a few hundred errors, work through them without changing runtime behavior", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.75, "slice": "boundary", "lang": "en"}
{"prompt": "the avatar images 404 for users who signed up before march. i think the url template changed and we never backfilled", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.45, "slice": "boundary", "lang": "en"}
{"prompt": "GET /v1/usage with from/to, group_by=day|hour, and a `metric` filter. reads from the rollup table, falls back to raw for today, and caps the range at 90 days", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "tell me whether the code in packages/pellet/src/merge.ts is a correct CRDT merge or just something that looks like one. i wrote it 8 months ago and i no longer trust myself", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "polish pass on the invoice pdf template — alignment on the totals column, the logo is 4px off center, and the footer legal text should be 8pt gray", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "core", "lang": "en"}
{"prompt": "one paragraph for the docs homepage that says what this product does without any marketing words. i've rewritten it four times and hate all of them", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.35, "slice": "core", "lang": "en"}
{"prompt": "the deploy workflow needs GH_TOKEN swapped for the new PAT secret name DEPLOY_PAT", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.1, "slice": "core", "lang": "en"}
Reference in New Issue Copy Permalink