blakeslee-research/nucleic-purpose-classifier
3
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

adding purpose-classifier data

Browse Source
This commit is contained in:
abkslm
2026-07-29 23:45:26 -07:00
parent a0f34b89cb
commit c9d21bfc43
61 changed files with 9839 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
data/opus-4.jsonl
+180
View File
@@ -0,0 +1,180 @@
{"prompt": "would you mind drafting the design for the replay system in our netcode? deterministic lockstep vs snapshot interpolation, 60hz sim, up to 8 players, and replays have to stay valid across patch versions", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.9, "slice": "core", "lang": "en"}
{"prompt": "help me carve the k8s migration into phases. we're on ECS today, nobody here has run kubernetes in prod, and we can't have a hard cutover weekend", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "work out the approach for sharding leaderboard writes and write it down. i don't want an implementation yet, just the shape and where it breaks", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "boundary", "lang": "en"}
{"prompt": "please implement the matchmaking queue — mmr buckets that widen the longer you wait, parties up to 4, and a soft cap so a stack of 4 never faces 4 solos", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "terraform module for a spot-backed gpu node group with an on-demand fallback group, taints and tolerations so only training jobs land there", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "the design in docs/adr-014 is already signed off, just build the outbox poller for telemetry events per that doc", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "boundary", "lang": "en"}
{"prompt": "floating damage numbers in the hud — spawn at hit position, drift up, fade over 0.8s, pooled so we don't allocate per hit", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "loading bar looks dead, animate it", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "from the ticket, build this screen:\n\nTitle: Squad detail sheet\nAs a player I want to see my squad's recent matches so I can tell who's been carrying.\n\n- opens as a half sheet from the squad row, drag to full height\n- top: squad crest (48pt), name, member count, join code with a copy button\n- segmented control: Matches / Members / Settings (settings only visible to the owner)\n- Matches tab: list of last 20, each row = map thumbnail, result pill (W/L/D), duration, relative time\n- empty state when a new squad has no matches yet\n- Members tab: avatar rows with online dot, owner gets a swipe-to-kick action\n- pull to refresh on both tabs, skeleton rows while loading\n\nSwiftUI, and reuse ResultPill from the match history screen", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "pasted-context", "lang": "en"}
{"prompt": "helm values, replicaCount 3", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.1, "slice": "core", "lang": "en"}
{"prompt": "tooltip says 'recieve', fix", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.1, "slice": "core", "lang": "en"}
{"prompt": "default port should be 8443 not 8080 — the config struct and the sample yaml both", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "boundary", "lang": "en"}
{"prompt": "train.py and eval.py each build the whole dataloader from scratch and they've drifted. one shared builder, both scripts use it, metrics should come out identical", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "our unity managers are all static singletons which makes the test scene impossible. put interfaces in front of them and inject through the bootstrapper", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "the pipeline configs use snake_case for column names in some yamls and camelCase in others. normalize on snake everywhere including the dbt models that reference them", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "gpu sits at 12% the whole training run, num_workers is 8, prefetch is on, i've been staring at nsight for two hours and cant find the stall", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "why is the mesh flickering", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "spark job dies on the join stage, third time today:\n\n25/07/29 03:12:44 WARN TaskSetManager: Lost task 118.0 in stage 7.0 (TID 4412) (10.0.7.44 executor 12): ExecutorLostFailure (executor 12 exited caused by one of the running tasks) Reason: Container killed by YARN for exceeding physical memory limits. 28.4 GB of 28 GB physical memory used. Consider boosting spark.executor.memoryOverhead.\n25/07/29 03:12:51 INFO DAGScheduler: Resubmitting ShuffleMapStage 6 (mapToPair) due to fetch failure\n25/07/29 03:14:09 WARN BlockManagerMasterEndpoint: No more replicas available for shuffle_3_118_0\n25/07/29 03:16:22 ERROR TaskSetManager: Task 118 in stage 7.0 failed 4 times; aborting job\n\nspark.sql.shuffle.partitions is 2000 already. the input only grew ~5% week over week so this feels like something else", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "pasted-context", "lang": "en"}
{"prompt": "explain how our terraform state locking works with that dynamodb table, and what actually happens if two applies race", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "is it expected that the physics step runs twice on the first frame after a scene load? trying to tell if that's by design in our bootstrap or not", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "boundary", "lang": "en"}
{"prompt": "look over the IAM policy attached to the CI role and tell me what it can do that it shouldn't", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "runbook for when the nightly training run fails — how to tell if it's data or infra, how to resume from the last checkpoint, who to page", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "document the /predict endpoint contract properly — input tensor shapes, the accepted dtypes, every error code and what triggers it", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "boundary", "lang": "en"}
{"prompt": "patch notes for the mod tools update, player-facing, upbeat but not cringe", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "same thing but for the eu region", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "vague-eval", "lang": "en"}
{"prompt": "i need a plan for getting off our homegrown migrations tool. 340 migration files, some hand-edited in prod, and two databases that have diverged. tell me how you'd sequence it and what we do about the drift", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.9, "slice": "core", "lang": "en"}
{"prompt": "roadmap please, one quarter, for making the cli usable offline. today every command hits the api", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "diseña el flujo de invitaciones para equipos: enlaces con expiración, roles preasignados, y qué pasa si el email ya tiene cuenta en otra org. sólo el diseño por ahora", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "es"}
{"prompt": "could you add a websocket presence channel — who's viewing a document right now, heartbeat every 20s, cleanup on disconnect, and it should survive a server restart without ghost users", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "sqlite fts5 index for the local notes cache in the mac app, incremental update on save", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "job that expires trial accounts at 14 days, sends the two reminder emails at day 10 and 13, and never double sends even if it runs twice", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "settings panel needs a two column layout on desktop, single column under 768. the section nav on the left should be sticky and highlight the section you're in", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "bar chart for weekly active devices, hover tooltip with exact counts, and the y axis shouldn't start at zero-ish weirdness like it does now", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "the badge on the notifications icon is 1px off center, and it should hide entirely at zero instead of showing '0'", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "boundary", "lang": "en"}
{"prompt": "mypy is red after i added the protocol, probably five minutes of work:\n\nsrc/acme/store/redis_store.py:31: error: Signature of \"get\" incompatible with supertype \"KeyValueStore\" [override]\nsrc/acme/store/redis_store.py:31: note: Superclass:\nsrc/acme/store/redis_store.py:31: note: def get(self, key: str) -> bytes | None\nsrc/acme/store/redis_store.py:31: note: Subclass:\nsrc/acme/store/redis_store.py:31: note: def get(self, key: str, default: bytes = ...) -> bytes\nsrc/acme/store/memory_store.py:19: error: Missing return statement [return]\nsrc/acme/api/deps.py:57: error: Argument 1 to \"Depends\" has incompatible type \"Callable[[Request], Coroutine[Any, Any, KeyValueStore]]\"; expected \"Callable[..., KeyValueStore]\" [arg-type]\nFound 3 errors in 3 files (checked 214 source files)", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "pasted-context", "lang": "en"}
{"prompt": "sentry dsn is hardcoded in main.ts, move it to an env var", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "prometheus scrape interval 15s -> 30s, we're drowning in cardinality anyway", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.1, "slice": "core", "lang": "en"}
{"prompt": "the api handlers each do their own auth check with slightly different logic. one middleware, applied at the router, identical decisions to what we have now — audit them first so nothing gets accidentally opened up", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "our css is 40% dead. we have tailwind AND a legacy sass bundle. consolidate onto tailwind, screens should render pixel identical", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "pull the magic numbers in the gameplay tuning code into a config asset. same values, designers should be able to edit them without a rebuild", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "search suggestions endpoint got 4x slower after we shipped last week. p50 fine, p99 terrible. no query changes in the diff that i can see", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "the android release build crashes at startup but debug is fine. proguard i assume but the mapping file isnt telling me much", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "prometheus fired this at 4am and i can't reproduce it now:\n\nALERT: KafkaConsumerLagGrowing\nSTATE: firing (7m)\nlabels:\n consumergroup=enrichment-v3\n topic=events.raw\n partition=11\n severity=warning\nannotations:\n summary=consumer lag 1.2M and climbing on partition 11 only\n runbook=https://wiki/int/kafka-lag\nvalue: 1214883\n\nother 23 partitions were at basically zero lag the whole time. rebalance count didn't change. what makes ONE partition fall behind like that", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "pasted-context", "lang": "en"}
{"prompt": "what happens to in-flight requests when we scale down? read the shutdown path and tell me if we actually drain", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "boundary", "lang": "en"}
{"prompt": "explain the diff between our two dockerfiles, dev and prod, i can't tell which differences are intentional", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "core", "lang": "en"}
{"prompt": "sanity check my understanding of the batching in the embedding service before i build on it — where does a request wait, and what's the worst case latency for a single-item request", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "we need an internal doc explaining the difference between our three environments and which one is safe to break. nobody knows and people keep asking in slack", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "summarize the design tradeoffs already captured in the comments of consensus.rs into a proper doc — i'd rather have it in one place than scattered across 900 lines", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "plan the move to server components and then convert the marketing pages as the first batch", "purpose": "planning", "secondary": "frontendImpl", "mixed": true, "difficulty": 0.8, "slice": "mixed", "lang": "en"}
{"prompt": "figure out why the dedupe is dropping legit rows and once you know, write the ticket update explaining it", "purpose": "debugging", "secondary": "writing", "mixed": true, "difficulty": 0.7, "slice": "mixed", "lang": "en"}
{"prompt": "finish it up", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "vague-eval", "lang": "en"}
{"prompt": "thinking about our on-call load. 60% of pages are the same three alerts and none of them are actionable. i want a plan to fix the alerting itself, not just tune thresholds", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "how would you structure a plugin api for the editor? third parties, sandboxed, versioned, and we don't want to be stuck supporting whatever we ship forever. write the proposal", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.9, "slice": "core", "lang": "en"}
{"prompt": "milestones for the postgres 14 -> 17 upgrade, logical replication, zero downtime, we have 4TB and one maintenance window a month", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "oauth device flow for the cli so people on headless boxes can log in. poll interval respected, and store the token in the os keyring not a dotfile", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "bloom filter in front of the object existence check, tune for 1% fp at 50M keys, rebuilt nightly", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "the swagger page documents a `fields` query param for sparse fieldsets on every resource. that param does nothing. implement it for real, at least on users and projects", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "boundary", "lang": "en"}
{"prompt": "diff viewer component — side by side, syntax highlighting, collapsible unchanged regions, line links. it's the last big thing before we can dogfood", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "the tab order on the invite modal is chaos and the close button isn't reachable by keyboard at all", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "core", "lang": "en"}
{"prompt": "put a subtle divider between the list sections, right now they run together", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "boundary", "lang": "en"}
{"prompt": "the retention env var says RETENTION_DAYS but the code reads RETAIN_DAYS, align them", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "swap the deprecated `substr` calls for `slice` in the two spots eslint flags", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.1, "slice": "boundary", "lang": "en"}
{"prompt": "cache-control on the static assets is 300s, make it a year with the hashed filenames", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "our error types are strings compared with ==. introduce a proper error enum and update every comparison, same control flow though", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "move the shared types out of the api package into a types package so the frontend stops importing server code", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "collapse UserService and AccountService, they're the same concept split by an old naming argument. keep the public methods that are actually called", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "boundary", "lang": "en"}
{"prompt": "some uploads finish but the file is 0 bytes. maybe 1 in 300. no error anywhere, the presigned put returns 200", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "xcodebuild fails only on CI:\n\n❌ /Users/runner/work/app/Sources/Sync/SyncEngine.swift:212:34: error: sending 'self.pendingOps' risks causing data races\n 212 | await withTaskGroup { group in\n 213 | for op in pendingOps {\n | `- note: sending task-isolated 'self.pendingOps' to nonisolated callee risks causing data races\n\n❌ /Users/runner/work/app/Sources/Sync/SyncEngine.swift:240:9: error: main actor-isolated property 'lastSyncedAt' can not be mutated from a nonisolated context\n\n** BUILD FAILED **\n\nThe following build commands failed:\n\tSwiftCompile normal arm64 Compiling SyncEngine.swift\n(1 failure)\n\nlocal xcode builds clean, same swift version supposedly. what's different", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "pasted-context", "lang": "en"}
{"prompt": "scroll position jumps to the top when the list refetches, but only sometimes, and only in the embedded webview", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "code review on the branch — mostly want your eyes on the transaction boundaries in the transfer handler", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "why do we set both `SameSite=Lax` and check the origin header? one of those seems redundant, is it", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "trace what happens end to end when someone clicks 'export' — i want every hop, including the lambda i keep forgetting about", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "api reference page for the rust crate — module overview, the three main entry points, and a worked example that compiles", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "release notes for the desktop app 5.2, and please mention the sparkle update fix prominently, people have been complaining", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "commit messages for these 6 commits i'm about to rebase, they all say 'wip'", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "ok so", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "vague-eval", "lang": "en"}
{"prompt": "wir müssen die Suche von Elasticsearch auf Typesense migrieren. schreib mir einen Migrationsplan mit Zwischenschritten, wir können nicht alles auf einmal umstellen", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "de"}
{"prompt": "design doc for background sync on the ios app — what we sync, when, how we handle the 30 second bg task budget, and conflict resolution", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "we have three services that each keep their own copy of the org hierarchy and they disagree. i want a strategy, not a patch", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "boundary", "lang": "en"}
{"prompt": "csv -> parquet converter cli in rust, streaming, schema inference with an override flag, progress bar to stderr", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "audit trail on every write to the settings table — who, when, before and after, using a trigger so nobody can bypass it from the app", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "graphql subscription for build status, redis pubsub behind it, and clients that reconnect should get the current state immediately not just future events", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "sortable columns on the results table — click to sort, shift click for secondary, and the sort should persist in the url", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "iPad layout: the split view collapses to a single column when it shouldn't, and the detail pane loses its scroll position on rotation", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "focus ring is invisible on our brand blue buttons, needs an offset outline that works on both", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "boundary", "lang": "en"}
{"prompt": "add `--json` to the status command, same data, machine readable", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "boundary", "lang": "en"}
{"prompt": "the copyright year in the footer is 2024", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.1, "slice": "core", "lang": "en"}
{"prompt": "set the k8s cpu request on the api deployment to 500m, it's currently unset and the scheduler is packing them badly", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "the ViewController is doing layout, networking, and analytics. break it apart following the pattern in the newer screens. no visual or behavioral change, screenshots should be identical", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "we have `getUser`, `fetchUser`, `loadUser` and `userById` and they all do the same thing with different caching. pick one, migrate everything, delete the rest", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "boundary", "lang": "en"}
{"prompt": "invert the dependency between the parser and the diagnostics collector so the parser stops knowing about our error formatting", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "customers on the enterprise plan report the export button does nothing. no js errors in their console, works for every account i test with", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "esp32 reboots every ~40 minutes, watchdog i assume, but the logs just stop:\n\nI (2412088) mqtt: publishing telemetry seq=4409 len=212\nI (2412094) mqtt: publish ok\nI (2417088) sensor: read ok t=22.4 h=41.2\nI (2422088) mqtt: publishing telemetry seq=4410 len=212\nRST:0x8 (TG1WDT_SYS_RST),boot:0x13 (SPI_FAST_FLASH_BOOT)\nconfigsip: 0, SPIWP:0xee\nload:0x3fff0030,len:4832\nentry 0x400806a8\nI (31) boot: ESP-IDF v5.2.1 2nd stage bootloader\n\nheap looks stable in the logs before it goes, ~84k free the whole time. TG1WDT is the task watchdog on timer group 1 right? nothing i wrote uses that timer directly", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "pasted-context", "lang": "en"}
{"prompt": "il y a une fuite mémoire quelque part dans le worker d'images, la RSS monte jusqu'à l'OOM en 6h. pas de fuite évidente côté go pprof", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "fr"}
{"prompt": "is this migration safe to run during business hours", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "compare the two approaches sitting in the repo for the pdf renderer — the puppeteer one and the typst one — and tell me which i should keep", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "boundary", "lang": "en"}
{"prompt": "read scripts/deploy.sh line by line and tell me what it does if the health check never passes", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "core", "lang": "en"}
{"prompt": "write the doc comments for our public swift package api, docc style, with a short discussion section on each type explaining when you'd use it", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "internal announcement post about the new deploy process. what changed, what people need to do differently, keep it under 300 words", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "translate the onboarding emails to ja, they're in emails/onboarding/*.mjml, keep the liquid tags intact", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "core", "lang": "en"}
{"prompt": "design the multi-region failover and then implement the health check propagation part of it", "purpose": "planning", "secondary": "backendImpl", "mixed": true, "difficulty": 0.9, "slice": "mixed", "lang": "en"}
{"prompt": "rename the Tenant model to Workspace everywhere and update the docs that reference it", "purpose": "refactor", "secondary": "writing", "mixed": true, "difficulty": 0.6, "slice": "mixed", "lang": "en"}
{"prompt": "handle it", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "vague-eval", "lang": "en"}
{"prompt": "i keep going back and forth on whether the sync engine should be crdt based or just server-authoritative with a merge ui. write up both, be opinionated at the end", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.9, "slice": "core", "lang": "en"}
{"prompt": "what's the migration path from our current session cookies to passkeys, with a period where both work", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "outline the test strategy for the payment integration — what's unit, what hits their sandbox, what we record and replay", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "implement optimistic concurrency on the document update endpoint, version column, 409 with the current version on conflict", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "connection pooling for the mysql driver in our go service, plus a circuit breaker that trips on repeated timeouts and half-opens after 30s", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "s3 multipart upload with resume from a local manifest, 8mb parts, 4 in flight", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "build the comparison table for the pricing page, 3 tiers, feature rows with checkmarks and tooltips, sticky header when you scroll past it, and it has to be readable on a phone", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "custom shader for the water surface, cheap — mobile targets. gerstner waves and a fresnel-ish edge foam", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "hide the beta ribbon on the dashboard, we launched", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.1, "slice": "boundary", "lang": "en"}
{"prompt": "content-security-policy is missing frame-ancestors, add 'none'", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "the seed script inserts 10 users, make it 1000 so the pagination is actually exercised locally", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "increase the sqs visibility timeout to 5 min, the handler takes longer than 30s now", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "the test suite has 40 helpers that build fixtures inline. introduce factories, port the tests, same assertions", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "our redux slices reach into each other's state directly. put selectors at the boundaries so we can move things later. behavior identical", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "s/AcmeCorpInternalHttpClientWrapper/HttpClient/g basically, but do it properly including the file names and the DI registration", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "boundary", "lang": "en"}
{"prompt": "two of our three api pods serve stale feature flags after a config push. the third is fine. same image, same env", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "playwright test is flaking on the login step, i've pasted the trace summary:\n\n 1) [chromium] › auth/login.spec.ts:24:5 › logs in with email ─────────────────\n\n TimeoutError: locator.click: Timeout 15000ms exceeded.\n Call log:\n - waiting for getByRole('button', { name: 'Continue' })\n - locator resolved to <button disabled type=\"submit\">Continue</button>\n - attempting click action\n - waiting for element to be enabled\n - element is not enabled - waiting...\n (repeated 74 times)\n\n at auth/login.spec.ts:31:52\n\n Retry #1 ─ passed (4.2s)\n\n 1 flaky, 89 passed (2.1m)\n\nthe button is disabled until the email validates, which is debounced 300ms. but 15 seconds?", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "pasted-context", "lang": "en"}
{"prompt": "requests from the mobile app get a 401 about 2% of the time right after a token refresh. clock skew maybe? not sure", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "how do we currently decide which shard a tenant lands on", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "core", "lang": "en"}
{"prompt": "the new caching decorator in api/decorators.py — is it going to cache per-user data across users? that's my worry, tell me if the key covers it", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "boundary", "lang": "en"}
{"prompt": "give me an honest assessment of the test coverage in the billing module. not the number, the actual risk", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "our http error responses aren't documented anywhere. write the reference — status, code string, when it happens, whether retrying helps", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "boundary", "lang": "en"}
{"prompt": "quickstart for the sdk, 5 minutes to first successful call, node and python", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "core", "lang": "en"}
{"prompt": "escreve a descrição do PR, tem uma mudança de comportamento no cálculo de imposto que precisa ficar bem clara", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "pt"}
{"prompt": "you know what needs doing", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "vague-eval", "lang": "en"}
{"prompt": "propose how we'd support self-hosted installs. licensing, telemetry opt-out, upgrade path, what we cut from cloud-only features. i need something i can show the exec team", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.9, "slice": "core", "lang": "en"}
{"prompt": "we're at 900 e2e tests and 50 minutes of CI. plan how we get under 15 without deleting coverage", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "sketch out an api design for bulk operations. i keep wanting to just add /bulk endpoints per resource but that feels wrong", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "boundary", "lang": "en"}
{"prompt": "server-sent events endpoint for the activity feed, resumable with Last-Event-ID, heartbeat comment every 15s to keep proxies happy", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "write the migration + backfill to split the `name` column into given/family. 4M rows, chunked, and keep `name` as a generated column during the transition", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "add a `serve` command to the cli that hosts the built site with live reload, pick a free port if 3000 is taken", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "photo grid with lazy loaded thumbnails, 3 across on phone, and tapping should zoom out of the tapped cell not cross-fade", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "our forms have no inline validation, everything blows up on submit. add field-level errors on blur across the signup flow", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "bump the modal max-width to 640, the form is squeezed", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.1, "slice": "boundary", "lang": "en"}
{"prompt": "we're logging the full request body including passwords on 400s. strip it, today", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "the rust edition in Cargo.toml is 2018, move to 2021 and fix the fallout", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "boundary", "lang": "en"}
{"prompt": "the makefile's `test` target doesn't run the integration tests, add them", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "hoist the shared validation out of the three signup paths (web, invite accept, sso jit) into one function, same rules, same error messages", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "the python package has modules importing from __init__ which imports from them. straighten out the import graph without changing the public surface", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "our sql lives in string literals scattered through the handlers. move it to .sql files and generate the accessors, identical queries", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "one customer's webhooks stopped delivering three days ago. our logs say 200 from their endpoint, they say they never got them", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "cpu pegged at 100% on one core in the render thread, profiler blames a function that should be trivial", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "ansible run bombs partway through, hosts end up half configured which is the worst outcome:\n\nTASK [postgres : ensure replication slot exists] *******************************\nfatal: [db-03.prod]: FAILED! => {\"changed\": false, \"msg\": \"unable to connect to database: connection to server at \\\"10.4.1.22\\\", port 5432 failed: FATAL: remaining connection slots are reserved for non-replication superuser connections\"}\n\nTASK [postgres : reload config] ***********************************************\nskipping: [db-03.prod]\n\nTASK [pgbouncer : write userlist] *********************************************\nchanged: [db-01.prod]\nchanged: [db-02.prod]\n\nPLAY RECAP ********************************************************************\ndb-01.prod : ok=41 changed=6 unreachable=0 failed=0\ndb-02.prod : ok=41 changed=6 unreachable=0 failed=0\ndb-03.prod : ok=28 changed=3 unreachable=0 failed=1\n\nmax_connections is 400 and we were at maybe 190 when this ran", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "pasted-context", "lang": "en"}
{"prompt": "walk me through the lock ordering in the scheduler, i want to know if a deadlock is possible before i add another lock", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "these two functions look like they should be the same but return different things for empty input. which one is right?", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "review the terraform plan diff before i apply, i'm nervous about the security group changes:\n\n # aws_security_group_rule.api_ingress will be destroyed\n - resource \"aws_security_group_rule\" \"api_ingress\" {\n - cidr_blocks = [\"10.0.0.0/8\"] -> null\n - from_port = 443 -> null\n - protocol = \"tcp\" -> null\n - security_group_id = \"sg-0a91f\" -> null\n - type = \"ingress\" -> null\n }\n\n # aws_security_group_rule.api_ingress_v2 will be created\n + resource \"aws_security_group_rule\" \"api_ingress_v2\" {\n + cidr_blocks = [\"0.0.0.0/0\"]\n + from_port = 443\n + protocol = \"tcp\"\n + security_group_id = \"sg-0a91f\"\n + type = \"ingress\"\n }\n\n # aws_lb_listener.https will be updated in-place\n ~ resource \"aws_lb_listener\" \"https\" {\n ~ ssl_policy = \"ELBSecurityPolicy-2016-08\" -> \"ELBSecurityPolicy-TLS13-1-2-2021-06\"\n }\n\nPlan: 1 to add, 1 to change, 1 to destroy.", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "pasted-context", "lang": "en"}
{"prompt": "the graphql schema has no descriptions on anything. add them, field by field, and be accurate about nullability semantics", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "summarize this week's merged PRs into a short changelog for the team channel", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "write up the security disclosure policy for the repo, SECURITY.md, include the response time we're committing to", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "figure out the plan for the search relevance work then knock out the tokenizer change first", "purpose": "planning", "secondary": "backendImpl", "mixed": true, "difficulty": 0.8, "slice": "mixed", "lang": "en"}
{"prompt": "explain what our rate limiter does today and then write the doc for it, we have neither", "purpose": "review", "secondary": "writing", "mixed": true, "difficulty": 0.5, "slice": "mixed", "lang": "en"}
{"prompt": "carry on with the widget stuff", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "vague-eval", "lang": "en"}
{"prompt": "我们要把日志系统从 Loki 换成 ClickHouse，先写个方案：成本、查询延迟、迁移期间双写怎么做", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "zh"}
{"prompt": "break down what it would take to support real-time collaboration on our forms builder. i want phases where each one ships something users notice", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.9, "slice": "core", "lang": "en"}
{"prompt": "plan the api versioning scheme before we have customers, i don't want to be stuck", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "implement scim user provisioning, users and groups, patch semantics done properly because okta is strict about it", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "distributed lock helper over redis with a fencing token, and make the api hard to misuse — no way to hold it past the ttl silently", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "the reference docs describe cursor pagination on /events with a `before` param. the handler only supports `after`. add `before`", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "command palette needs a recent-files section and file icons by extension. match the visual density of the sidebar", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "the swiftui list scrolls like garbage with 2000 rows, rows have async images. make it smooth", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "boundary", "lang": "en"}
{"prompt": "onboarding checklist card on the dashboard — 4 steps, progress ring, collapses once complete, dismissible", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "trailing whitespace in the generated file breaks our lint, add a strip", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "the retry decorator doesn't retry on 429, add it to the retryable set", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "pyproject requires-python says >=3.9 but we use match statements, so >=3.10", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.1, "slice": "core", "lang": "en"}
{"prompt": "our components take 14 props each because we kept adding variants. introduce composition — slots instead of booleans — and migrate the call sites without changing what renders", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "move everything under src/util into named modules, util is a graveyard. no behavior change, keep a shim for one release", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "the emailer takes 9 positional args. make it a struct/options object, update the 30 call sites", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "since the node 22 upgrade our streaming responses get truncated at exactly 65536 bytes sometimes. only through the cdn, direct to origin is fine", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "the numbers in the finance report are off by a few cents for about 1 in 200 invoices and finance is not happy", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "dbt run failed overnight, here's the tail:\n\n05:12:44 1 of 84 START sql incremental model analytics.fct_sessions ..... [RUN]\n05:14:02 1 of 84 ERROR creating sql incremental model analytics.fct_sessions [ERROR in 78.1s]\n05:14:02 Database Error in model fct_sessions (models/marts/fct_sessions.sql)\n05:14:02 Cannot insert a NULL value into column: 'session_key' in table 'FCT_SESSIONS'\n05:14:02 compiled Code at target/run/marts/fct_sessions.sql\n05:14:09 16 of 84 SKIP relation analytics.fct_engagement ................ [SKIP]\n05:14:09 Done. PASS=61 WARN=0 ERROR=1 SKIP=22 TOTAL=84\n\nsession_key is a surrogate key from dbt_utils.generate_surrogate_key over (user_id, session_start). so one of those is null now, which shouldn't be possible upstream", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "pasted-context", "lang": "en"}
{"prompt": "does our soft delete actually prevent deleted records from showing up in search? i see the filter in the query builder but not in the reindex job", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "boundary", "lang": "en"}
{"prompt": "what's the blast radius if the config service is down", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "read through the new contributor's PR and give me a review i can paste, be kind but thorough, they're a first timer", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "changelog + upgrade notes for the breaking change in the config format. include a before/after yaml", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "core", "lang": "en"}
{"prompt": "explain the caching headers we set, in prose, for the docs site's performance page", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "boundary", "lang": "en"}
{"prompt": "the architecture diagram in the wiki is 2 years stale. write an updated description of the system, i'll draw from that", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "nope", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "vague-eval", "lang": "en"}