162 lines
46 KiB
JSON
162 lines
46 KiB
JSON
{"prompt":"Design a shared feature-store architecture for fraud, ranking, and support models. Cover offline/online consistency, feature ownership, point-in-time correctness, freshness tiers, backfills, schema evolution, regional availability, and a phased migration from team-owned Redis tables. Deliver an architecture proposal only.","purpose":"planning","secondary":null,"mixed":false,"difficulty":0.91,"slice":"core","lang":"en"}
|
|
{"prompt":"Outline the rollout for signed OTA firmware updates across 180,000 intermittently connected devices. Include key rotation, staged cohorts, downgrade protection, interrupted flashes, health gates, and recovery through the factory image; no code yet.","purpose":"planning","secondary":null,"mixed":false,"difficulty":0.9,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Add zstd support to the trace ingester","purpose":"backendImpl","secondary":null,"mixed":false,"difficulty":0.4,"slice":"core","lang":"en"}
|
|
{"prompt":"Inference incident excerpt:\n\n```text\nmodel=churn-v17 runtime=onnxruntime-gpu\nbatch=128 input_checksum=69f204\nnode-a output_checksum=5c1ab8 top1=retain score=0.5031\nnode-b output_checksum=b70de2 top1=churn score=0.5028\nnode-a gpu=A10 driver=580.65 cuda=13.0\nnode-b gpu=L4 driver=580.65 cuda=13.0\npreprocessor image=features:2026.07.28\nORT graph optimization=all\nTensor `account_age_days`: min=-1 max=4412 dtype=float32\nTensor `recent_sessions`: min=0 max=32767 dtype=int16\nWARNING Cast_91 saturating conversion differs from reference evaluator\n```\n\nBoth nodes use the same model artifact and input checksum, but about 0.4% of borderline predictions disagree. Determine whether preprocessing, numeric kernels, or graph optimization causes the divergence and identify what evidence would isolate it.","purpose":"debugging","secondary":null,"mixed":false,"difficulty":0.9,"slice":"pasted-context","lang":"en"}
|
|
{"prompt":"Build the in-game squad HUD with health rings, voice activity, status effects, controller focus, and a compact spectator variant.","purpose":"frontendImpl","secondary":null,"mixed":false,"difficulty":0.62,"slice":"core","lang":"en"}
|
|
{"prompt":"Plan and implement a short-lived credential broker for CI jobs. Begin with a design covering workload identity, policy evaluation, token lifetimes, audit trails, revocation, regional failure, and migration from static repository secrets. After the design is reviewable, add the Go service, Kubernetes authentication, provider adapters, metrics, and integration tests. Treat the security model and rollout plan as the primary outcome.","purpose":"planning","secondary":"backendImpl","mixed":true,"difficulty":0.96,"slice":"mixed","lang":"en"}
|
|
{"prompt":"Set `shuffle_buffer` to 50000","purpose":"quickFix","secondary":null,"mixed":false,"difficulty":0.07,"slice":"core","lang":"en"}
|
|
{"prompt":"Separate feature extraction from model invocation in `ScoringPipeline` while preserving tensor names, batching, timeouts, metrics, and prediction output exactly.","purpose":"refactor","secondary":null,"mixed":false,"difficulty":0.67,"slice":"core","lang":"en"}
|
|
{"prompt":"Expose the active model metadata at `/internal/model-info`, including artifact digest, feature-schema version, and load timestamp, with tests for the JSON contract.","purpose":"backendImpl","secondary":null,"mixed":false,"difficulty":0.42,"slice":"boundary","lang":"en"}
|
|
{"prompt":"make the battle screen cooler","purpose":"frontendImpl","secondary":null,"mixed":false,"difficulty":0.44,"slice":"vague-eval","lang":"en"}
|
|
{"prompt":"Find why the STM32 bootloader jumps into the application with an invalid stack pointer only after a watchdog reset.","purpose":"debugging","secondary":null,"mixed":false,"difficulty":0.79,"slice":"core","lang":"en"}
|
|
{"prompt":"Explain how `ShardPlanner` balances hot embeddings across GPU replicas and whether its scoring function can strand capacity. Review only.","purpose":"review","secondary":null,"mixed":false,"difficulty":0.66,"slice":"core","lang":"en"}
|
|
{"prompt":"Raw notes from the platform channel:\n\n```text\nTopic: model rollback procedure\nCurrent deployment: 10% shadow, 5% serving canary, then regional ramp\nArtifacts live in `ml-artifacts-prod`; manifests are signed\nRollback means switching the routing alias, not deleting the bad version\nFeature schema can advance independently, but only one version backward is guaranteed\nOnline cache entries include model version but not preprocessor digest\nDashboards: `Inference / Quality`, `Inference / Saturation`\nOn-call concern: quality labels arrive 4-12 hours late\nSupport concern: customers ask which predictions used the reverted version\nCompliance requires preserving decision lineage for 7 years\n```\n\nTurn this into an operator runbook for rolling back a model release. Include decision criteria, exact verification categories, delayed-label handling, cache considerations, lineage checks, and post-rollback communication.","purpose":"writing","secondary":null,"mixed":false,"difficulty":0.59,"slice":"pasted-context","lang":"en"}
|
|
{"prompt":"Nudge the minimap legend below the safe-area inset","purpose":"frontendImpl","secondary":null,"mixed":false,"difficulty":0.14,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Write the integration guide for publishing custom metrics through the edge agent, including supported types, label limits, buffering, retries, and sample configuration.","purpose":"writing","secondary":null,"mixed":false,"difficulty":0.38,"slice":"core","lang":"en"}
|
|
{"prompt":"Implement the dataset snapshot endpoint with filtering, signed download URLs, and retention enforcement, then document its request parameters, asynchronous states, examples, and failure responses for SDK authors.","purpose":"backendImpl","secondary":"writing","mixed":true,"difficulty":0.72,"slice":"mixed","lang":"en"}
|
|
{"prompt":"Map out a migration from Jenkins-hosted game builds to ephemeral Buildkite agents, covering asset caches, console SDK isolation, signing, capacity spikes, observability, and rollback milestones.","purpose":"planning","secondary":null,"mixed":false,"difficulty":0.86,"slice":"core","lang":"en"}
|
|
{"prompt":"Change `max_retrys` to `max_retries` in the unreleased config schema.","purpose":"quickFix","secondary":null,"mixed":false,"difficulty":0.09,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Implement deterministic replay files for the physics server with versioned inputs, checksums, and divergence reporting.","purpose":"backendImpl","secondary":null,"mixed":false,"difficulty":0.76,"slice":"core","lang":"en"}
|
|
{"prompt":"Polish the crafting inventory grid","purpose":"frontendImpl","secondary":null,"mixed":false,"difficulty":0.31,"slice":"core","lang":"en"}
|
|
{"prompt":"Bump `prometheus-operator` to `v0.85.0`","purpose":"quickFix","secondary":null,"mixed":false,"difficulty":0.08,"slice":"core","lang":"en"}
|
|
{"prompt":"Rename `MasterNode` to `CoordinatorNode` across the scheduler, protobufs, dashboards, and tests while preserving wire compatibility and runtime behavior.","purpose":"refactor","secondary":null,"mixed":false,"difficulty":0.68,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Split the monolithic `TrainingJobController` into validation, scheduling, status, and cleanup collaborators without changing its Kubernetes API behavior.","purpose":"refactor","secondary":null,"mixed":false,"difficulty":0.74,"slice":"core","lang":"en"}
|
|
{"prompt":"Work item copied from the infrastructure backlog:\n\n```text\nNET-304 — reconcile BGP advertisements from Kubernetes Services\nInput: Services labeled `edge.example.com/advertise=true`\nOutput: prefixes announced through two route reflectors\nRules:\n- only LoadBalancer IPs from configured pools\n- withdraw within 20s after Service deletion\n- duplicate events are expected\n- retain last accepted router generation in Postgres\n- reject an update that would advertise more than 256 prefixes per tenant\n- dry-run mode emits proposed changes but touches no routers\n- expose reconcile duration, rejected prefix count, and router generation lag\nFailure policy: retry transient gRPC failures; configuration errors go to status conditions\n```\n\nImplement the controller, persistence, router client calls, status updates, limits, metrics, and integration tests.","purpose":"backendImpl","secondary":null,"mixed":false,"difficulty":0.88,"slice":"pasted-context","lang":"en"}
|
|
{"prompt":"Investigate why Unreal Engine streaming levels intermittently remain invisible after their assets report loaded.","purpose":"debugging","secondary":null,"mixed":false,"difficulty":0.72,"slice":"core","lang":"en"}
|
|
{"prompt":"Create the GPU allocation heatmap and write a short legend explaining capacity, reservations, fragmentation, and unavailable devices. The view needs zoom, time-range selection, accessible values, and a useful empty state.","purpose":"frontendImpl","secondary":"writing","mixed":true,"difficulty":0.67,"slice":"mixed","lang":"en"}
|
|
{"prompt":"Why does the telemetry collector quantize timestamps to 8-millisecond buckets? Explain the existing behavior from code and tests; nothing is broken.","purpose":"review","secondary":null,"mixed":false,"difficulty":0.44,"slice":"core","lang":"en"}
|
|
{"prompt":"Draft a setup page for running the emulator farm locally with fake devices and a MinIO-backed artifact store.","purpose":"writing","secondary":null,"mixed":false,"difficulty":0.33,"slice":"core","lang":"en"}
|
|
{"prompt":"After moving the agent to cgroup v2, this integration test started hanging:\n\n```text\n=== RUN TestMemoryPressureTriggersEviction\nfixture: limit=256MiB high=192MiB\nworkload: allocated 220MiB\nmemory.events before: high 0 oom 0 oom_kill 0\nwaiting for watcher notification...\ncontext deadline exceeded after 45s\nmemory.current=230776832\nmemory.high=201326592\nmemory.events after: high 184 oom 0 oom_kill 0\nwatcher epoll fd=17 events=0\n```\n\nDetermine why the kernel counter changes without our watcher waking. I do not know whether the bug is event subscription, file descriptor setup, or an incorrect assumption carried over from cgroup v1.","purpose":"debugging","secondary":null,"mixed":false,"difficulty":0.81,"slice":"boundary","lang":"en"}
|
|
{"prompt":"finish the controller stuff","purpose":"backendImpl","secondary":null,"mixed":false,"difficulty":0.48,"slice":"vague-eval","lang":"en"}
|
|
{"prompt":"Propose a storage design for petabyte-scale training checkpoints, including chunking, deduplication, consistency, lifecycle rules, regional recovery, and cost checkpoints.","purpose":"planning","secondary":null,"mixed":false,"difficulty":0.93,"slice":"core","lang":"en"}
|
|
{"prompt":"Add support for loading GGUF tensor metadata without materializing tensor contents.","purpose":"backendImpl","secondary":null,"mixed":false,"difficulty":0.56,"slice":"core","lang":"en"}
|
|
{"prompt":"Change request under review:\n\n```diff\n func selectReplica(key string, replicas []Replica) Replica {\n- return replicas[xxhash.Sum64String(key)%uint64(len(replicas))]\n+ best := replicas[0]\n+ bestScore := uint64(0)\n+ for _, r := range replicas {\n+ score := xxhash.Sum64String(key + r.ID)\n+ if score > bestScore {\n+ best, bestScore = r, score\n+ }\n+ }\n+ return best\n }\n```\n\nThe author says rendezvous hashing will reduce cache churn during scaling. Review this implementation for correctness, distribution, empty-input behavior, stable identifiers, and what happens when replica weights differ. Do not edit the patch.","purpose":"review","secondary":null,"mixed":false,"difficulty":0.68,"slice":"pasted-context","lang":"en"}
|
|
{"prompt":"编写一份故障排查文档，说明 GPU worker 无法注册时要检查的证书、NVIDIA runtime、网络策略和日志。不要修改代码。","purpose":"writing","secondary":null,"mixed":false,"difficulty":0.4,"slice":"boundary","lang":"zh"}
|
|
{"prompt":"Build the spectator timeline with event markers, scrub previews, bookmarks, and controller-friendly navigation.","purpose":"frontendImpl","secondary":null,"mixed":false,"difficulty":0.6,"slice":"core","lang":"en"}
|
|
{"prompt":"Review the existing autoscaler decision code, explain its stabilization windows, and turn the explanation into an operator-facing page with worked scale-up and scale-down examples.","purpose":"writing","secondary":"review","mixed":true,"difficulty":0.62,"slice":"mixed","lang":"en"}
|
|
{"prompt":"Pin the base image digest in `worker.Dockerfile`","purpose":"quickFix","secondary":null,"mixed":false,"difficulty":0.08,"slice":"core","lang":"en"}
|
|
{"prompt":"把 `WorkerPool` 在整个仓库中改名为 `ExecutorPool`，包括文件名、测试和指标前缀；除了名称以外，行为必须保持不变。","purpose":"refactor","secondary":null,"mixed":false,"difficulty":0.59,"slice":"boundary","lang":"zh"}
|
|
{"prompt":"Consolidate the CUDA capability checks used by the trainer and serving runtime without altering device eligibility.","purpose":"refactor","secondary":null,"mixed":false,"difficulty":0.55,"slice":"core","lang":"en"}
|
|
{"prompt":"Diagnose the missing final audio buffer","purpose":"debugging","secondary":null,"mixed":false,"difficulty":0.48,"slice":"core","lang":"en"}
|
|
{"prompt":"Explain how the rollout controller decides a canary is healthy when latency improves but the error budget worsens.","purpose":"review","secondary":null,"mixed":false,"difficulty":0.57,"slice":"core","lang":"en"}
|
|
{"prompt":"Write an API reference for the existing model-evaluation endpoint, including multipart uploads, metric selection, asynchronous status polling, limits, and result retention.","purpose":"writing","secondary":null,"mixed":false,"difficulty":0.43,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Plan a transition from mutable model tags to immutable content-addressed references. Cover registry compatibility, CI changes, cache behavior, lineage, rollout gates, and rollback.","purpose":"planning","secondary":null,"mixed":false,"difficulty":0.82,"slice":"core","lang":"en"}
|
|
{"prompt":"Backlog item:\n\n```text\nUI-771: fleet health overview\nHeader metrics: online, degraded, offline, updating\nMain chart: device count by state over selected range\nTable: fleet, firmware, last contact, health, rollout cohort\nFilters: region, hardware revision, firmware channel, customer\nScale: 250k rows total; server pagination already exists\nLive updates: websocket patches can arrive out of order\nResponsive: table becomes cards below 720px\nAccessibility: chart needs text summary; status cannot rely on color\nError states: stale data banner, socket reconnecting, initial load failure\nUse design tokens from `packages/fleet-theme`\n```\n\nImplement the dashboard views and interaction states. Do not change the fleet API.","purpose":"frontendImpl","secondary":null,"mixed":false,"difficulty":0.77,"slice":"pasted-context","lang":"en"}
|
|
{"prompt":"Implement `/v1/models/{id}/card.md` by rendering stored model metadata and evaluation results into deterministic Markdown with caching.","purpose":"backendImpl","secondary":null,"mixed":false,"difficulty":0.46,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Increase the watchdog threshold to 12 seconds","purpose":"quickFix","secondary":null,"mixed":false,"difficulty":0.09,"slice":"core","lang":"en"}
|
|
{"prompt":"Fix the clipped squad names and extract the shared truncation rule used by the lobby and scoreboard. Preserve all other layout behavior.","purpose":"quickFix","secondary":"refactor","mixed":true,"difficulty":0.38,"slice":"mixed","lang":"en"}
|
|
{"prompt":"Flatten the rule-evaluation branches into named stages without changing prediction scores or rejection reasons.","purpose":"refactor","secondary":null,"mixed":false,"difficulty":0.61,"slice":"core","lang":"en"}
|
|
{"prompt":"Corrige `enviroment` a `environment` solamente en `deploy/values-dev.yaml`.","purpose":"quickFix","secondary":null,"mixed":false,"difficulty":0.04,"slice":"boundary","lang":"es"}
|
|
{"prompt":"Investigate why the ARM inference service becomes 30% slower after several model hot-swaps but recovers on restart.","purpose":"debugging","secondary":null,"mixed":false,"difficulty":0.8,"slice":"core","lang":"en"}
|
|
{"prompt":"Summarize the existing anti-cheat event pipeline for the privacy review, including collection, enrichment, retention, and human access.","purpose":"writing","secondary":null,"mixed":false,"difficulty":0.42,"slice":"core","lang":"en"}
|
|
{"prompt":"Develop a capacity roadmap for expanding our Kafka telemetry cluster from three to twelve regions. Include partition ownership, replication traffic, failure domains, quotas, cost gates, and load-test milestones.","purpose":"planning","secondary":null,"mixed":false,"difficulty":0.92,"slice":"core","lang":"en"}
|
|
{"prompt":"Implement WebAssembly plugin execution in the edge agent with memory, fuel, syscall, and output limits.","purpose":"backendImpl","secondary":null,"mixed":false,"difficulty":0.88,"slice":"core","lang":"en"}
|
|
{"prompt":"Create the model-comparison screen with side-by-side metrics, confidence intervals, slice filters, and accessible chart summaries.","purpose":"frontendImpl","secondary":null,"mixed":false,"difficulty":0.69,"slice":"core","lang":"en"}
|
|
{"prompt":"Remove the duplicate colon in the alert title","purpose":"quickFix","secondary":null,"mixed":false,"difficulty":0.03,"slice":"core","lang":"en"}
|
|
{"prompt":"Merge the two ECS task-definition builders while preserving JSON output, environment ordering, and deployment behavior.","purpose":"refactor","secondary":null,"mixed":false,"difficulty":0.6,"slice":"core","lang":"en"}
|
|
{"prompt":"Trainer output from the failed run:\n\n```text\nstep=18420 loss=2.481 grad_norm=1.91 lr=0.000031\nstep=18421 loss=2.476 grad_norm=1.88 lr=0.000031\nrank=3 NCCL WARN Watchdog caught collective operation timeout\nWorkNCCL(SeqNum=77219, OpType=ALLREDUCE, Timeout(ms)=600000)\nrank=0 dataloader batch=18422 ready=true\nrank=1 dataloader batch=18422 ready=true\nrank=2 dataloader batch=18422 ready=true\nrank=3 dataloader batch=18422 ready=false\nrank=3 worker stderr: PIL.UnidentifiedImageError asset/93/88201.avif\nrank=3 exception handler skipped batch and continued\n```\n\nTraining hangs rather than failing after one corrupt image. Determine how rank 3 diverges from the collective schedule and why the exception handler does not keep distributed steps aligned.","purpose":"debugging","secondary":null,"mixed":false,"difficulty":0.83,"slice":"pasted-context","lang":"en"}
|
|
{"prompt":"Change the dashboard label from `GPU's` to `GPUs`.","purpose":"quickFix","secondary":null,"mixed":false,"difficulty":0.03,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Assess how the current enclave attestation verifier handles certificate expiry, nonce reuse, and an unknown measurement. No code changes.","purpose":"review","secondary":null,"mixed":false,"difficulty":0.69,"slice":"core","lang":"en"}
|
|
{"prompt":"Reorganize the Terraform provider into resource-specific packages while preserving schema, state upgrades, diagnostics, and generated documentation; also write a maintainer note describing the new package boundaries.","purpose":"refactor","secondary":"writing","mixed":true,"difficulty":0.79,"slice":"mixed","lang":"en"}
|
|
{"prompt":"Plan the replacement of our proprietary metrics format with OpenTelemetry Protocol, including semantic mappings, dual emission, cardinality validation, SDK migration, and retirement gates.","purpose":"planning","secondary":null,"mixed":false,"difficulty":0.86,"slice":"core","lang":"en"}
|
|
{"prompt":"Add a retention worker for expired experiment assignments with legal-hold exclusions and idempotent batches.","purpose":"backendImpl","secondary":null,"mixed":false,"difficulty":0.57,"slice":"core","lang":"en"}
|
|
{"prompt":"Build the game-server region picker with latency probes, recommended-region highlighting, party constraints, and keyboard navigation.","purpose":"frontendImpl","secondary":null,"mixed":false,"difficulty":0.59,"slice":"core","lang":"en"}
|
|
{"prompt":"do the infra bit","purpose":"backendImpl","secondary":null,"mixed":false,"difficulty":0.49,"slice":"vague-eval","lang":"en"}
|
|
{"prompt":"Replace raw device IDs with the `DeviceKey` value type across the firmware service while preserving JSON, SQL, and protobuf representations.","purpose":"refactor","secondary":null,"mixed":false,"difficulty":0.65,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Find the cause of corrupted save games after cloud conflict resolution, implement the smallest verified fix, and add a troubleshooting note explaining recovery options for support.","purpose":"debugging","secondary":"writing","mixed":true,"difficulty":0.84,"slice":"mixed","lang":"en"}
|
|
{"prompt":"Why does the batch scheduler reserve two CPU cores for every GPU task? Review the existing placement rationale and utilization tradeoffs; nothing is malfunctioning.","purpose":"review","secondary":null,"mixed":false,"difficulty":0.51,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Track down the intermittent 90-second pause in the DNS controller reconcile loop.","purpose":"debugging","secondary":null,"mixed":false,"difficulty":0.72,"slice":"core","lang":"en"}
|
|
{"prompt":"Write release notes for the new quantized model runtime, calling out supported hardware, accuracy caveats, fallback behavior, and upgrade steps.","purpose":"writing","secondary":null,"mixed":false,"difficulty":0.41,"slice":"core","lang":"en"}
|
|
{"prompt":"Architecture scratchpad:\n\n```text\nGoal: isolate user-submitted Python training code\nCurrent: containers run in a shared Kubernetes node pool\nThreats: kernel escape, credential theft, lateral network movement, data exfiltration, resource abuse\nCandidate isolation: gVisor, Kata, Firecracker workers\nNeeds: NVIDIA GPUs for 30% of jobs; local NVMe cache; startup under 45s\nTenants select approved base images but can pip install at runtime\nTraining data arrives through signed object-store URLs\nNo inbound connectivity required\nLogs and checkpoints must survive worker termination\nCompliance wants per-job network and filesystem evidence\n```\n\nProduce an implementation-free isolation architecture and evaluation plan. Include threat assumptions, prototype benchmarks, GPU constraints, network policy, credential delivery, observability, rollout cohorts, and kill criteria.","purpose":"planning","secondary":null,"mixed":false,"difficulty":0.96,"slice":"pasted-context","lang":"en"}
|
|
{"prompt":"Give the firmware updater a proper progress timeline","purpose":"frontendImpl","secondary":null,"mixed":false,"difficulty":0.45,"slice":"core","lang":"en"}
|
|
{"prompt":"Turn the existing data-retention implementation into an administrator guide, and separately review whether the documented guarantees match deletion workers, backups, and analytics exports.","purpose":"writing","secondary":"review","mixed":true,"difficulty":0.71,"slice":"mixed","lang":"en"}
|
|
{"prompt":"Outline the architecture and rollout for multi-cluster service discovery across Kubernetes regions, without changing manifests or controllers yet.","purpose":"planning","secondary":null,"mixed":false,"difficulty":0.86,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Implement a lock-free ring buffer for the audio capture path with overflow counters and deterministic tests.","purpose":"backendImpl","secondary":null,"mixed":false,"difficulty":0.8,"slice":"core","lang":"en"}
|
|
{"prompt":"Produce a field guide for interpreting the device agent's LED error patterns and serial diagnostic codes.","purpose":"writing","secondary":null,"mixed":false,"difficulty":0.32,"slice":"core","lang":"en"}
|
|
{"prompt":"Move optimizer construction out of `Trainer` into a factory without changing parameter groups, defaults, checkpoint keys, or training results.","purpose":"refactor","secondary":null,"mixed":false,"difficulty":0.63,"slice":"core","lang":"en"}
|
|
{"prompt":"Investigate why the sidecar sometimes reports healthy while its Envoy admin endpoint is no longer responding.","purpose":"debugging","secondary":null,"mixed":false,"difficulty":0.68,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Add atomic lease renewal to the batch coordinator and reject renewals from stale worker generations.","purpose":"backendImpl","secondary":null,"mixed":false,"difficulty":0.65,"slice":"core","lang":"en"}
|
|
{"prompt":"Set `replicaCount` to 5 in production","purpose":"quickFix","secondary":null,"mixed":false,"difficulty":0.08,"slice":"core","lang":"en"}
|
|
{"prompt":"Create the embedded-device provisioning wizard with QR scanning, Wi-Fi setup, certificate enrollment, retry states, and offline guidance.","purpose":"frontendImpl","secondary":null,"mixed":false,"difficulty":0.71,"slice":"core","lang":"en"}
|
|
{"prompt":"Renombra `tmpModel` a `candidateModel` solo en `Evaluator.kt`; no cambies ninguna lógica.","purpose":"quickFix","secondary":null,"mixed":false,"difficulty":0.07,"slice":"boundary","lang":"es"}
|
|
{"prompt":"Review the checkpoint loader for unsafe pickle execution and explain which artifact sources can reach it. Report findings only.","purpose":"review","secondary":null,"mixed":false,"difficulty":0.74,"slice":"core","lang":"en"}
|
|
{"prompt":"Add the model-lineage endpoint and write its API documentation. Include parent artifacts, datasets, code revisions, pagination, authorization, retention gaps, JSON examples, and error cases.","purpose":"backendImpl","secondary":"writing","mixed":true,"difficulty":0.72,"slice":"mixed","lang":"en"}
|
|
{"prompt":"Design a staged migration from node-local container caches to a regional registry mirror, including eviction, integrity, outage behavior, bandwidth estimates, and rollback.","purpose":"planning","secondary":null,"mixed":false,"difficulty":0.82,"slice":"core","lang":"en"}
|
|
{"prompt":"Implement HLS segment encryption with rotating keys and deterministic IV derivation in the streaming packager.","purpose":"backendImpl","secondary":null,"mixed":false,"difficulty":0.77,"slice":"core","lang":"en"}
|
|
{"prompt":"Build the match-results ceremony with rank transitions, reward reveals, skip behavior, and reduced-motion alternatives.","purpose":"frontendImpl","secondary":null,"mixed":false,"difficulty":0.64,"slice":"core","lang":"en"}
|
|
{"prompt":"continue the weird crash work","purpose":"debugging","secondary":null,"mixed":false,"difficulty":0.55,"slice":"vague-eval","lang":"en"}
|
|
{"prompt":"Extract environment detection from the deployment CLI and keep every flag, precedence rule, prompt, and exit code unchanged.","purpose":"refactor","secondary":null,"mixed":false,"difficulty":0.56,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Diagnose why the feature generator produces duplicate rows after a Spark executor retry, implement the proven correction, and document the idempotency invariant beside the job configuration.","purpose":"debugging","secondary":"writing","mixed":true,"difficulty":0.86,"slice":"mixed","lang":"en"}
|
|
{"prompt":"Assess the proposal to store experiment assignments in browser local storage. Cover tampering, privacy, cross-device behavior, expiration, and compatibility; return review comments only.","purpose":"review","secondary":null,"mixed":false,"difficulty":0.59,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Determine why packet loss jumps when the embedded gateway rotates its TLS certificate.","purpose":"debugging","secondary":null,"mixed":false,"difficulty":0.72,"slice":"core","lang":"en"}
|
|
{"prompt":"Draft the operator handbook section for pausing, resuming, and cancelling distributed training jobs safely.","purpose":"writing","secondary":null,"mixed":false,"difficulty":0.39,"slice":"core","lang":"en"}
|
|
{"prompt":"Migration notes:\n\n```text\nCurrent state:\n- 420 Prometheus servers scrape about 2.8M targets\n- Thanos sidecars upload 2-hour blocks\n- global queries fan out through three query tiers\nPain:\n- rule evaluation differs by region\n- compactors compete for object-store bandwidth\n- tenant attribution is incomplete\nProposal: centralize ingestion through remote write and shard by tenant\nConstraints:\n- agents cannot all upgrade in one quarter\n- recording rule names are customer-visible\n- historical queries must span old and new stores\n- maximum tolerated metrics loss is 30 seconds\n- finance needs projected storage and egress by phase\n```\n\nTurn this into a multi-quarter migration roadmap with compatibility, capacity testing, dual-query verification, rule migration, cost gates, rollback, and ownership. Do not implement changes.","purpose":"planning","secondary":null,"mixed":false,"difficulty":0.94,"slice":"pasted-context","lang":"en"}
|
|
{"prompt":"Add keyboard navigation to the shader graph editor","purpose":"frontendImpl","secondary":null,"mixed":false,"difficulty":0.5,"slice":"core","lang":"en"}
|
|
{"prompt":"Restructure the rollout controller and document its extension API. Preserve rollout ordering, retries, status conditions, and metrics while creating stable hooks for custom health analysis.","purpose":"refactor","secondary":"writing","mixed":true,"difficulty":0.78,"slice":"mixed","lang":"en"}
|
|
{"prompt":"Architect a privacy-preserving federated learning workflow for mobile keyboards, covering secure aggregation, consent, poisoning defenses, update eligibility, model rollback, and rollout milestones.","purpose":"planning","secondary":null,"mixed":false,"difficulty":0.97,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Implement incremental Parquet footer scanning over object-store range requests with bounded concurrency.","purpose":"backendImpl","secondary":null,"mixed":false,"difficulty":0.73,"slice":"core","lang":"en"}
|
|
{"prompt":"Write an onboarding guide for adding a new board to the Zephyr firmware repository, including devicetree, Kconfig, flashing, and hardware tests.","purpose":"writing","secondary":null,"mixed":false,"difficulty":0.46,"slice":"core","lang":"en"}
|
|
{"prompt":"Consolidate the three game-event serializers behind a versioned codec while keeping replay bytes identical.","purpose":"refactor","secondary":null,"mixed":false,"difficulty":0.68,"slice":"core","lang":"en"}
|
|
{"prompt":"Find why read replicas occasionally return an older feature vector after the writer confirms commit.","purpose":"debugging","secondary":null,"mixed":false,"difficulty":0.76,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Add automatic certificate renewal to the MQTT broker with overlapping trust and connection draining.","purpose":"backendImpl","secondary":null,"mixed":false,"difficulty":0.73,"slice":"core","lang":"en"}
|
|
{"prompt":"Change the typo `frametime_ms` to `frame_time_ms` in the new dashboard query.","purpose":"quickFix","secondary":null,"mixed":false,"difficulty":0.1,"slice":"core","lang":"en"}
|
|
{"prompt":"Build the training-run detail page with live logs, metric charts, resource usage, artifacts, cancellation, and reconnect recovery.","purpose":"frontendImpl","secondary":null,"mixed":false,"difficulty":0.73,"slice":"core","lang":"en"}
|
|
{"prompt":"設定ファイルの `batch_sze` を `batch_size` に直して。まだ未リリースなので互換処理は不要です。","purpose":"quickFix","secondary":null,"mixed":false,"difficulty":0.08,"slice":"boundary","lang":"ja"}
|
|
{"prompt":"Evaluate whether the eBPF network observer can miss short-lived connections during map rotation. Explain the current synchronization approach; no patch.","purpose":"review","secondary":null,"mixed":false,"difficulty":0.71,"slice":"core","lang":"en"}
|
|
{"prompt":"Correct the cramped mobile leaderboard header and consolidate its duplicated breakpoint styles. Only the documented spacing and wrapping defects should change.","purpose":"quickFix","secondary":"refactor","mixed":true,"difficulty":0.43,"slice":"mixed","lang":"en"}
|
|
{"prompt":"Develop a deprecation plan for the legacy inference protocol, covering client inventory, adapters, shadow traffic, parity metrics, customer notices, and removal gates.","purpose":"planning","secondary":null,"mixed":false,"difficulty":0.81,"slice":"core","lang":"en"}
|
|
{"prompt":"Implement a streaming tar extractor that rejects path traversal, sparse-file abuse, and size-limit violations.","purpose":"backendImpl","secondary":null,"mixed":false,"difficulty":0.69,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Style the vehicle damage overlay with directional indicators, severity animation, color-blind-safe modes, and ultrawide support.","purpose":"frontendImpl","secondary":null,"mixed":false,"difficulty":0.58,"slice":"core","lang":"en"}
|
|
{"prompt":"Disable `ENABLE_OLD_MATCHMAKER`","purpose":"quickFix","secondary":null,"mixed":false,"difficulty":0.05,"slice":"core","lang":"en"}
|
|
{"prompt":"Split the data-loader package into sampling, decoding, augmentation, and collation modules while preserving batches for a fixed seed.","purpose":"refactor","secondary":null,"mixed":false,"difficulty":0.76,"slice":"core","lang":"en"}
|
|
{"prompt":"Failure from the hardware-in-loop suite:\n\n```text\ncase=power_cycle_during_slot_swap iteration=418\nactive_slot=A pending_slot=B\nbootloader wrote B metadata crc=0x72a1\nrelay power-off at +14ms\nrelay power-on after 3000ms\nboot reason=brownout\nslot A valid=true generation=91\nslot B valid=true generation=92 confirmed=false\nselected slot=B\napplication heartbeat missing\nreset after 30s\nselected slot=B\napplication heartbeat missing\nexpected fallback to A after 1 failed boot\n```\n\nThe bootloader keeps selecting the unconfirmed image instead of falling back. Determine whether the attempt counter, metadata commit order, or brownout reset classification prevents the failure from being recorded.","purpose":"debugging","secondary":null,"mixed":false,"difficulty":0.85,"slice":"pasted-context","lang":"en"}
|
|
{"prompt":"Explain why the event schema includes both `device_time` and `received_time`, and how downstream jobs are expected to use each.","purpose":"review","secondary":null,"mixed":false,"difficulty":0.4,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Create a reference page for the existing checkpoint manifest format with field definitions, compatibility rules, integrity checks, and examples.","purpose":"writing","secondary":null,"mixed":false,"difficulty":0.43,"slice":"core","lang":"en"}
|
|
{"prompt":"Plan the device-twin service and implement its first storage layer. The design should settle desired versus reported state, version conflicts, offline retention, fan-out, authorization, and migration from retained MQTT messages; then add the schema, repository, reconciliation primitives, and integration tests.","purpose":"planning","secondary":"backendImpl","mixed":true,"difficulty":0.93,"slice":"mixed","lang":"en"}
|
|
{"prompt":"Outline a roadmap for making game replays portable across engine upgrades, including versioned state, deterministic validation, compatibility windows, and archival tests.","purpose":"planning","secondary":null,"mixed":false,"difficulty":0.86,"slice":"core","lang":"en"}
|
|
{"prompt":"Implement cache-aware routing for embedding requests using model version, tenant, and GPU memory pressure.","purpose":"backendImpl","secondary":null,"mixed":false,"difficulty":0.76,"slice":"core","lang":"en"}
|
|
{"prompt":"Create the incident-map view with clustered sites, live severity updates, timeline playback, and non-map accessibility fallback.","purpose":"frontendImpl","secondary":null,"mixed":false,"difficulty":0.72,"slice":"core","lang":"en"}
|
|
{"prompt":"clean up that module","purpose":"refactor","secondary":null,"mixed":false,"difficulty":0.45,"slice":"vague-eval","lang":"en"}
|
|
{"prompt":"Move the packet checksum functions into `wire/checksum` across firmware and simulator targets, preserving generated bytes and timing behavior.","purpose":"refactor","secondary":null,"mixed":false,"difficulty":0.62,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Investigate the intermittent model-registry 409 responses during concurrent uploads, fix the verified transaction race, and add an operator note distinguishing safe retries from conflicting artifacts.","purpose":"debugging","secondary":"writing","mixed":true,"difficulty":0.81,"slice":"mixed","lang":"en"}
|
|
{"prompt":"Why does the deployment agent retain the previous two manifests on disk? Explain the existing recovery behavior and storage tradeoff; no failure is reported.","purpose":"review","secondary":null,"mixed":false,"difficulty":0.42,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Trace the source of nondeterministic particle positions in replay verification after the compiler upgrade.","purpose":"debugging","secondary":null,"mixed":false,"difficulty":0.77,"slice":"core","lang":"en"}
|
|
{"prompt":"Write a migration guide from handwritten Kubernetes manifests to the new Helm chart, with value mappings and rollback instructions.","purpose":"writing","secondary":null,"mixed":false,"difficulty":0.41,"slice":"core","lang":"en"}
|
|
{"prompt":"Planning fragment from an internal RFC:\n\n```text\nWe need online feature joins under 12 ms p99.\nSources: account DB CDC, session Kafka stream, daily warehouse aggregates\nEntity keys: account, device, household\nFreshness: sessions under 30s; account attributes under 5m; aggregates under 30h\nMissing values must be distinguishable from source lag\nModel requests include an explicit feature-set version\nPotential stores: Redis Cluster, ScyllaDB, custom RocksDB service\nPeak: 240k reads/sec, 18k writes/sec\nMulti-region active-active required for reads\nDeletion requests must propagate within 24h\n```\n\nProduce an architecture evaluation and rollout plan. Define benchmarks, consistency semantics, schema ownership, deletion verification, failure behavior, cost comparison, and migration gates without implementing a store.","purpose":"planning","secondary":null,"mixed":false,"difficulty":0.95,"slice":"pasted-context","lang":"en"}
|
|
{"prompt":"Give the patch notes screen readable typography","purpose":"frontendImpl","secondary":null,"mixed":false,"difficulty":0.31,"slice":"core","lang":"en"}
|
|
{"prompt":"Write a design for the internal package registry and then implement immutable publishing plus provenance verification. Cover namespaces, authorization, retention, replication, compromise recovery, compatibility, and staged client adoption before building the service paths.","purpose":"planning","secondary":"backendImpl","mixed":true,"difficulty":0.95,"slice":"mixed","lang":"en"}
|
|
{"prompt":"Plan the extraction of matchmaking from the game monolith, including state ownership, party consistency, regional routing, shadow decisions, and rollout milestones.","purpose":"planning","secondary":null,"mixed":false,"difficulty":0.91,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Add durable consumer offsets to the CAN bus gateway so it resumes uploads after power loss.","purpose":"backendImpl","secondary":null,"mixed":false,"difficulty":0.65,"slice":"core","lang":"en"}
|
|
{"prompt":"Turn the existing admission-controller behavior into user documentation covering policies, exemptions, dry runs, and rejection messages.","purpose":"writing","secondary":null,"mixed":false,"difficulty":0.37,"slice":"core","lang":"en"}
|
|
{"prompt":"Combine the three image-normalization implementations while preserving pixel output for every fixture.","purpose":"refactor","secondary":null,"mixed":false,"difficulty":0.63,"slice":"core","lang":"en"}
|
|
{"prompt":"Investigate why the Linux game server reports a stable tick rate while clients observe periodic 200 ms freezes.","purpose":"debugging","secondary":null,"mixed":false,"difficulty":0.75,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Implement tenant-aware compaction scheduling in the time-series database with fairness and I/O budgets.","purpose":"backendImpl","secondary":null,"mixed":false,"difficulty":0.79,"slice":"core","lang":"en"}
|
|
{"prompt":"Pin the Node exporter chart to `4.49.0`","purpose":"quickFix","secondary":null,"mixed":false,"difficulty":0.06,"slice":"core","lang":"en"}
|
|
{"prompt":"Build the drone mission planner canvas with waypoints, altitude editing, no-fly-zone warnings, undo, and touch support.","purpose":"frontendImpl","secondary":null,"mixed":false,"difficulty":0.75,"slice":"core","lang":"en"}
|
|
{"prompt":"Ändere nur `scrape_interval: 30s` auf `15s` in `clusters/staging/prometheus.yaml`.","purpose":"quickFix","secondary":null,"mixed":false,"difficulty":0.07,"slice":"boundary","lang":"de"}
|
|
{"prompt":"Inspect the shader cache loader for malformed-offset reads and explain whether downloaded cache files are trusted before validation.","purpose":"review","secondary":null,"mixed":false,"difficulty":0.72,"slice":"core","lang":"en"}
|
|
{"prompt":"Implement the cluster-cost export endpoint and write its consumer documentation. Include attribution windows, delayed billing data, pagination, currency, permissions, examples, and known reconciliation gaps.","purpose":"backendImpl","secondary":"writing","mixed":true,"difficulty":0.71,"slice":"mixed","lang":"en"}
|
|
{"prompt":"Create a disaster-recovery plan for the model registry, including artifact integrity, metadata restoration, cross-region lag, signing keys, restore drills, and RTO tiers.","purpose":"planning","secondary":null,"mixed":false,"difficulty":0.89,"slice":"core","lang":"en"}
|
|
{"prompt":"Add cooperative cancellation to the C++ asset compiler without leaving partial cache entries.","purpose":"backendImpl","secondary":null,"mixed":false,"difficulty":0.68,"slice":"core","lang":"en"}
|
|
{"prompt":"Implement the warehouse robot status board with floor zones, battery warnings, blocked-path states, and wall-display scaling.","purpose":"frontendImpl","secondary":null,"mixed":false,"difficulty":0.63,"slice":"core","lang":"en"}
|
|
{"prompt":"do the model screen","purpose":"frontendImpl","secondary":null,"mixed":false,"difficulty":0.43,"slice":"vague-eval","lang":"en"}
|
|
{"prompt":"Extract snapshot retention from `CheckpointManager` into a policy object while keeping deletion order and recovery behavior identical.","purpose":"refactor","secondary":null,"mixed":false,"difficulty":0.59,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Diagnose the intermittent NCCL initialization failure on mixed A100 and H100 nodes, implement the proven scheduling or runtime correction, and write a troubleshooting section listing the decisive logs.","purpose":"debugging","secondary":"writing","mixed":true,"difficulty":0.88,"slice":"mixed","lang":"en"}
|
|
{"prompt":"Assess whether the proposed device enrollment flow resists cloned QR codes and replayed bootstrap tokens. Provide review findings only.","purpose":"review","secondary":null,"mixed":false,"difficulty":0.69,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Find why the log shipper occasionally rereads a rotated file from byte zero and duplicates several gigabytes.","purpose":"debugging","secondary":null,"mixed":false,"difficulty":0.74,"slice":"core","lang":"en"}
|
|
{"prompt":"Write the release announcement for firmware 3.14, including rollout schedule, compatibility, visible changes, and recovery steps.","purpose":"writing","secondary":null,"mixed":false,"difficulty":0.34,"slice":"core","lang":"en"}
|
|
{"prompt":"Incident review material:\n\n```text\nService: feature-cache\nImpact: recommendation requests fell back to defaults for 11 minutes\nTimeline:\n14:02 deploy began in eu-central\n14:04 new pods failed readiness `schema generation 84 unavailable`\n14:05 autoscaler doubled old pods due latency\n14:07 registry request rate limited at 429\n14:09 rollout paused automatically\n14:11 on-call increased registry quota\n14:13 new pods became ready\nContributing notes:\n- image startup downloads the schema synchronously\n- readiness timeout is 90s\n- schema artifacts are immutable\n- node-local cache is disabled on spot pools\n- no alert exists for registry throttling by consumer\n```\n\nProduce a blameless incident report with impact, contributing factors, detection gaps, corrective actions, owners as placeholders, and measurable follow-up criteria. Do not invent a single root cause beyond the evidence.","purpose":"writing","secondary":null,"mixed":false,"difficulty":0.56,"slice":"pasted-context","lang":"en"}
|
|
{"prompt":"Add a compact mode to the server browser","purpose":"frontendImpl","secondary":null,"mixed":false,"difficulty":0.39,"slice":"core","lang":"en"}
|
|
{"prompt":"Design and implement certificate-based device enrollment. First document manufacturing trust, bootstrap identities, rotation, revocation, offline enrollment, compromised batches, auditability, and rollout. Then build the enrollment endpoints, signing integration, persistence, and conformance tests.","purpose":"planning","secondary":"backendImpl","mixed":true,"difficulty":0.96,"slice":"mixed","lang":"en"}
|
|
{"prompt":"Propose a phased architecture for replacing nightly warehouse retraining with continuous training, including data gates, evaluation, promotion, rollback, and cost limits.","purpose":"planning","secondary":null,"mixed":false,"difficulty":0.91,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Implement a priority-aware admission queue for GPU inference with tenant quotas and deadline expiry.","purpose":"backendImpl","secondary":null,"mixed":false,"difficulty":0.79,"slice":"core","lang":"en"}
|
|
{"prompt":"Rédige un guide pour créer et valider un nouveau scénario de charge dans `benchmarks/scenarios`, avec exemples de configuration.","purpose":"writing","secondary":null,"mixed":false,"difficulty":0.36,"slice":"core","lang":"fr"}
|
|
{"prompt":"Restructure the embedded logging package into formatting, buffering, and transport layers with byte-identical serial output.","purpose":"refactor","secondary":null,"mixed":false,"difficulty":0.69,"slice":"core","lang":"en"}
|
|
{"prompt":"Investigate why the multiplayer relay accepts a session token for several seconds after explicit revocation.","purpose":"debugging","secondary":null,"mixed":false,"difficulty":0.71,"slice":"boundary","lang":"en"}
|
|
{"prompt":"Add garbage collection for abandoned multipart model uploads with retention metrics and legal-hold exemptions.","purpose":"backendImpl","secondary":null,"mixed":false,"difficulty":0.61,"slice":"core","lang":"en"}
|
|
{"prompt":"Remove the stale `beta` suffix from the CLI help","purpose":"quickFix","secondary":null,"mixed":false,"difficulty":0.05,"slice":"core","lang":"en"}
|
|
{"prompt":"Create the data-labeling workspace with keyboard shortcuts, annotation history, confidence controls, and conflict indicators.","purpose":"frontendImpl","secondary":null,"mixed":false,"difficulty":0.7,"slice":"core","lang":"en"}
|
|
{"prompt":"`traning_jobs_total` を `training_jobs_total` に修正して。新しい未公開メトリクスなので互換エイリアスは不要。","purpose":"quickFix","secondary":null,"mixed":false,"difficulty":0.08,"slice":"boundary","lang":"ja"}
|
|
{"prompt":"Review the OTA manifest parser for integer overflows, duplicate component names, and signature-verification ordering. Do not implement fixes.","purpose":"review","secondary":null,"mixed":false,"difficulty":0.75,"slice":"core","lang":"en"}
|
|
{"prompt":"Build the experiment audit endpoint and produce its API guide. Support actor, model, date, and action filters; stable pagination; retention boundaries; authorization tests; JSON examples; and an explanation of delayed events.","purpose":"backendImpl","secondary":"writing","mixed":true,"difficulty":0.71,"slice":"mixed","lang":"en"}
|