System Design on the Road

Episodes

• 
  S2 E26 — Design an AI Evaluation Platform: Reproducibility, Grading, and Release Gates 22.10.2026 34m

  An evaluation score improves even though the model has not changed. Daniel and Maya design a platform that can trace the difference to data, graders, configuration, infrastructure, or actual model behavior.Frozen inputs, task attempts, grading versions, failure classification, and aggregation form a reproducible run. Infrastructure errors and retries must not distort model comparisons. The interview ends with uncertainty and release gates, asking what evidence a higher score really provides for a deployment decision.Chapters00:00:00 Welcome and an unexplained score gain00:02:26 Frozen datasets, models, and run inputs00:05:29 Task execution, grading, and aggregation00:08:46 Grader changes and comparison validity00:12:32 Worker failures, retries, and fairness00:16:33 Uncertainty and reproducible comparisons00:20:15 Release gates and evaluation operations00:31:14 Final score interpretation and recapDaniel is the interviewer, voiced by Liam. Maya is the engineer, voiced by Bella. AI narration at a calm pace, with three thinking pauses.Opening music: Stealth Groover by Kevin MacLeod (incompetech.com). Licensed under Creative Commons Attribution 4.0: https://creativecommons.org/licenses/by/4.0/ . Edited excerpt with EQ, volume automation, a smooth fade, and narration.
• 
  S2 E25 — Design an AI Research Assistant: Sources, Budgets, and Checkpoints 22.10.2026 33m

  Three sources tell two different stories and the research deadline is approaching. Daniel and Maya design an assistant that can choose what to investigate, preserve its evidence, and stop within time and spending limits.A working lookup expands into parallel research, conflicting sources, checkpoints, and recovery. Tool results and pages remain evidence rather than authorities. Evaluation asks whether the answer is supported and useful, including when uncertainty or insufficient evidence is the conclusion the assistant should report.Chapters00:00:00 Welcome and conflicting research sources00:02:30 Answer requirements and evidence records00:05:47 A complete research lookup00:09:29 Conflicting sources and uncertainty00:13:24 Parallel research and bounded budgets00:17:15 Checkpoints, tools, and untrusted content00:21:29 Evaluating useful, supported research00:30:22 Final answer and evidence trailDaniel is the interviewer, voiced by Liam. Maya is the engineer, voiced by Bella. AI narration at a calm pace, with three thinking pauses.Opening music: Stealth Groover by Kevin MacLeod (incompetech.com). Licensed under Creative Commons Attribution 4.0: https://creativecommons.org/licenses/by/4.0/ . Edited excerpt with EQ, volume automation, a smooth fade, and narration.
• 
  S2 E24 — Design Enterprise RAG: Retrieval, Permissions, and Citations 22.10.2026 34m

  An employee may read a document when it is retrieved, but loses access before the generated answer is released. Daniel and Maya design enterprise RAG around that permission race rather than relevance alone.Document versions, ingestion, retrieval, evidence selection, and answer generation form the baseline. Deletion and indexing delays test the copies between source and response. Citations and evaluation ask whether the final answer has adequate evidence and whether the user is still authorized to receive it.Chapters00:00:00 Welcome and a wrongly disclosed answer00:02:24 Source records, versions, and permissions00:05:41 Ingestion, retrieval, and generation00:09:12 Revocation races and release-time checks00:13:14 Relevance, evidence, and citations00:17:08 Index failures, deletion, and freshness00:20:58 Evaluating retrieval and answer evidence00:30:50 Final authorized answer flow and recapDaniel is the interviewer, voiced by Liam. Maya is the engineer, voiced by Bella. AI narration at a calm pace, with three thinking pauses.Opening music: Stealth Groover by Kevin MacLeod (incompetech.com). Licensed under Creative Commons Attribution 4.0: https://creativecommons.org/licenses/by/4.0/ . Edited excerpt with EQ, volume automation, a smooth fade, and narration.
• 
  S2 E23 — Design a Cloud Coding Environment: Sandboxing, Secrets, and Workspace Recovery 22.10.2026 34m

  A repository's test script runs hostile code inside a cloud development workspace. Daniel and Maya design isolation while preserving useful files and developer progress.Sandboxing, network access, secrets, and permission to cause external effects require distinct boundaries. A machine vanishing before a patch is saved tests durable workspace state. Prepared images, capacity, cleanup, and recovery connect the security model to the everyday experience of opening a project, running code, and returning to saved work.Chapters00:00:00 Welcome and an untrusted test script00:02:32 Workspace state and execution boundaries00:05:48 A complete cloud development session00:09:07 Sandboxing hostile repository code00:13:11 Secrets, network access, and side effects00:17:03 Workspace persistence and machine loss00:20:56 Images, capacity, and cleanup00:30:42 Final recovery and isolation guaranteesDaniel is the interviewer, voiced by Liam. Maya is the engineer, voiced by Bella. AI narration at a calm pace, with three thinking pauses.Opening music: Stealth Groover by Kevin MacLeod (incompetech.com). Licensed under Creative Commons Attribution 4.0: https://creativecommons.org/licenses/by/4.0/ . Edited excerpt with EQ, volume automation, a smooth fade, and narration.
• 
  S2 E22 — Design a Browser Agent: Sessions, Permissions, and Recovery 22.10.2026 33m

  A browser agent clicks Submit and disconnects before seeing the result. Daniel and Maya design recovery that preserves the uncertainty instead of automatically repeating a potentially consequential action.Goals, sessions, attempts, observations, and permissions have separate records and authorities. Untrusted page instructions test that separation. Checkpoints, human takeover, cleanup, and monitoring complete the lifecycle, with a clear distinction between an attempted click and a verified result.Chapters00:00:00 Welcome and an uncertain submit action00:02:32 Goals, sessions, and permission boundaries00:05:40 An action-observation session baseline00:08:56 Uncertain effects and safe retries00:12:57 Untrusted pages and prompt injection00:17:07 Checkpoints, human takeover, and cleanup00:21:01 Monitoring attempts and verified outcomes00:30:17 Final authority and recovery boundariesDaniel is the interviewer, voiced by Liam. Maya is the engineer, voiced by Bella. AI narration at a calm pace, with three thinking pauses.Opening music: Stealth Groover by Kevin MacLeod (incompetech.com). Licensed under Creative Commons Attribution 4.0: https://creativecommons.org/licenses/by/4.0/ . Edited excerpt with EQ, volume automation, a smooth fade, and narration.
• 
  S2 E21 — Design a Collaborative Prompt Playground: Versioning and Reproducible Runs 22.10.2026 33m

  A teammate changes a prompt while an experiment is running, and later nobody knows which instructions produced the answer. Daniel and Maya design a collaborative playground that allows editable drafts without losing immutable run inputs.Concurrent edits, run snapshots, model settings, and result history establish traceability. Reconnection and revoked access test the shared room. Comparing outputs becomes a recorded experiment with explicit conditions, rather than a comparison between whatever happens to be in the editor now.Chapters00:00:00 Welcome and an untraceable experiment00:02:22 Drafts, versions, and run identities00:05:44 Editing and running a prompt00:09:02 Concurrent edits and save conflicts00:13:10 Immutable snapshots for reproducible runs00:17:20 Reconnection and revoked workspace access00:21:15 Comparing runs under recorded conditions00:30:01 Final draft-to-result traceabilityDaniel is the interviewer, voiced by Liam. Maya is the engineer, voiced by Bella. AI narration at a calm pace, with three thinking pauses.Opening music: Stealth Groover by Kevin MacLeod (incompetech.com). Licensed under Creative Commons Attribution 4.0: https://creativecommons.org/licenses/by/4.0/ . Edited excerpt with EQ, volume automation, a smooth fade, and narration.
• 
  S2 E20 — Design a Model Distribution System: Downloads, Rollouts, and Rollbacks 22.10.2026 33m

  The model files have downloaded, but the worker is not ready to serve. Daniel and Maya design distribution as a sequence of transfer, verification, staging, loading, and traffic activation.Immutable manifests and content identities define the release. Fleet-wide downloads, interrupted transfers, partially staged files, and mixed versions test the plan. A rollout needs readiness evidence and a usable rollback target; moving all the bytes successfully is only one step toward safely changing the model behind a request.Chapters00:00:00 Welcome and a model that is not ready00:02:41 Immutable releases and artifact identity00:06:08 Download, verify, load, and activate00:09:19 Fleet-wide downloads and bandwidth00:12:42 Resuming downloads and staging safely00:16:37 Mixed versions, rollout, and rollback00:20:41 Security, readiness, and operational cost00:30:06 Final controlled release and recapDaniel is the interviewer, voiced by Liam. Maya is the engineer, voiced by Bella. AI narration at a calm pace, with three thinking pauses.Opening music: Stealth Groover by Kevin MacLeod (incompetech.com). Licensed under Creative Commons Attribution 4.0: https://creativecommons.org/licenses/by/4.0/ . Edited excerpt with EQ, volume automation, a smooth fade, and narration.
• 
  S2 E19 — Design an LLM Inference Service: Batching, Admission Control, and Fairness 22.10.2026 33m

  A short question waits behind a prompt containing an entire book. Daniel and Maya design a shared inference service that uses GPUs efficiently while protecting latency and tenant fairness.Batch formation, request length, resource budgets, admission, and queueing shape the working path. Slow consumers and worker failures complicate an answer already streaming to a client. The final design explains which throughput gains are worth their waiting time and what can still be promised after partial output.Chapters00:00:00 Welcome and a blocked short request00:02:45 Latency contracts and request ownership00:06:17 A working inference request path00:10:01 Batching, long prompts, and latency00:13:54 Admission budgets and tenant fairness00:18:22 Worker failure during streamed output00:22:25 Monitoring and safe service changes00:29:30 Final inference flow and tradeoffsDaniel is the interviewer, voiced by Liam. Maya is the engineer, voiced by Bella. AI narration at a calm pace, with three thinking pauses.Opening music: Stealth Groover by Kevin MacLeod (incompetech.com). Licensed under Creative Commons Attribution 4.0: https://creativecommons.org/licenses/by/4.0/ . Edited excerpt with EQ, volume automation, a smooth fade, and narration.
• 
  S2 E18 — Design a Video Generation Service: GPU Scheduling and Job Recovery 22.10.2026 34m

  A video-generation job is preempted, canceled, and then completed by an obsolete worker. Daniel and Maya design the lifecycle that keeps this late result from becoming the customer's published output.Durable jobs, GPU placement, fair admission, checkpointing, and resumption establish the baseline. Scarce compute makes wasted work and queue policy visible costs. The interview separates producing output bytes from holding current authority to publish them, then follows cancellation and recovery through that boundary.Chapters00:00:00 Welcome and an obsolete video result00:03:07 Inputs, budgets, and durable job state00:07:04 Prompt-to-download baseline00:10:57 GPU placement, admission, and fairness00:15:23 Preemption, checkpoints, and resumption00:19:29 Cancellation and result-publication fencing00:23:38 Monitoring and protecting scarce compute00:30:48 Final job and publication guaranteesDaniel is the interviewer, voiced by Liam. Maya is the engineer, voiced by Bella. AI narration at a calm pace, with three thinking pauses.Opening music: Stealth Groover by Kevin MacLeod (incompetech.com). Licensed under Creative Commons Attribution 4.0: https://creativecommons.org/licenses/by/4.0/ . Edited excerpt with EQ, volume automation, a smooth fade, and narration.
• 
  S2 E17 — Design an AI Chat Service: Streaming, Reconnection, and Cancellation 22.10.2026 33m

  An AI answer is still being generated when its browser connection breaks. Daniel and Maya design a chat service that resumes the existing answer rather than silently starting and charging for another generation.Conversation branches, generation IDs, durable stream events, cursors, and worker ownership connect the interface to model execution. Cancellation has its own outcome; closing a socket is not enough. Admission, latency, and cost limits test which work the system accepts and how it handles a worker that keeps running.Chapters00:00:00 Welcome and an interrupted AI answer00:03:08 Conversation branches and generation IDs00:07:09 A complete streamed generation00:11:05 Durable events, cursors, and reconnection00:15:41 Cancellation and stale-worker ownership00:19:40 Admission, latency, and generation cost00:23:41 Lifecycle monitoring and recovery00:31:01 Final resume and cancellation guaranteesDaniel is the interviewer, voiced by Liam. Maya is the engineer, voiced by Bella. AI narration at a calm pace, with three thinking pauses.Opening music: Stealth Groover by Kevin MacLeod (incompetech.com). Licensed under Creative Commons Attribution 4.0: https://creativecommons.org/licenses/by/4.0/ . Edited excerpt with EQ, volume automation, a smooth fade, and narration.
• 
  S2 E16 — Design a Collaborative Editor: CRDTs, Offline Edits, and Permissions 22.10.2026 33m

  One collaborator deletes a paragraph while another edits it offline. Daniel and Maya compare approaches to shared editing before developing local operations, durable acceptance, reconnect history, and snapshots.Concurrent insertion and deletion test convergence. A returning device whose access was revoked tests a different guarantee: an edit may be mergeable but no longer authorized. Busy documents, ownership, presence, and realistic recovery tests connect the operation model to what people experience while typing and reconnecting.Chapters00:00:00 Welcome and conflicting offline edits00:03:00 Requirements and collaboration approaches00:06:58 Local edits and durable operations00:10:35 Concurrent insertion and deletion00:15:20 Offline reconnection and revoked access00:19:31 Snapshots, ownership, and hot documents00:23:30 Convergence and recovery testing00:30:06 Final durability and permission guaranteesDaniel is the interviewer, voiced by Liam. Maya is the engineer, voiced by Bella. AI narration at a calm pace, with three thinking pauses.Opening music: Stealth Groover by Kevin MacLeod (incompetech.com). Licensed under Creative Commons Attribution 4.0: https://creativecommons.org/licenses/by/4.0/ . Edited excerpt with EQ, volume automation, a smooth fade, and narration.
• 
  S2 E15 — Design a Web Crawler: URL Scheduling, Deduplication, and Politeness 22.10.2026 33m

  A crawler has ten million queued URLs, but many point to one small website. Daniel and Maya design a frontier whose scheduling rules limit pressure on remote origins instead of letting queue size drive request volume.Discovery, deduplication, fetches, parsing, and publication form a complete path. Shared politeness, infinite calendars, changing pages, redirects, and hostile content test its limits. Recovery must preserve both accepted results and the permissions and timing rules governing the next fetch.Chapters00:00:00 Welcome and a crowded URL frontier00:02:56 Crawl scope and URL scheduling00:06:53 Discovery, fetch, parse, and publication00:10:51 Per-origin politeness across workers00:15:18 Crawl traps, revisits, and changing pages00:19:55 Untrusted content and bounded fetches00:23:52 Recovery, monitoring, and result publication00:30:19 Final frontier policy and recapDaniel is the interviewer, voiced by Liam. Maya is the engineer, voiced by Bella. AI narration at a calm pace, with three thinking pauses.Opening music: Stealth Groover by Kevin MacLeod (incompetech.com). Licensed under Creative Commons Attribution 4.0: https://creativecommons.org/licenses/by/4.0/ . Edited excerpt with EQ, volume automation, a smooth fade, and narration.
• 
  S2 E14 — Design Ad Click Analytics: Deduplication, Late Events, and Reprocessing 22.10.2026 32m

  Yesterday's click count changes and an advertiser disputes the result. Daniel and Maya design an analytics pipeline that can explain which events were counted, which day they belong to, and which rules were used.Late events, duplicate processing, hot campaigns, and replay challenge the baseline. Event time and processing time answer different questions. Rebuilding historical aggregates requires preserving definitions as well as data, so a newly computed number is not silently presented as an equivalent measurement.Chapters00:00:00 Welcome and a disputed click count00:03:07 Event identity and counting contracts00:06:54 A click through the aggregation pipeline00:10:42 Event time, late data, and corrections00:15:14 Hot campaigns and duplicate processing00:19:13 Replay and versioned counting rules00:22:44 Cost, monitoring, and bounded recovery00:29:53 Final count and supporting evidenceDaniel is the interviewer, voiced by Liam. Maya is the engineer, voiced by Bella. AI narration at a calm pace, with three thinking pauses.Opening music: Stealth Groover by Kevin MacLeod (incompetech.com). Licensed under Creative Commons Attribution 4.0: https://creativecommons.org/licenses/by/4.0/ . Edited excerpt with EQ, volume automation, a smooth fade, and narration.
• 
  S2 E13 — Design a Payment System: Idempotency, Ledgers, and Reconciliation 22.10.2026 33m

  A provider charges a customer, but the payment service crashes before saving the answer. Daniel and Maya design recovery around durable payment identities and uncertain outcomes rather than treating a timeout as a declined charge.Provider attempts, idempotency, ledger entries, refunds, and payouts are connected in one lifecycle. A balanced ledger can still record the wrong business event. Reconciliation supplies independent evidence for missing or mismatched results, with controlled repair instead of an unexplained adjustment to a balance.Chapters00:00:00 Welcome and an uncertain charge00:03:08 Payment IDs, money, and lifecycle states00:07:05 A complete payment and its records00:11:32 Provider timeouts and idempotent recovery00:16:08 Ledger entries and business correctness00:20:44 Refunds, payouts, and new requirements00:24:49 Reconciliation and controlled repair00:30:44 Final money flow and interview recapDaniel is the interviewer, voiced by Liam. Maya is the engineer, voiced by Bella. AI narration at a calm pace, with three thinking pauses.Opening music: Stealth Groover by Kevin MacLeod (incompetech.com). Licensed under Creative Commons Attribution 4.0: https://creativecommons.org/licenses/by/4.0/ . Edited excerpt with EQ, volume automation, a smooth fade, and narration.
• 
  S2 E12 — Design a Distributed Job Scheduler: Leases, Retries, and Cancellation 22.10.2026 33m

  A worker loses its lease, its replacement starts, and both eventually finish the same export. Daniel and Maya design a scheduler that distinguishes a job, a scheduled run, and an execution attempt.The flow follows durable scheduling, assignment, lease expiry, cancellation, and dependency recovery. Midnight bursts add fairness and missed-schedule decisions. Detecting an expired worker is only part of recovery: the design must also prevent that old worker from later publishing a result it no longer owns.Chapters00:00:00 Welcome and duplicate job completion00:03:17 Jobs, scheduled runs, and attempts00:07:11 A recurring job through the scheduler00:11:16 Leases, stale workers, and fencing00:16:00 Cancellation and dependency recovery00:19:57 Schedule bursts, fairness, and missed runs00:23:55 Monitoring and safe scheduler changes00:30:18 Result publication and final recapDaniel is the interviewer, voiced by Liam. Maya is the engineer, voiced by Bella. AI narration at a calm pace, with three thinking pauses.Opening music: Stealth Groover by Kevin MacLeod (incompetech.com). Licensed under Creative Commons Attribution 4.0: https://creativecommons.org/licenses/by/4.0/ . Edited excerpt with EQ, volume automation, a smooth fade, and narration.
• 
  S2 E11 — Design Notifications and Webhooks: Retries, Deduplication, and Delivery 22.10.2026 33m

  The delivery system records one logical notification, but the customer receives two emails. Daniel and Maya trace durable intent through scheduling, workers, provider calls, acknowledgments, and retries to find the uncertainty.A slow webhook endpoint tests fleet fairness, while outbound requests introduce security boundaries. Local deduplication alone cannot erase an external side effect whose response was lost. Monitoring and repair tools provide a way to investigate an uncertain delivery before sending it again.Chapters00:00:00 Welcome and a duplicate email00:03:09 Delivery contracts and channel ownership00:06:50 Durable intent to provider delivery00:10:41 Lost acknowledgments and deduplication00:15:23 Slow endpoints and fair worker use00:20:05 Webhook security and outbound requests00:24:07 Monitoring, repair, and safe changes00:30:42 Final delivery flow and guaranteesDaniel is the interviewer, voiced by Liam. Maya is the engineer, voiced by Bella. AI narration at a calm pace, with three thinking pauses.Opening music: Stealth Groover by Kevin MacLeod (incompetech.com). Licensed under Creative Commons Attribution 4.0: https://creativecommons.org/licenses/by/4.0/ . Edited excerpt with EQ, volume automation, a smooth fade, and narration.
• 
  S2 E10 — Design a Distributed Rate Limiter: Token Buckets and Global Quotas 22.10.2026 32m

  A customer sends requests through three regions and stays below each local limit while exceeding the intended total. Daniel and Maya define the quota contract before choosing token buckets or counters.They build a regional limiter, spend its last credit concurrently, and lose an admission response. Global budgets, hot tenants, combined limits, and partitions force a choice between coordination and allocated capacity. The final design states its allowed error and failure policy alongside the algorithm.Chapters00:00:00 Welcome and a customer in three regions00:02:22 Defining quotas, windows, and errors00:05:55 A regional token-bucket limiter00:09:44 Last-token races and lost responses00:14:50 Global budgets during partitions00:21:51 Hot tenants and combined limits00:25:40 Failure policy, monitoring, and recovery00:29:13 Final rate-limit contract and recapDaniel is the interviewer, voiced by Liam. Maya is the engineer, voiced by Bella. AI narration at a calm pace, with three thinking pauses.Opening music: Stealth Groover by Kevin MacLeod (incompetech.com). Licensed under Creative Commons Attribution 4.0: https://creativecommons.org/licenses/by/4.0/ . Edited excerpt with EQ, volume automation, a smooth fade, and narration.
• 
  S2 E09 — Design a Ride-Hailing System: Location Search and Driver Matching 22.10.2026 32m

  Two riders are offered the same nearby driver. Daniel and Maya design the path from location updates and geographic candidate search to an assignment that one authority can confirm.A driver's latest reported position is not a reservation of their time. Stale locations, simultaneous offers, offer expiry, and city-level ownership test that distinction. The interview follows matching under scale and network partitions, then reconstructs the final flow to show exactly who may promise a driver to a rider.Chapters00:00:00 Welcome and two riders, one driver00:02:42 Matching contracts and city ownership00:06:05 Location-to-assignment baseline00:09:21 Stale locations and candidate search00:14:59 Concurrent offers and driver assignment00:20:18 Scaling matching during partitions00:26:14 Monitoring and safe recovery00:29:47 Final assignment authority and recapDaniel is the interviewer, voiced by Liam. Maya is the engineer, voiced by Bella. AI narration at a calm pace, with three thinking pauses.Opening music: Stealth Groover by Kevin MacLeod (incompetech.com). Licensed under Creative Commons Attribution 4.0: https://creativecommons.org/licenses/by/4.0/ . Edited excerpt with EQ, volume automation, a smooth fade, and narration.
• 
  S2 E08 — Design File Sync: Offline Edits, Conflicts, and Versioning 22.10.2026 33m

  Two laptops edit the same file offline and reconnect after a flight. Daniel and Maya design synchronization that recognizes both versions rather than silently choosing the later timestamp.Stable file identities, version relationships, local edits, and a durable change history establish the baseline. The interview then follows conflict detection and safe application of a remote version while local work continues. Chunks, storage reclamation, client rollouts, and recovery complete the path from one edit to a converged file history.Chapters00:00:00 Welcome and two offline edits00:02:52 File identity, paths, and versions00:06:55 A local edit through the sync service00:10:52 Detecting concurrent file versions00:15:51 Applying remote edits without data loss00:21:31 Chunks, storage, and garbage collection00:26:54 Client rollout, monitoring, and recovery00:30:26 Final conflict resolution and recapDaniel is the interviewer, voiced by Liam. Maya is the engineer, voiced by Bella. AI narration at a calm pace, with three thinking pauses.Opening music: Stealth Groover by Kevin MacLeod (incompetech.com). Licensed under Creative Commons Attribution 4.0: https://creativecommons.org/licenses/by/4.0/ . Edited excerpt with EQ, volume automation, a smooth fade, and narration.
• 
  S2 E07 — Design a Video Platform: Uploads, Transcoding, and Playback 22.10.2026 33m

  A worker dies near the end of processing a large video. Daniel and Maya build a video platform that can retry useful work without publishing an incomplete playback result.The flow connects resumable uploads, object storage, transcoding attempts, renditions, manifests, and delivery. Queue pressure and scarce processing capacity introduce backpressure and fairness. The interview asks who may publish a completed version, how a viewer learns that it is ready, and what deletion and recovery must preserve.Chapters00:00:00 Welcome and a failed processing job00:02:48 Control services versus media delivery00:06:22 Upload, processing, and playback baseline00:10:06 Resumable uploads and safe retries00:15:08 Renditions and complete manifests00:20:53 Backpressure and fair transcoding00:27:25 Monitoring, rollouts, and deletion00:30:26 Final upload-to-playback flowDaniel is the interviewer, voiced by Liam. Maya is the engineer, voiced by Bella. AI narration at a calm pace, with three thinking pauses.Opening music: Stealth Groover by Kevin MacLeod (incompetech.com). Licensed under Creative Commons Attribution 4.0: https://creativecommons.org/licenses/by/4.0/ . Edited excerpt with EQ, volume automation, a smooth fade, and narration.