Topic Backlog

The cron job can use this as inspiration, but should still avoid duplicates and may choose better current/adjacent topics.

Prediction markets / trading / Coinbase-relevant

• Order books vs automated market makers: why market structure changes system design
• Matching engines: deterministic sequencing, fairness, and latency
• Market data fanout: snapshots, deltas, sequence numbers, and backpressure
• Idempotency keys in trading and payments APIs
• Double-entry ledgers and why balances are derived, not edited
• Event sourcing for financial systems
• Exactly-once vs effectively-once processing in settlement flows
• Risk checks and kill switches in trading platforms
• Reconciliation pipelines between internal ledgers and external venues/chains
• Hot/warm/cold wallet architecture as a system design problem
• Circuit breakers in exchanges and distributed services
• Time, clocks, and ordering: Lamport clocks, vector clocks, and exchange sequencing

Social apps / apps at scale

• Fanout-on-write vs fanout-on-read for social feeds
• Ranking pipelines: candidate generation, scoring, and serving
• Notification systems: batching, dedupe, quiet hours, and preference fanout
• Rate limiting: token buckets, leaky buckets, and abuse prevention
• Caches: cache-aside, write-through, write-behind, stampede protection
• Redis from zero: in-memory key-value store, TTLs, commands, and common use cases
• Redis data structures: strings, hashes, lists, sets, sorted sets, streams
• Cache invalidation and stale data failure modes
• Cache stampedes, hot keys, jitter, request coalescing, stale-while-revalidate
• Redis for rate limiting, sessions, leaderboards, queues, and ephemeral state
• Redis persistence, replication, Sentinel, Cluster, and failover
• Cache observability: hit rate, latency, memory pressure, evictions, key cardinality
• What not to cache in trading systems: balances, orders, risk state, and correctness-critical data
• Multi-layer caching: browser, CDN, edge, app cache, database cache
• CDN architecture and media delivery
• Real-time presence and websockets at scale
• Graph data models: followers, blocks, mutes, recommendations
• Moderation queues and human-in-the-loop systems

Platform / reliability fundamentals

• Load balancers: L4 vs L7, health checks, retries, and connection draining
• Queues and streams: SQS-style queues vs Kafka-style logs
• Sharding and partition keys
• Replication, quorums, and consistency tradeoffs
• Backpressure and overload protection
• SLOs, SLIs, error budgets, and alert design
• Observability: logs, metrics, traces, exemplars
• Deployments: blue/green, canary, feature flags, rollbacks
• Multi-region design: active-active vs active-passive
• Data migrations with zero downtime

Case-study-driven lesson ideas

• Coinbase 2026-05-07 incident: thermal event, failed isolation boundary, single-zone exchange primary, and latency vs availability tradeoffs
• Coinbase 2026-05-07 incident: why Kubernetes workloads could be drained but exchange/Kafka could not
• Coinbase 2026-05-07 incident: matching-engine quorum, leader election, split-brain avoidance, and safe trading halt
• Coinbase 2026-05-07 incident: Kafka partition failover, TiB-scale recovery, replication catch-up, and delayed balance streams
• Coinbase 2026-05-07 incident: Datadog/observability-style detection of quote failures, Sev1 response, runbooks, and RCA
• Coinbase 2026-05-07 incident: cancel-only mode, auction mode, product-state audit, and safe market reopening

Source inspiration

• Designing Data-Intensive Applications
• System Design Interview / Alex Xu series
• Google SRE books
• Kleppmann papers/talks
• Coinbase, Stripe, Uber, Meta, Twitter/X, Netflix, Discord, Slack, Cloudflare, AWS, and Google engineering blogs