# Byte-Sized Design Master system design concepts, engineering fundamentals, and interview basics. Weekly summaries, post-mortems, and advice for 43,000+ engineers. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## Multi-Agent Architecture and Microservices: What Still Works and Where It Breaks DevFeed: [Multi-Agent Architecture and Microservices: What Still Works and Where It Breaks]() Original publisher: [Read original article]() Author: Byte-Sized Design Published: 2026-09-12T17:56:36Z Content type: opinion Language: en Sources: [Byte-Sized Design]() Topics: [Microservice](), [Architecture & Design]() Tags: [agent](), [architecture](), [microservices](), [multi-agent]() ### AI overview An opinion article comparing multi-agent architecture with microservices. It discusses what still works, three places the approach breaks, and one current heuristic. ### Source excerpt The parts that still work, the three places it breaks, and one heuristic for right now ## 🔌 Stripe's Proxy Was Opening Hundreds Of Thousands Of Connections DevFeed: [🔌 Stripe's Proxy Was Opening Hundreds Of Thousands Of Connections]() Original publisher: [Read original article]() Author: Byte-Sized Design Published: 2026-09-08T15:36:04Z Content type: article Language: en Sources: [Byte-Sized Design]() Topics: [stripe](), [Go Language](), [Latency](), [servers](), [WebSocket]() Tags: [go](), [latency](), [proxy](), [servers](), [stripe](), [websocket]() ### AI overview The article explains why Stripe replaced Envoy with its own Go proxy, mesh-proxy. It attributes the change to connection scaling problems and reports lower CPU usage and latency, fewer connections, and additional burst capacity from Ruby load balancing. ### Source excerpt A Go proxy, one connection per host, and 3,000 days of latency saved every day ## The Four Vendor Relationships Commonly Involved in a Production Voice Feature DevFeed: [The Four Vendor Relationships Commonly Involved in a Production Voice Feature]() Original publisher: [Read original article]() Author: Byte-Sized Design Published: 2026-08-28T21:47:43Z Content type: article Language: en Sources: [Byte-Sized Design]() Topics: [API](), [Large Language Model](), [pii]() Tags: [2026](), [api](), [compliance](), [llm](), [pii](), [pii-redaction](), [production](), [transcription](), [voice]() ### AI overview The article describes how a production voice feature typically involves four vendor relationships: transcription, PII redaction, an LLM provider, and compliance review. ### Source excerpt TLDR A production voice feature in 2026 typically ships with four vendor relationships: a transcription API, a PII redaction step somebody built in a sprint, an LLM provider, and a compliance review stretched across all of it. ## Linear's Case for Building a Custom Agent Harness Instead of Using Agent Frameworks DevFeed: [Linear's Case for Building a Custom Agent Harness Instead of Using Agent Frameworks]() Original publisher: [Read original article]() Author: Byte-Sized Design Published: 2026-08-10T17:59:45Z Content type: opinion Language: en Sources: [Byte-Sized Design]() Topics: [Frameworks]() Tags: [agent](), [frameworks](), [harness](), [tools]() ### AI overview The article presents a case for building a custom agent harness instead of using agent frameworks, focusing on shallow tools and just-in-time skills. ### Source excerpt Shallow tools, just-in-time skills, and a defensible case for a custom harness ## How AI Agents Actually Talk to Each Other DevFeed: [How AI Agents Actually Talk to Each Other]() Original publisher: [Read original article]() Author: Byte-Sized Design Published: 2026-08-06T06:31:30Z Content type: opinion Language: en Sources: [Byte-Sized Design]() Topics: [Artificial Intelligence](), [Model Context Protocol]() Tags: [a2a](), [agents](), [ai](), [ai-agents](), [mcp]() ### AI overview The article discusses how AI agents communicate and argues that framing MCP and A2A as competing technologies describes a conflict that does not exist. ### Source excerpt And why "MCP vs A2A" is a fight that doesn't exist ## GitHub Observes That Debating Some Changes Can Cost More Than Building Them DevFeed: [GitHub Observes That Debating Some Changes Can Cost More Than Building Them]() Original publisher: [Read original article]() Author: Byte-Sized Design Published: 2026-07-20T18:07:06Z Content type: opinion Language: en Sources: [Byte-Sized Design]() Topics: [GitHub]() Tags: [github]() ### AI overview The article says GitHub is observing that, for one class of change, debating whether to proceed can cost more than building the change. ### Source excerpt Saying no used to be free. GitHub is noticing that for one class of change, the debate now costs more than just building the thing. ## Millions of Batch Jobs Migrated in Four Weeks DevFeed: [Millions of Batch Jobs Migrated in Four Weeks]() Original publisher: [Read original article]() Author: Byte-Sized Design Published: 2026-07-13T17:15:37Z Content type: article Language: en Sources: [Byte-Sized Design]() Topics: [jobs]() Tags: [batch](), [jobs]() ### AI overview The article describes migrating millions of batch jobs in four weeks without users noticing. ### Source excerpt Millions of batch jobs, migrated in 4 weeks, and nobody noticed ## OpenAI identified a race condition in a 100-picosecond window DevFeed: [OpenAI identified a race condition in a 100-picosecond window]() Original publisher: [Read original article]() Author: Byte-Sized Design Published: 2026-07-06T19:05:51Z Content type: article Language: en Sources: [Byte-Sized Design]() Topics: [race-condition](), [debugging](), [bug](), [OpenAI]() Tags: [bug](), [debugging](), [openai](), [race-condition]() ### AI overview The article describes how OpenAI identified a race condition occurring within a 100-picosecond window. ### Source excerpt How OpenAI stopped debugging like a doctor and started thinking like an epidemiologist and caught a race condition in the process. ## Cloudflare Found a Race Condition in a Widely Deployed Rust Library DevFeed: [Cloudflare Found a Race Condition in a Widely Deployed Rust Library]() Original publisher: [Read original article]() Author: Byte-Sized Design Published: 2026-06-30T17:31:07Z Content type: article Language: en Sources: [Byte-Sized Design]() Topics: [race-condition](), [Rust](), [Cloudflare]() Tags: [bug](), [cloudflare](), [race-condition](), [rust]() ### AI overview Cloudflare found a race condition in one of Rust's most-deployed libraries. The reported fix was four lines. ### Source excerpt Cloudflare found a race condition buried in one of Rust's most-deployed libraries, and the fix was four lines ## How Discord Automates ScyllaDB Clusters at Scale DevFeed: [How Discord Automates ScyllaDB Clusters at Scale]() Original publisher: [Read original article]() Author: Byte-Sized Design Published: 2026-06-15T06:30:00Z Content type: article Language: en Sources: [Byte-Sized Design]() Topics: [Discord](), [scylladb](), [Framework](), [Database]() Tags: [database](), [discord](), [framework](), [scale](), [scylladb]() ### AI overview An article about how Discord automates ScyllaDB cluster operations at scale, describing a framework that reduced a 36-hour database operation to two hours of mostly waiting. ### Source excerpt The framework that turned a 36-hour database operation into two hours of mostly waiting ## How MongoDB Atlas Built a Predictive Auto-Scaler DevFeed: [How MongoDB Atlas Built a Predictive Auto-Scaler]() Original publisher: [Read original article]() Author: Byte-Sized Design Published: 2026-06-01T06:07:00Z Content type: article Language: en Sources: [Byte-Sized Design]() Topics: [MongoDB](), [cloud-infrastructure](), [Server]() Tags: [atlas](), [mongodb](), [servers]() ### AI overview The article describes how MongoDB Atlas built a predictive auto-scaler designed to scale infrastructure before server capacity becomes a problem. ### Source excerpt How Atlas built a predictive auto-scaler that scales before your servers break a sweat ## Coinbase Outage Raises Questions About Distributed Backups, Latency, and Resilience During an AWS Thermal Event DevFeed: [Coinbase Outage Raises Questions About Distributed Backups, Latency, and Resilience During an AWS Thermal Event]() Original publisher: [Read original article]() Author: Byte-Sized Design Published: 2026-05-16T19:15:36Z Content type: article Language: en Sources: [Byte-Sized Design]() Topics: [coinbase](), [Resilience](), [Amazon Web Services](), [Latency]() Tags: [2026](), [aws](), [coinbase](), [distributed](), [latency](), [outage](), [resilience]() ### AI overview The article discusses a May 2026 AWS thermal event, a Coinbase outage, the tradeoff between latency and resilience, and the limits of backup systems described as distributed. ### Source excerpt The May 2026 AWS thermal event, the latency-versus-resilience tradeoff that broke coinbase, and what happens when "distributed backup" isn't actually distributed ## How Slack Uses a Three-Channel Context System for AI-Agent Security Investigations DevFeed: [How Slack Uses a Three-Channel Context System for AI-Agent Security Investigations]() Original publisher: [Read original article]() Author: Byte-Sized Design Published: 2026-05-08T07:42:17Z Content type: article Language: en Sources: [Byte-Sized Design]() Topics: [Slack](), [Artificial Intelligence](), [Security]() Tags: [agents](), [ai](), [ai-agents](), [channel](), [context](), [security](), [slack]() ### AI overview The article examines Slack's three-channel context system for a team of AI agents involved in security investigations. ### Source excerpt Inside the three-channel context system behind slack's security investigations ## HubSpot's 37-Minute Lesson in Why HTTP 200 Can Lie DevFeed: [HubSpot's 37-Minute Lesson in Why HTTP 200 Can Lie]() Original publisher: [Read original article]() Author: Byte-Sized Design Published: 2026-04-22T16:34:15Z Content type: article Language: en Sources: [Byte-Sized Design]() Topics: [hubspot](), [HTTP](), [Availability](), [Monitoring]() Tags: [availability](), [http](), [hubspot](), [monitoring](), [permission]() ### AI overview The article examines a HubSpot access incident in which a permission check passed while users were locked out, showing that an HTTP 200 response and conventional availability monitoring may not reflect whether users can actually access a service. ### Source excerpt The permission check that passed, the users who were locked out, and what monitoring for "availability" actually misses ## Slack Rebuilt Notifications for Millions of Users DevFeed: [Slack Rebuilt Notifications for Millions of Users]() Original publisher: [Read original article]() Author: Byte-Sized Design Published: 2026-03-30T01:14:20Z Content type: article Language: en Sources: [Byte-Sized Design]() Topics: [Slack]() Tags: [architecture](), [notifications](), [slack]() ### AI overview The article examines how Slack rebuilt its notification system and describes an architecture decision that enabled the change without breaking existing users. ### Source excerpt Slack rebuilt its notification system from scratch, here's the architecture decision that made it possible without breaking millions of users. ## Uber's Analysis of Hours-Old Stale Data DevFeed: [Uber's Analysis of Hours-Old Stale Data]() Original publisher: [Read original article]() Author: Byte-Sized Design Published: 2026-03-24T05:16:10Z Content type: opinion Language: en Sources: [Byte-Sized Design]() Topics: [data]() Tags: [batch](), [data]() ### AI overview The article discusses what Uber found when analyzing stale data that was hours old, and frames batch jobs as a potential liability. ### Source excerpt What they found when they finally did the math on stale data.