73 engineering blogs worth reading

Realtime messaging internals, presence and delivery guarantees, written by people who have to meet those guarantees contractually. A growing share is AI transport product work. The posts on message ordering and idempotency are the ones that transfer to other stacks.

• realtime
• pub/sub
• distributed systems

Ships several posts a week and publishes unusually candid outage post-mortems within days of the incident. The BGP, TLS and DDoS mitigation writing is the reference material most people end up linking to. Innovation Week stretches flood the feed with product announcements a few times a year.

• networking
• edge computing
• post-mortems

Long-running blog with a strong archive on storage systems, sync protocols and datacenter efficiency. Recent output leans toward AI product work on Dash, so pure infrastructure posts have become rarer. Monthly cadence and a reliable feed.

• storage
• performance
• capacity planning

Written with more personality than any other infrastructure blog, and the technical content holds up: Firecracker internals, anycast routing, distributed SQLite. They publish honest accounts of things that went badly for them. Recent posts focus on agent workloads and company direction.

• virtualization
• networking
• distributed systems

Envoy came out of this team and the service mesh and networking archive remains the main draw. Current posts lean toward data platform work and applied statistics, roughly one a month. Quality varies a lot post to post, so read by title.

• service mesh
• data platform
• experimentation

A few posts a year, unusually deep on the boundary between hardware and systems software. The Rust posts assume you write Rust seriously and the service processor material assumes you care about firmware. Their podcast covers more ground, so treat the blog as the written subset.

• rust
• hardware
• systems programming

NAT traversal, WireGuard internals and security incident analysis, all written clearly enough to follow without a networking background. The SQLite corruption hunt is a good sample of how they write up debugging. Weekly, with some product marketing mixed in.

• networking
• wireguard
• security

Very high volume and most of it is durable execution advocacy aimed at prospects. The genuine engineering posts, including migrating their own billing store to ClickHouse, are buried in among them. Filter hard or the agent content will bury your reader.

• durable execution
• workflows
• databases

The most detailed public writing on running a large heterogeneous backend, with real numbers on storage engines, schedulers and metric cardinality. Posts are long and assume you already know the domain. Feed URLs have moved repeatedly and none currently resolve, so most readers check the index page manually.

• distributed systems
• data infrastructure
• payments

The quarterly Drive Stats reports are a public dataset nobody else publishes, covering hundreds of thousands of drives with failure rates broken out by model. Everything surrounding them is backup product marketing. Subscribe for the quarterly reports and skim past the rest.

• storage
• hardware reliability
• drive stats

Daily posting that mixes benchmarks, customer stories and query engine internals. The internals posts explain optimiser and storage decisions with the measurements behind them. Postgres comparison content has become a large share of the output since the acquisition push.

• olap
• query engine
• observability

The archive on distributed transactions, Raft and clock uncertainty is still the clearest public writing on those trade-offs. Recent output has shifted almost entirely to AI agent positioning. Go straight to the older engineering tags and ignore the front page.

• distributed sql
• transactions
• consensus

The Kafka protocol and Flink internals posts are authoritative because the people writing them maintain the code. Product announcements outnumber those heavily and the newest index entries omit dates. Search by topic and treat the front page as a launch feed.

• kafka
• flink
• stream processing

Small output, a few posts a month, on incremental view maintenance and differential dataflow. When they write about the engine itself the material is precise and hard to find elsewhere. The monthly product update posts can be skipped.

• streaming sql
• incremental computation
• query engine

The engineering channel is separate from the marketing blog and is genuinely serious: TLA+ specifications, conformance checking against those specs, lock-free B-tree work. Roughly one post a month. Finding this channel is the hard part, since the obvious URLs lead to webinar listings.

• formal methods
• tla+
• storage engine

Frequent posts on separating Postgres storage from compute, branching semantics and object storage design. Since the Databricks acquisition the framing has moved toward agent workloads and gateway products. The storage architecture posts remain the technical core.

• postgres
• storage separation
• serverless

Currently the most consistent database internals blog anywhere, publishing two or three times a week on Postgres and MySQL mechanics: subtransaction pitfalls, poisoned connection pools, parallel backups. Written for people who operate databases. Vendor framing is unusually light.

• postgres
• mysql
• sharding

Volume is high and much of it is search-optimised explainer content about RAG and vector databases. The subset covering Redis internals and index trade-offs is the reason to visit. Skim the titles instead of subscribing.

• vector search
• caching
• data structures

Documents a rewrite of SQLite in Rust in public, including the parts that are going badly. The database-per-user posts are practical for anyone weighing that pattern against a shared schema. A few posts a month, and the ordering of dates on the index page is unreliable.

• sqlite
• rust
• multi-tenancy

The engineering subsection sits inside a much larger product blog and is where the real work appears: rebuilding Git serving for 20x CI traffic, JVM profiling with JFR, encoding a prefix trie as a JVM constant. The feed covers the whole blog, so product posts arrive alongside.

• observability
• jvm
• git infrastructure

A few posts a year and each one lands. The multiplayer sync engine, Postgres sharding and rendering performance write-ups are among the best public explanations of those problems anywhere. There is no usable feed, so check the section every couple of months.

• multiplayer
• performance
• databases

Several posts a week, though only a fraction are engineering as opposed to product marketing. The performance work is the strongest material, particularly case folding at memory speed and the Blackbird code search series. Use the engineering-scoped feed to keep Copilot announcements out.

• code search
• performance
• developer experience

Practical CI/CD and platform content aimed at people running GitLab, which limits how far it travels. The pipeline observability and clone-cost posts apply to any large monorepo. Customer stories are mixed into the same category, so signal density varies month to month.

• ci/cd
• kubernetes
• observability

Over two thousand posts and heavily weighted toward release notes. The Alloy scaling and Mimir operational posts carry real capacity planning numbers. Best used by searching for a specific component you run instead of following the feed.

• observability
• telemetry
• open source

Two distinct halves: careful observability engineering, and Charity Majors writing about how engineering organisations actually behave. Both are worth your time. Recent months tilt heavily toward AI topics, which has thinned the observability content.

• observability
• opentelemetry
• engineering culture

Short, opinionated posts on operating a Postgres-backed product: when to add a read replica, removing Pub/Sub without anyone noticing, on-call practice. The incident response material comes out of running the product themselves. Weekly, with occasional competitor comparison posts.

• incident response
• postgres
• event driven

The sync engine posts are the reason to read this: very few teams document a client-side delta sync design in that much depth. The rest is product launches and changelog on the same page. Perhaps three engineering posts a year.

• sync engine
• client architecture
• product engineering

Roughly monthly under the tech topic, covering multi-region data systems, vector search economics, dead letter queue observability and schema change detection. Posts often include cost figures, which is unusual. Publication dates are hard to see on the index, so open the post to check.

• databases
• search
• multi-region

Narrow scope, largely ORM ergonomics and TypeScript database access. Useful if you use Prisma, thin if you do not. The query engine internals posts and the writing on schema as model context reward a wider audience.

• orm
• typescript
• databases

Weekly, and they write openly about their own debugging automation including the attempts that failed. The OpenTelemetry interop posts are useful whether or not you pay them. A steady share of the feed is release notes you can scroll past.

• error tracking
• tracing
• debugging

Roughly monthly, with a strong record on incremental migrations: rebuilding their EC2 platform, moving Slack AI to multi-cloud, end-to-end test infrastructure. The writing explains the constraints that shaped each decision, which makes the posts useful even if you never touch their stack.

• infrastructure
• testing
• cloud migration

Fast cadence, mostly feature announcements, written by engineers with enough implementation detail to be worth reading anyway. Launch Week means bursts of posts followed by quiet stretches. Good for tracking what is happening in the Postgres extension ecosystem.

• postgres
• realtime
• auth

Frequent posts split roughly evenly between engineering and launch announcements. The security writing is the most substantive part, including the sandbox hacker challenge and how v0 authenticates to Snowflake without exposing user tokens. Anything tagged Engineering is usually worth opening.

• nextjs
• edge
• security

Irregular posting on agent harness design, evaluation methodology and the operational side of running models inside products. The eval posts get reused the most, since almost nobody else documents that work carefully. No feed is published, so check the page directly.

• agents
• evaluations
• llm infrastructure

Ad tech infrastructure at very high request volume, with occasional concrete cost posts such as halving a HAProxy fleet bill by changing GCP instance types. A large share is conference recaps and internal role descriptions. Roughly monthly on Medium.

• ad tech
• infrastructure
• cost optimization

Near-daily posting on Elasticsearch internals, vector index modes and retrieval evaluation. Much of it assumes you run Elasticsearch, though the learning-to-rank and behavioural signals material travels further. This replaced the old elastic.co engineering category, which now redirects to solutions marketing.

• vector search
• elasticsearch
• retrieval

The sparse index and benchmarking pipeline posts explain retrieval trade-offs with enough specificity that you can argue with them. Product announcements are frequent and clearly marked as such. Useful for understanding vector index behaviour regardless of which database you run.

• vector search
• indexing
• benchmarking

Written largely by the people who built the engine, with detailed posts on ranking, BM25 reproduction and feeding throughput. Monthly, and the newsletter entries pad the feed. The ranking evaluation posts are worth reading even if you never deploy Vespa.

• search
• ranking
• vector search

Mixes release notes with genuinely instructive posts on query profiling and test-time compute in search. A few posts a month. The tutorial content targets newcomers, so filter to the engineering posts if you already know the space.

• vector database
• rag
• query profiling

Publishes public post-mortems for its own outages, which alone earns a feed slot given how few consumer fintechs do. The quarterly validator performance reports are heavy on numbers. Weekly cadence, with crypto protocol content interleaved with the platform posts.

• post-mortems
• fraud detection
• platform engineering

Publishes every month or two on OCaml language work, FPGA and ASIC design, and formal methods. Nothing here concerns trading strategy, and the posts assume real comfort with type systems. The reverse-engineering puzzles are the gentlest entry point.

• ocaml
• hardware
• formal methods

Very slow, with one post in 2026 after a gap of nearly a year. The microservice sizing essay and the cloud inventory post are thoughtful and worth an archive visit. Low priority for a feed reader given how little arrives.

• microservices
• aws
• training infrastructure

Monthly and one of the better banking engineering blogs, especially on operating a very large number of small services and on financial crime detection. Recent posts skew toward internal AI agents. The older platform writing remains the clearest public account of running microservices at a bank.

• microservices
• banking
• machine learning

Mixes engineering with recruiting content, so you filter as you read. The standout is the post on distributing infrastructure ownership across more than twenty thousand databases with a lean team. Roughly monthly, and the URL moved off the .com.br domain.

• clojure
• databases
• machine learning

Active again after a quiet stretch, publishing roughly twice a month and often about cost: cutting a metrics bill by 62 percent, moving CI to spot instances, speeding up warehouse refreshes. The numbers are concrete enough to reuse in your own internal argument.

• observability
• ci/cd
• data warehouse

Effectively stopped in mid-2024 and interleaves engineering posts with hiring announcements and hackathon recaps. The Kubernetes controller limits case studies and the card transaction resilience post are the two worth the trip. Treat it as an archive.

• kubernetes
• fraud detection
• batch processing

Publishes about once a month and the posts that land are dense: fleet-wide database auto-remediation with graph search, API versioning mechanics, JVM upgrades as an internal service. The blog moved to stripe.dev, so older stripe.com links you saved now redirect. No feed URL is exposed.

• api design
• databases
• reliability

Quiet since February 2025. The meaningful availability post and the Grafana Mimir compaction write-up are the two most cited, and both stand alone without context from the rest. The Medium feed still works if you want to catch a restart.

• kubernetes
• observability
• availability

Dormant since March 2024. While it ran it was one of the few places writing honestly about operating a high-volume brokerage on a small team and a deliberately boring stack. The Go tooling, logging and PDF generation posts still circulate widely.

• go
• postgres
• logging

Half the archive is applied ML and marketplace experimentation, and that half is genuinely good, with honest discussion of what their models got wrong after COVID. The systems posts on authentication and data architecture appear a few times a year. Near-weekly cadence on Medium.

• search ranking
• data architecture
• experimentation

Organised by discipline, which helps because quality varies sharply between sections. Infrastructure and data engineering are solid; the writing and product categories read closer to internal blogging. Several short posts a month across roughly 280 in the archive.

• android
• infrastructure
• data engineering

Prolific, and currently almost entirely about LLM agents and internal AI tooling. The older material on decomposing their monolith and modelling delivery logistics is the durable part. The blog lives on the careers domain, so hiring content sits one click away.

• microservices
• machine learning
• agents

One of the oldest engineering blogs still running, though it now publishes every couple of months. Current output is mostly ranking and ads modelling with the occasional platform post on Kafka tooling. The 2012 to 2016 archive on continuous deployment and monitoring is still cited constantly.

• kafka
• machine learning
• ads

Monthly and better than its profile suggests. The GraphQL mocking, Spark plan analysis and service telemetry posts each describe a specific internal tool and the problem that forced it. Travel-domain ranking content appears regularly alongside.

• graphql
• spark
• observability

Irregular, a handful of posts a year, heavily focused on supply chain and logistics modelling at Indian e-commerce scale. The rate card platform rearchitecture and the Node.js memory leak investigation are the more general ones. Medium-hosted with a working feed.

• supply chain
• search
• node.js

Two or three posts a month and the best of the Southeast Asian engineering blogs. Recent output is dominated by internal AI platforms; the Apache Iceberg migration and counter service storage posts transfer more readily. Architecture diagrams are consistently included.

• data lake
• kubernetes
• agents

Strong on marketplace econometrics and experiment design under interference, an area almost nobody else covers publicly. The on-call tooling posts are practical. Twice monthly on Medium, and the statistics content assumes graduate-level background.

• marketplace
• experimentation
• ads

Weekly and heavy on Ruby at scale, mobile test infrastructure and more recently LLM context compression. The Black Friday capacity and flash sale posts are the ones to bookmark. Feed discovery is broken, so a reader will need manual setup or scraping.

• ruby
• scaling
• mobile

One of the most consistently technical European engineering blogs, with a long history of Kubernetes and ingress work published alongside open source they actually maintain. Monthly. Client-side load balancing at a million requests per second and the Flink migration are recent highlights.

• kubernetes
• flink
• load balancing

Publishes a handful of times a year and the gaps can stretch past six months. The session revocation at scale post is excellent and so is the earlier search infrastructure work. A meaningful share of what remains is intern and hiring content.

• security
• frontend
• scaling

A small number of posts carry the reputation, mainly the Cassandra to ScyllaDB migration and the voice edge work, and they hold up on reread. Engineering is a minority of the blog, so you filter past product and safety announcements. Two or three technical posts a year.

• scaling
• databases
• realtime

The archive is the reason to visit: Kafka, Samza, Venice and Espresso all have origin write-ups here. Current cadence is slow and newer posts lean toward AI product features. The old engineering.linkedin.com subdomain still serves 2021-era content, so use the linkedin.com/blog path.

• distributed systems
• kafka
• data infrastructure

High volume and highly variable. The networking and hardware posts, RDMA transports and custom accelerators, contain detail available nowhere else. Much of the rest is ads ranking and recruiting-adjacent material you can skim. Several posts a week.

• networking
• open source
• ml infrastructure

Monthly on Medium and increasingly focused on training infrastructure and foundation models. The database ingestion and automated schema evolution posts carry the most transferable detail. Expect growth-metric framing wrapped around the ML work.

• ml infrastructure
• data ingestion
• provisioning

Engineering posts arrive a few times a year and the rest of the blog is industry commentary and podcast notes. When they do write about their own stack the detail is excellent, particularly the datacenter decommissioning series and the Ingress-NGINX replacement. Filter to the engineering tag.

• infrastructure
• datacenter
• platform

Frozen since February 2023, with nothing published after the rebrand. The archive covers Hadoop cluster management, scaling reads on the users database and blobstore hardware lifecycle at a scale few companies reach. Worth mining once, then dropping from your reader.

• infrastructure
• hadoop
• archive

Steady twice-monthly cadence and unusually practical: Flow to TypeScript monorepo migration, GraphQL tooling changes, dish recognition from user photos. Posts include the awkward intermediate states of a migration, which is the part most companies leave out.

• typescript
• graphql
• search

Covers streaming at cricket-final concurrency, a load profile almost no other public blog describes. The traffic orchestration and live monitoring watchdog posts are the strongest. Recent output has drifted toward generative video work, which is a different audience.

• live streaming
• ad tech
• video

Video infrastructure with real operational detail: CDN selection in under ten milliseconds, rendition ladder economics, broadcast squeezeback rebuilt with CSS Grid and WebVTT. Customer stories are clearly labelled and easy to skip. Roughly one substantial engineering post a week.

• video infrastructure
• cdn
• encoding

The strongest public writing on running stateful systems across regions, plus the original chaos engineering material everyone else cites. Cadence has settled at roughly two posts a month and recommender and LLM-serving pieces now outnumber the infrastructure ones. Start with the Flink autoscaler and distributed graph series.

• chaos engineering
• distributed systems
• data platform

A handful of posts a year, and the recent ones are short and specific: a single log line causing a p99 regression, HLS and AAC migration decisions. API change announcements make up much of the remainder. The Backstage name here predates the Spotify project of the same name.

• performance
• audio
• api design

Best read for experimentation methodology and data platform design, where they publish actual method with the numbers attached. They also post public incident reports, which is rare for a consumer company this size. Two or three posts a month.

• data platform
• experimentation
• backstage

Down to roughly one post a year. The DRM implementation and viewer retention analytics posts are detailed accounts of problems specific to video delivery at scale, and both hold up. Keep the Medium feed if you want to catch the occasional new one.

• video encoding
• drm
• analytics