Digs

Searching X architecture returns material written to be found by that query, which means tutorials and vendor pages. The material worth having was written by an engineer explaining a real system to their peers, and it uses different words. A dig is what comes back from going after that layer on one topic: the architecture reconstructed across real systems, the decisions and the conditions that flip them, what actually broke, and numbers you can plan against. Every claim is graded, linked and dated.

91 field guides 2462 graded sources 422 postmortems 8 categories
Platform & Infrastructure 17 Sep 2026

Emulate, then own: ten years of Microsoft leaving its own compatibility layers

Three unrelated parts of Microsoft's developer platform ran the same experiment in public and ended the same way: ship a clever layer that imitates another system, operate it at scale, then replace it with the…

What surprised meNone of the three replacements removed the cost; each moved it to a narrower boundary and made it measurable, and in two cases Microsoft still recomm…

31 sources · 5 organisations · 8 postmortems 34 min read
compatibilitymonorepovirtualizationmigrationgit
Platform & Infrastructure 16 Sep 2026

Maintenance mode: a decade of Netflix retiring code that runs in other people's processes

Reconstructs a decade of Netflix platform retirement from package-registry timestamps, README status blocks, archive banners, unmerged pull requests and security advisories. A reader leaves with a way to predi…

What surprised meSurvival tracks what a component does, not how old or popular it is: every Netflix library that held policy an application had to configure is frozen…

29 sources · 6 organisations · 4 postmortems 35 min read
open-sourcedeprecationclient-librariesservice-discoverydependency-risk
Reliability & Operations 16 Sep 2026

The router finds out last: taking a server out of service without dropping requests

Reconstructs the five-step removal sequence that Google, Kubernetes, AWS, Envoy and gRPC all converge on, and shows that implementations differ on exactly one axis: whether the drain closes the loop by subscri…

What surprised meAfter a decade of deploy-time 502s, upstream Kubernetes codified the sleep itself as the fix (KEP-3960, GA v1.34) while the proposal to sequence SIGT…

30 sources · 22 organisations · 3 postmortems 32 min read
graceful-shutdownload-balancingkubernetesdeploysconnection-draining
Data Platforms 15 Sep 2026

Own the layer above the engine: ten years of LinkedIn's data platform

Reconstructs a decade of data-platform decisions at LinkedIn from artefacts rather than announcements: archive notices, dated release listings on Maven Central, PyPI, npm, Docker Hub and the Go proxy, publishe…

What surprised meThe registry record inverts the production record: Voldemort, retired from production in 2018, is fully downloadable, while Venice, the system Linked…

22 sources · 6 organisations · 5 postmortems 31 min read
linkedinderived-dataservingcatalogopen-source
AI & LLM Systems 15 Sep 2026

The fastest change in the stack is a block of text

A field guide to the change lifecycle of the production prompt, the highest-leverage, lowest-friction change surface in an LLM product. Built from the published incidents of OpenAI, xAI, DPD, Cursor and Air Ca…

What surprised meThe industry is running the same experiment in both directions at once: xAI re-coupled prompt changes to code's controls after three incidents in twe…

26 sources · 22 organisations · 4 postmortems 33 min read
promptsllm-operationschange-managementevalsrollback
Platform & Infrastructure 14 Sep 2026

Deciding what to stop building: ten years of Spotify, read from its own archive notices

Reconstructs a decade of platform decisions at Spotify from 289 public repositories and the dated notices attached to them, the package registries that record when software actually stopped shipping, and the 6…

What surprised meArchiving a repository at Spotify is a publishing decision rather than a retirement decision: the Styx scheduler's repository was discontinued in Jul…

36 sources · 10 organisations · 5 postmortems 35 min read
developer-platformopen-sourcedeprecationbackstagebuild-vs-buy
Cost & Efficiency 14 Sep 2026

The bill is not a brake

Reconstructs the spend-containment architecture from six published billing incidents (Milkie Way, Troy Hunt, the Netlify and empty-S3-bucket bills, Cara, the 2026 AWS estimates outage) and the platform mechani…

What surprised meThe hard cap was not never-built: Google built it, removed it (2019-2023) because its coverage could not keep up with the platform, and reintroduced …

34 sources · 31 organisations · 5 postmortems 36 min read
cloud-billingspend-capsdenial-of-walletserverlessquotas
Platform & Infrastructure 13 Sep 2026

Ten years of changing a runtime you are not allowed to break

A hosted runtime cannot ask its tenants to upgrade, so Cloudflare pins semantics to a date per deployment and releases the runtime every day. This guide reads the contract out of the one file where it is writt…

What surprised meThe flag namespace is accelerating rather than converging: 4 flags took a default date in 2021 and 26 have taken one in the first nine months of 2026…

24 sources · 5 organisations · 2 postmortems 30 min read
cloudflareworkersbackwards-compatibilityjavascript-runtimeplatform-evolution
Security & Identity 13 Sep 2026

The breach is a missing check. The outage is the checker.

Reconstructs the authorization layer from the accounts of Google, Airbnb, Carta, Netflix, Figma, Slack and Gojek: relationship tuples, precomputed indexes, caches with a consistency story, and the fail-open ch…

What surprised meThe paper's central contribution, consistency tokens against the new-enemy problem, is the least-copied part of the design: Ory Keto's zookie issue h…

26 sources · 24 organisations · 5 postmortems 29 min read
authorizationzanzibarrebacfail-openidor