Multi‑Ratio Video Resizing for Mobile Feed and Marketing — Netflix Tech Blog
An automated three‑stage pipeline—shot detection with active‑speaker split, focus‑region discovery using faces and saliency, and temporally smoothed virtual‑camera cropping—generates 9:16, 4:5 and 1:1 variants in a single pass and beats…
💡 Worth it if you build mobile video pipelines that need fast multi‑aspect cropping
If an AI agent can read untrusted input, access private data, and send data out, it forms the Lethal Trifecta; splitting duties into separate agents so no single agent has all three stops the attack.
💡 Worth it if you design AI agents that handle external data and secret access
Beyond API Compatibility: Building a Tool to Measure the Blast Radius of Library Changes — Netflix Tech Blog
Quasar indexes symbol definitions and references across Netflix repos, then on a PR reports the exact count of consumer repositories that call a changed API (e.g., 137) with version breakdown, so producers can gauge blast radius before…
💡 Worth it if you manage library releases across many services
LLMs are making formal software verification feasible at a much larger scale — Anthropic
Claude proved that fuzz tests for 239 entry points across 123 OSS‑Fuzz libraries never crash for any input under 64 MB, generating ~180 million lines of Rocq proofs and surfacing hundreds of hidden bugs.
💡 Worth it if you need formal security guarantees for many open‑source libraries