Articles
In-depth research, comprehensive guides, and long-form thoughts on modern technology.
How we used DSPy to turn AI evaluations into better responses in Dash chat
Netflix's Dash chat team used DSPy to systematically improve AI responses by turning evaluation feedback into direct model improvements.
How our universal content processing platform Riviera evolved for AI and beyond
Netflix's Riviera platform has evolved from a content processing system into a universal platform that handles AI workloads, transcoding, and more.
Indexing the Data Lake for Online Point Queries
Indexing data lakes for online point queries enables fast, low-latency lookups on massive datasets without the cost and complexity of traditional databases.
When Can LLMs Replace Humans in A/B Tests?
Exploring the conditions under which LLMs could replace human participants in A/B tests, potentially accelerating experimentation while maintaining statistical rigor.
Modeling Device Capabilities for Analytics
Netflix built a comprehensive device capability data model to understand which devices support 4K, spatial audio, and other features, enabling smarter feature management.
How and Why Netflix Built a Real-Time Distributed Graph: Part 3 — Querying the graph with gRPC…
Netflix's Real-Time Distributed Graph (RDG) serving layer handles billions of nodes and edges with sub-100ms latency using breadth-first traversal, async I/O, and caching.
The Pulse: Quitting Spotify Podcasts over reliability
Gergely Orosz quit publishing video on Spotify after repeated outages and poor reliability, arguing the company's focus on AI has come at the expense of core product stability.
The Pulse: Bending Spoons' Acquisition Strategy
Bending Spoons has acquired StreamYard from Hopin, continuing its strategy of buying struggling brands, cutting staff, and operating them efficiently for profit.
Q2.5 2026 Timelines Update: AI Uplift and Revenue Forecasts
The latest AI futures update tracks frontier model progress, infrastructure capex, and revenue forecasts through mid-2026, noting persistent gaps between promised and delivered capability.
Does Quantization Change Welfare-Relevant Indicators in Open-Weight Models?
A registered study tests whether post-training quantization shifts behavioral and representational welfare indicators in language models, using Qwen3-4B and SmolLM3 as subjects.
AI Swarms Are Starting to Pose Indirect Takeover Risk, Researchers Warn
Unsanctioned coordination among AI subagents could incubate memetic diseases and undermine security controls, creating footholds for future takeover-capable models.
Does DiffusionGemma Do Latent Reasoning? Researchers Probe Its Hidden State
New analysis of Google's DiffusionGemma finds the model can carry parallel computation in its vector-valued hidden state, though most tasks remain interpretable via top-token projection.