engblogs

Local-first software empowers users with full ownership and control of their data by prioritizing local storage and real-time collaboration, overcoming cloud app limitations such as dependence on servers, privacy concerns, and offline access issues through technologies like CRDTs and peer-to-peer networking.

Exploring the concept of local-first software and its ability to empower users by giving them ownership of their data, even in a cloud-based environment.

Exploring the latest advancements in AI technology through the OpenAI Five Finals

SoundCloud slashed reindexing time from weeks to one hour by implementing a Kappa-architecture streaming pipeline with Kafka and Elasticsearch, using bulk API sizing, optimized primary shards, and translog/refresh tuning to reindex over one billion documents.

SoundCloud shows how to reindex a billion documents in one hour by deploying a Kappa Architecture with Kafka-backed ingestion and bulk-optimized Elasticsearch settings, including shard tuning and translog/refresh adjustments to accelerate indexing and reduce search latency.

Exploring implicit generation and generalization methods for energy-based models

Optimizing Elasticsearch to reindex one billion documents in one hour at SoundCloud

Highlighting the achievements of OpenAI Scholars 2019

Exploring the potential of OpenAI LP in the current technological landscape.

A technical overview of SoundCloud's off-platform validation architecture, detailing a hybrid of automatic and manual metadata checks, asynchronous feedback, and cached evidence to enable fast, compliant distribution across partners.

A deep dive into a hybrid off-platform validation system for music distribution that blends automatic and manual checks to deliver fast, asynchronous feedback while enforcing metadata quality and streaming partner requirements.

Exploring the power of Activation Atlases in neural networks