https://sparsenotes.com/series/weekly-videos-2026-08/ Recent content in weekly-videos-2026-08 on Sparse Notes https://sparsenotes.com/images/og-default.png https://sparsenotes.com/images/og-default.png Hugo -- gohugo.io Sat, 08 Aug 2026 00:00:00 +0000 https://sparsenotes.com/posts/2026/08/inventing-on-principle-bret-victor/ Sat, 08 Aug 2026 00:00:00 +0000 https://sparsenotes.com/posts/2026/08/inventing-on-principle-bret-victor/ Bret Victor's 2012 CUSEC talk argues that creators need an immediate connection to what they create, then reframes technical work as a form of activism. https://sparsenotes.com/posts/2026/08/reliability-lessons-from-sqlite/ Sat, 08 Aug 2026 00:00:00 +0000 https://sparsenotes.com/posts/2026/08/reliability-lessons-from-sqlite/ Richard Hipp explains how avionics-grade MC/DC test coverage, fault injection, and simulated power loss let three committers maintain the most widely deployed software on Earth. https://sparsenotes.com/posts/2026/08/stanford-cs329a-self-improving-ai-agents/ Sat, 08 Aug 2026 00:00:00 +0000 https://sparsenotes.com/posts/2026/08/stanford-cs329a-self-improving-ai-agents/ Akanksha Chowdhery and Azalia Mirhoseini trace the arc from pre-training scaling laws through inference-time scaling to agentic workflows, and explain why verification is the bottleneck for self-improvement. https://sparsenotes.com/posts/2026/08/dynamo-amazon-key-value-store/ Sat, 01 Aug 2026 00:00:00 +0000 https://sparsenotes.com/posts/2026/08/dynamo-amazon-key-value-store/ A Papers We Love Tokyo walkthrough of the 2007 Dynamo paper — consistent hashing, virtual nodes, sloppy quorums, vector clocks, Merkle trees — and how much modern DynamoDB quietly abandoned. https://sparsenotes.com/posts/2026/08/scaling-to-long-horizons/ Sat, 01 Aug 2026 00:00:00 +0000 https://sparsenotes.com/posts/2026/08/scaling-to-long-horizons/ Ross Taylor and Chengxi Taylor of General Reasoning trace the line from Galactica's failure to the long-horizon RL frontier — and why compaction, value models, and open-ended environments are the next bottleneck. https://sparsenotes.com/posts/2026/08/misaligned-incentives-ai-coding-agents/ Sat, 01 Aug 2026 00:00:00 +0000 https://sparsenotes.com/posts/2026/08/misaligned-incentives-ai-coding-agents/ Cognition president Russell Kaplan on why 'best model' is now the wrong question, how Frontier Code measures mergeability instead of correctness, and why Cognition underwrote a $10M productivity guarantee.