https://sparsenotes.com/series/weekly-videos-2026-07/ Recent content in weekly-videos-2026-07 on Sparse Notes https://sparsenotes.com/images/og-default.png https://sparsenotes.com/images/og-default.png Hugo -- gohugo.io Sat, 18 Jul 2026 00:00:00 +0000 https://sparsenotes.com/posts/2026/07/dont-ship-skills-without-evals/ Sat, 18 Jul 2026 00:00:00 +0000 https://sparsenotes.com/posts/2026/07/dont-ship-skills-without-evals/ Philipp Schmid (Google DeepMind) on why almost none of the 50k+ published agent skills have evals — and a concrete recipe for building the smallest useful eval harness for yours. https://sparsenotes.com/posts/2026/07/preventing-collapse-of-civilization/ Sat, 18 Jul 2026 00:00:00 +0000 https://sparsenotes.com/posts/2026/07/preventing-collapse-of-civilization/ Jonathan Blow's 2019 DevGAMM talk on why technological knowledge — including software — does not just automatically compound. A meditation on decay, complexity, and what programmers should be doing about it. https://sparsenotes.com/posts/2026/07/recursive-model-improvement-cursor/ Sat, 18 Jul 2026 00:00:00 +0000 https://sparsenotes.com/posts/2026/07/recursive-model-improvement-cursor/ Lee Robinson (Cursor) on Composer 2.5, why reward hacking is the real bottleneck, and how Cursor is closing the loop so their models help train the next Cursor models. https://sparsenotes.com/posts/2026/07/hammock-driven-development/ Sat, 11 Jul 2026 00:00:00 +0000 https://sparsenotes.com/posts/2026/07/hammock-driven-development/ Classic-of-the-week: Rich Hickey's 2010 talk on why hard problems are solved away from the keyboard, and what a disciplined thinking practice actually looks like. https://sparsenotes.com/posts/2026/07/understanding-is-the-new-bottleneck/ Sat, 11 Jul 2026 00:00:00 +0000 https://sparsenotes.com/posts/2026/07/understanding-is-the-new-bottleneck/ Geoffrey Litt (Notion) on why understanding — not writing — has become the scarce resource in an era where agents land 50,000-line PRs, and three concrete practices to keep humans in the loop. https://sparsenotes.com/posts/2026/07/inner-thoughts-of-ai-interpretability/ Sat, 11 Jul 2026 00:00:00 +0000 https://sparsenotes.com/posts/2026/07/inner-thoughts-of-ai-interpretability/ A long-form Google DeepMind podcast conversation on mechanistic interpretability: what sparse autoencoders actually see inside a model, why chain-of-thought is only sometimes faithful, and how auditing games are becoming the empirical loop for alignment.