<p><strong>Intro topic: Asymmetric Returns<br></strong><br></p><p><br></p><p><strong>News/Links:<br></strong><br></p><ul><li>NanoChat by Andrej Karpathy<ul><li><a href="https://github.com/karpathy/nanochat">https://github.com/karpathy/nanochat</a></li></ul></li><li>Pydantic AI<ul><li><a href="https://www.marktechpost.com/2025/03/25/pydanticai-advancing-generative-ai-agent-development-through-intelligent-framework-design/">https://www.marktechpost.com/2025/03/25/pydanticai-advancing-generative-ai-agent-development-through-intelligent-framework-design/</a></li></ul></li><li>1000th Starlink this year<ul><li><a href="https://spaceflightnow.com/2025/05/16/live-coverage-spacex-plans-morning-launch-of-starlink-satellites-from-california/">https://spaceflightnow.com/2025/05/16/live-coverage-spacex-plans-morning-launch-of-starlink-satellites-from-california/</a></li></ul></li><li>ChatGPT Apps SDK<ul><li><a href="https://openai.com/index/introducing-apps-in-chatgpt/">https://openai.com/index/introducing-apps-in-chatgpt/<br></a><br></li></ul></li></ul><p><strong>Book of the Show</strong></p><ul><li>Patrick<ul><li>The Will of the Many by James Islington<ul><li><a href="https://amzn.to/43IfU8Q">https://amzn.to/43IfU8Q</a></li></ul></li></ul></li><li>Jason<ul><li>Interview with DHH (Founder of Ruby on Rails)<ul><li><a href="https://www.youtube.com/watch?v=vagyIcmIGOQ">https://www.youtube.com/watch?v=vagyIcmIGOQ<br></a><br></li></ul></li></ul></li></ul><p><br></p><p>Patreon Plug <a href="https://www.patreon.com/programmingthrowdown?ty=h">https://www.patreon.com/programmingthrowdown?ty=h</a></p><p><br></p><p><strong>Tool of the Show</strong></p><ul><li>Patrick<ul><li>Factorio<ul><li><a href="https://www.factorio.com/">https://www.factorio.com/</a> </li></ul></li></ul></li><li>Jason<ul><li><a href="http://nip.io">nip.io</a> <p></p></li></ul></li></ul><p><br></p><p><strong>Topic: Workflow Orchestrators<br></strong><br></p><ul><li>Why<ul><li>Batch jobs (embarrassingly parallel)</li><li>Long-running tasks (e.g. transcoding video)</li><li>Checkpointing/resuming</li></ul></li><li>How<ul><li>Message Queues</li><li>Containerization</li><li>Worker Pools &amp; Autoscaling</li><li>History &amp; Backfill</li></ul></li><li>Steps to run workflows:<ul><li>Containerize the workflow definition and send to the cloud</li><li>Containerize all the individual tasks</li><li>Submit job(s)</li></ul></li><li>Examples<ul><li>Airflow<ul><li>Legacy but dominant</li></ul></li><li>Dagster<ul><li>Great UX for python developers</li></ul></li><li>Temporal: <a href="https://temporal.io/">https://temporal.io/</a><ul><li>The new hotness</li></ul></li><li>Ray<ul><li>Low-level but very powerful</li></ul></li><li>Kubeflow<ul><li>Designed for ML workflows, integrated dashboard<p></p></li></ul></li></ul></li></ul>
<strong>
  <a href="https://www.patreon.com/programmingthrowdown" rel="payment" title="★ Support this podcast on Patreon ★">★ Support this podcast on Patreon ★</a>
</strong>

Programming Throwdown

Patrick Wheeler and Jason Gauci

185: Workflow Orchestrators

NOV 4, 202592 MIN
Programming Throwdown

185: Workflow Orchestrators

NOV 4, 202592 MIN

Description

Intro topic: Asymmetric ReturnsNews/Links:NanoChat by Andrej Karpathyhttps://github.com/karpathy/nanochatPydantic AIhttps://www.marktechpost.com/2025/03/25/pydanticai-advancing-generative-ai-agent-development-through-intelligent-framework-design/1000th Starlink this yearhttps://spaceflightnow.com/2025/05/16/live-coverage-spacex-plans-morning-launch-of-starlink-satellites-from-california/ChatGPT Apps SDKhttps://openai.com/index/introducing-apps-in-chatgpt/Book of the ShowPatrickThe Will of the Many by James Islingtonhttps://amzn.to/43IfU8QJasonInterview with DHH (Founder of Ruby on Rails)https://www.youtube.com/watch?v=vagyIcmIGOQPatreon Plug https://www.patreon.com/programmingthrowdown?ty=hTool of the ShowPatrickFactoriohttps://www.factorio.com/ Jasonnip.io Topic: Workflow OrchestratorsWhyBatch jobs (embarrassingly parallel)Long-running tasks (e.g. transcoding video)Checkpointing/resumingHowMessage QueuesContainerizationWorker Pools & AutoscalingHistory & BackfillSteps to run workflows:Containerize the workflow definition and send to the cloudContainerize all the individual tasksSubmit job(s)ExamplesAirflowLegacy but dominantDagsterGreat UX for python developersTemporal: https://temporal.io/The new hotnessRayLow-level but very powerfulKubeflowDesigned for ML workflows, integrated dashboard ★ Support this podcast on Patreon ★