seangoedecke.com

Latest posts

• Build agents, not pipelines

  May 31, 2026
• The famous o3 "GeoGuessr" prompt did not work

  May 21, 2026
• Prompts are technical debt too

  May 20, 2026
• The just-say-no engineer was a ZIRP phenomenon

  May 18, 2026
• How I use LLMs as a staff engineer in 2026

  May 17, 2026
• DeepSeek-V4-Flash means LLM steering is interesting again

  May 16, 2026
• AI datacenters in space do not have a cooling problem

  May 13, 2026
• Thinking Machines and interaction models

  May 12, 2026
• The left-wing case for AI

  May 10, 2026
• AI makes weak engineers less harmful

  May 09, 2026
• Notes on incidents

  May 08, 2026
• Why hasn't longer-horizon training slowed AI progress?

  May 07, 2026
• Why I don't like the "staff engineer archetypes"

  May 03, 2026
• Software engineering may no longer be a lifetime career

  Apr 24, 2026
• Luddites and burning down AI datacenters

  Apr 22, 2026
• Many anti-AI arguments are conservative arguments

  Apr 18, 2026
• Programming (with AI agents) as theory building

  Apr 03, 2026
• Working on products people hate

  Mar 27, 2026
• Engineers do get promoted for writing simple code

  Mar 26, 2026
• Big tech engineers need big egos

  Mar 14, 2026
• I don't know if my job will still exist in ten years

  Mar 06, 2026
• Giving LLMs a personality is just good engineering

  Mar 03, 2026
• What's so hard about continuous learning?

  Feb 23, 2026
• Insider amnesia

  Feb 23, 2026
• LLM-generated skills work, if you generate them afterwards

  Feb 17, 2026
• Two different tricks for fast LLM inference

  Feb 15, 2026
• On screwing up

  Feb 11, 2026
• Large tech companies don't need heroes

  Feb 08, 2026
• Getting the main thing right

  Feb 05, 2026
• How does AI impact skill formation?

  Jan 31, 2026