<?xml version="1.0" encoding="utf-8"?>
<feed xmlns="http://www.w3.org/2005/Atom">
  <id>https://greglum.com/learn</id>
  <title>Learn AI · Greg Lum</title>
  <subtitle>Understand AI through everyday questions, examples, and sources. Start with Making Sense of AI, or follow a path for using, building with, or reviewing AI.</subtitle>
  <link rel="self" type="application/atom+xml" href="https://greglum.com/learn/feed.xml"/>
  <link rel="alternate" type="text/html" href="https://greglum.com/learn"/>
  <updated>2026-10-29T00:00:00Z</updated>
  <author><name>Greg Lum</name><uri>https://greglum.com/</uri></author>
  <entry>
    <id>https://greglum.com/learn/finding-before-answering</id>
    <title>Finding Before Answering</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/finding-before-answering"/>
    <published>2026-10-28T00:00:00Z</published>
    <updated>2026-10-29T00:00:00Z</updated>
    <summary>Follow a holiday returns question from search to reply, see why a better choice cannot recover a missing policy, and check each stage.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/bounded-decisions</id>
    <title>Should a Model Make This Decision?</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/bounded-decisions"/>
    <published>2026-10-25T00:00:00Z</published>
    <updated>2026-10-29T00:00:00Z</updated>
    <summary>Compare code, classifiers, and models for a fixed choice, set an error limit, and test on fresh cases.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/cheapest-model-that-passes</id>
    <title>Choosing the Cheapest Model That Passes</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/cheapest-model-that-passes"/>
    <published>2026-10-29T00:00:00Z</published>
    <updated>2026-10-29T00:00:00Z</updated>
    <summary>Write the bar first, test the smallest model on frozen cases, and move up one size only when a check fails.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/designing-ai-systems</id>
    <title>Designing the System Around a Model</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/designing-ai-systems"/>
    <published>2026-10-26T00:00:00Z</published>
    <updated>2026-10-29T00:00:00Z</updated>
    <summary>Choose the simplest AI design that does the job, then decide its context, limits, recovery, and evidence, with Ask the Archive as the example.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/reading-model-claims</id>
    <title>How to Read a Model Claim</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/reading-model-claims"/>
    <published>2026-10-28T00:00:00Z</published>
    <updated>2026-10-29T00:00:00Z</updated>
    <summary>Ask 7 questions about a model comparison: its baseline, version, sample size, total cost, and who ran it.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/ai-and-work</id>
    <title>Will AI take my job, or change it?</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/ai-and-work"/>
    <published>2026-10-28T00:00:00Z</published>
    <updated>2026-10-29T00:00:00Z</updated>
    <summary>Find out what studies measure about AI and work, then test which tasks an assistant could help you finish and check.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/how-ai-answers-work</id>
    <title>How does an AI answer get made?</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/how-ai-answers-work"/>
    <published>2026-10-27T00:00:00Z</published>
    <updated>2026-10-28T00:00:00Z</updated>
    <summary>Follow an event notice into an AI answer, and check what context, generation, and search each contribute.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/getting-help-from-ai</id>
    <title>Getting Useful Help from AI</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/getting-help-from-ai"/>
    <published>2026-10-27T00:00:00Z</published>
    <updated>2026-10-28T00:00:00Z</updated>
    <summary>Turn a vague request into a task an assistant can work on, then check and improve its reply.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/ai-privacy-and-actions</id>
    <title>What should I share, and what can it do?</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/ai-privacy-and-actions"/>
    <published>2026-10-27T00:00:00Z</published>
    <updated>2026-10-28T00:00:00Z</updated>
    <summary>Check saved chats, training choices, and connected tools, and see why drafting and sending need different permissions.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/understanding-ai-news</id>
    <title>Making Sense of What You Hear About AI</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/understanding-ai-news"/>
    <published>2026-10-27T00:00:00Z</published>
    <updated>2026-10-28T00:00:00Z</updated>
    <summary>Turn an AI headline into a task you can assess, and separate results from predictions.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/what-agents-do</id>
    <title>What does an AI agent actually do?</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/what-agents-do"/>
    <published>2026-10-28T00:00:00Z</published>
    <updated>2026-10-28T00:00:00Z</updated>
    <summary>Follow a book-swap task through tools, choices, and approvals, then decide when a fixed workflow would do.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/writing-versus-choosing</id>
    <title>Writing an Answer or Choosing One</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/writing-versus-choosing"/>
    <published>2026-10-28T00:00:00Z</published>
    <updated>2026-10-28T00:00:00Z</updated>
    <summary>Follow a bike shop request through writing, finding, ranking, choosing, and the jobs for ordinary software.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/ai-costs-and-tokens</id>
    <title>What makes an AI request cost more?</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/ai-costs-and-tokens"/>
    <published>2026-10-28T00:00:00Z</published>
    <updated>2026-10-28T00:00:00Z</updated>
    <summary>Work out the token bill for a book-swap question box, then compare caching, batching, and sending less.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/where-models-run</id>
    <title>Where should an AI model run?</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/where-models-run"/>
    <published>2026-10-28T00:00:00Z</published>
    <updated>2026-10-28T00:00:00Z</updated>
    <summary>Compare who sees your text, what you pay, and who maintains the model when you use a hosted service, a cloud platform, or your own machine.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/ai-rules-and-dates</id>
    <title>Which AI rules apply, and when?</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/ai-rules-and-dates"/>
    <published>2026-10-28T00:00:00Z</published>
    <updated>2026-10-28T00:00:00Z</updated>
    <summary>Read an AI rule by place, role, use, and date, with examples from the EU, the US, and South Korea.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/tricking-ai-systems</id>
    <title>How can someone trick an AI assistant?</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/tricking-ai-systems"/>
    <published>2026-10-28T00:00:00Z</published>
    <updated>2026-10-28T00:00:00Z</updated>
    <summary>Follow hidden instructions in a book-swap email and see which permissions can limit the harm.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/ai-across-industries</id>
    <title>How is AI used in different fields?</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/ai-across-industries"/>
    <published>2026-10-28T00:00:00Z</published>
    <updated>2026-10-28T00:00:00Z</updated>
    <summary>Follow documented AI uses across 9 fields: what goes in, who checks the result, and what a mistake costs.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/ai-energy-and-compute</id>
    <title>How much energy does AI use?</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/ai-energy-and-compute"/>
    <published>2026-10-28T00:00:00Z</published>
    <updated>2026-10-28T00:00:00Z</updated>
    <summary>Check what electricity and water estimates count, from a single AI question to training a model and running data centers.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/cited-answers</id>
    <title>How can I check an AI answer?</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/cited-answers"/>
    <published>2026-10-23T00:00:00Z</published>
    <updated>2026-10-27T00:00:00Z</updated>
    <summary>Check the facts, open the sources, and notice what is missing. A confident answer or a citation can still be wrong.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/what-is-ai</id>
    <title>What do people mean by AI?</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/what-is-ai"/>
    <published>2026-10-27T00:00:00Z</published>
    <updated>2026-10-27T00:00:00Z</updated>
    <summary>See how predicting, recognizing, recommending, and creating content fit under the name AI.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/why-ai-now</id>
    <title>Why is everyone talking about AI now?</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/why-ai-now"/>
    <published>2026-10-27T00:00:00Z</published>
    <updated>2026-10-27T00:00:00Z</updated>
    <summary>Follow the changes that brought AI from learning patterns to tools you can talk to.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/ai-names-explained</id>
    <title>ChatGPT, Claude, and All Those Names</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/ai-names-explained"/>
    <published>2026-10-27T00:00:00Z</published>
    <updated>2026-10-27T00:00:00Z</updated>
    <summary>Untangle the company, model, assistant, and tool behind familiar AI names, with no technical background needed.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/one-rulebook-two-agents</id>
    <title>One Rulebook for Two Agents</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/one-rulebook-two-agents"/>
    <published>2026-10-23T00:00:00Z</published>
    <updated>2026-10-23T00:00:00Z</updated>
    <summary>Give Claude Code and Codex the same user-level rules through an import and a symbolic link.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/public-repo-to-better-skill</id>
    <title>From a Public Repo to a Better Skill</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/public-repo-to-better-skill"/>
    <published>2026-10-23T00:00:00Z</published>
    <updated>2026-10-23T00:00:00Z</updated>
    <summary>Find advice your skill needs, check its source and license, and test the change against a known failure.</summary>
  </entry>
  <entry>
    <id>https://greglum.com/learn/catching-what-agents-miss</id>
    <title>Catching What Agents Miss</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/catching-what-agents-miss"/>
    <published>2026-10-23T00:00:00Z</published>
    <updated>2026-10-23T00:00:00Z</updated>
    <summary>See how fresh review, browser measurements, and accessibility tests caught 4 defects in a Draftroom change that passed local checks.</summary>
  </entry>
  <entry>
    <id>tag:greglum.com,2026-10-23:learn-log/2</id>
    <title>Prepared my skills repository for public release</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/skills"/>
    <published>2026-10-23T00:00:00Z</published>
    <updated>2026-10-23T00:00:00Z</updated>
    <summary>A cleaned export of the catalog, with an installer dry run, a validator, and license notices.</summary>
  </entry>
  <entry>
    <id>tag:greglum.com,2026-10-23:learn-log/1</id>
    <title>Studied public motion skills for the film skill</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/public-repo-to-better-skill"/>
    <published>2026-10-23T00:00:00Z</published>
    <updated>2026-10-23T00:00:00Z</updated>
    <summary>Adopted timing defaults, script checks, and a critique loop into motion-film, with each source and license recorded.</summary>
  </entry>
  <entry>
    <id>tag:greglum.com,2026-10-22:learn-log/2</id>
    <title>Cleaned up my development environment</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/how-i-build#tried"/>
    <published>2026-10-22T00:00:00Z</published>
    <updated>2026-10-22T00:00:00Z</updated>
    <summary>Retired Zed, Neovim, Warp, Podman, Caveman, an offline semantic search tool, and a local docs index. Context7 replaced the docs index.</summary>
  </entry>
  <entry>
    <id>tag:greglum.com,2026-10-22:learn-log/1</id>
    <title>Wrote down how I split work between agents</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/log"/>
    <published>2026-10-22T00:00:00Z</published>
    <updated>2026-10-22T00:00:00Z</updated>
    <summary>Claude Code mainly implements. Codex mainly refines writing, researches, and configures the machine. It is a habit, not a rule.</summary>
  </entry>
  <entry>
    <id>tag:greglum.com,2026-10-20:learn-log/1</id>
    <title>Four problems caught before one release</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/catching-what-agents-miss"/>
    <published>2026-10-20T00:00:00Z</published>
    <updated>2026-10-20T00:00:00Z</updated>
    <summary>An agent-built Draftroom change passed the local gate. Reviewers, a phone screenshot, and CI still caught four problems before merge.</summary>
  </entry>
  <entry>
    <id>tag:greglum.com,2026-10-18:learn-log/1</id>
    <title>Added How I Build to this site</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/how-i-build"/>
    <published>2026-10-18T00:00:00Z</published>
    <updated>2026-10-18T00:00:00Z</updated>
    <summary>My Claude Code and Codex setup, shared rules, skills, and checks, on one page.</summary>
  </entry>
  <entry>
    <id>tag:greglum.com,2026-10-16:learn-log/1</id>
    <title>Gave the site a written point of view</title>
    <link rel="alternate" type="text/html" href="https://greglum.com/learn/log"/>
    <published>2026-10-16T00:00:00Z</published>
    <updated>2026-10-16T00:00:00Z</updated>
    <summary>After Katie Dill's talk on &quot;zombie UI&quot;, the design spec gained a point of view and principles that agents follow.</summary>
  </entry>
</feed>
