Learn anythingin the right order
Too much to read. Too little signal. Ology turns a topic into a seminar: one essay, five questions, and a few sessions of readings that earn their place.
10,003readers so far
Browse the Reading ListsReal lists, real readings
- When the Judge Becomes the BenchmarkFundamentals of using large language models as judges to evaluate AI-generated outputs, including rubrics, prompting, comparison, calibration, reliability, and bias5 sessionsLikes: 2Shares: 3Bookmarks: 1
- Code Is Not a File: Reading Graphs of SoftwareCode graphs model source code through syntax, control flow, data flow, dependencies, and semantics for analysis, navigation, security, and tooling5 sessionsLikes: 1Shares: 2Bookmarks: 1
- Memory Is a Context-Management ProblemAI agent memory and context management: architectures, retrieval, context windows, personalization, evaluation, implementation trade-offs, reliability, privacy, and open questions.5 sessionsLikes: 1Shares: 6Bookmarks: 1
How it works
- Name a topic
A curator types one line: “memory in AI agents”, “how consensus works”. Ology asks what you already know and how deep to go, then sharpens it into a real question.
how do agents remember things across sessions?how deep? and have you built one? - Get the seminar
An opening essay, five questions worth arguing about, and three to five sessions. Each session states a problem and links the three readings that answer it. Ology searches the open web and checks every link.
ContextMemoryEvals - Read in order
Every session sets up the next. The essay and session one are open to anyone with the link. Sign in once with Google for the rest.
1 · Why memory is a design problem2 · What to keep, what to drop3 · Retrieval is not remembering4 · Testing an agent’s memory - Follow it, share it
Follow a list and it waits in your library. Share the link and it lands on the list itself, never on a sign-up page.
ologyai.com/c/…5 sessions · 16 readings
FAQ
Who makes the lists?
Curators. A curator names a topic, Ology sharpens it and searches the web for readings, and the curator reviews and publishes. Every list has a person behind it.
Where do the readings come from?
Primary sources: papers, engineering blogs, documentation, essays. Every reading links out to the original. Nothing is copied or rehosted.
What can I read without signing in?
The opening essay, the first two questions, and the whole first session, for anyone with the link. Sign in with Google once to unlock every session and every reading.
How long does a list take?
A session is an evening: a problem statement and three readings. Most lists are three to five sessions, so a list is a week or two at one sitting per session.
Is it free?
Reading is free. You sign in with Google. No password, no credit card.