Skip to main content

Start Here

Pick the closest path and move into the full README from there.

How entries are organized

Every resource on the site is classified along three independent axes:

  • Lifecycle stage — where it fits in the discovery loop. See workflows for the 6 canonical stages (e.g., literature-intelligence, evaluation).
  • Domain — the scientific field it targets. See domains for the 6 canonical domain slugs (5 scientific domains + cross-domain).
  • Resource type — the kind of artefact it is. See resource-types for the 10 canonical types (e.g., paper, framework, benchmark).

Axes are independent and filterable: a resource may carry multiple values per axis when appropriate (lifecycle and type frequently do; domain is usually singular and falls back to cross-domain when no field is primary). Tags are written inline in the form:

AlphaFold — protein structure prediction across the proteome (AlphaFold 2 and 3 weights and code). Tags: lifecycle:experiment-planning,evaluation · domain:genomics-biology · type:model,paper

When an entry carries multiple lifecycle or type slugs, list the primary slug first — the one that matches the section the entry sits in. The same entry can appear in two sections with the order reversed (see STORM in workflows and scientific communication).

Maintainers: the rationale behind this taxonomy lives in the local specs/2026-04-29-information-architecture/ folder (gitignored).

Pick your path

New readers

New to AI Scientist systems

Get the landscape first, then scan the standout entries.

Open foundational systems

Open surveys & overviews

Open highlights

Builders

Building agent systems

Start with the implementation surfaces, then fill in the workflow gaps.

Open frameworks

Open tool use

Open models

Researchers

Evaluating or extending systems

Start with benchmarks and datasets, then drill into the relevant domain track.

Open benchmarks

Open datasets

Open domain tracks