Research knowledge base

Status Stable · Last verified 2026-07-11 · API v10

The research section is the atlas’s “lab notebook”: how we know what we know, what we’re watching, and why the architecture is what it is.

What’s here

• Changelog watch — curated Discord-side changes we track, with dates and verification status.
• Decision log — numbered architecture decisions (name, single-source content, verifiability, stack, i18n, examples, secrets).
• Research backlog — prioritized next work and honestly unresolved questions.

Source inventory & capability data (in the repo)

• research/sources.md — the authoritative sources we cite, ranked by trust, with what each is good for.
• research/capability-matrix.yaml — machine-readable capability list (area, surface, status, sources). The human-readable view is the platform map.

Method (how a claim earns a status)

1. Prefer primary sources in this order: Discord developer docs → official Discord repos/SDKs → the official change log → announcements/support → specs & source code → high-quality community research (only to fill gaps, clearly marked).
2. Separate fact from hypothesis. Verified claims cite a source and a date; first-hand behavior is an Observation; anything else is a Hypothesis.
3. Date everything. lastVerified on every page; re-verify numeric limits and availability first, since they drift fastest.
4. Test where practical. The examples are how we check real behavior, not just documented behavior.

What we deliberately exclude

Terms-of-Service-violating techniques (self-bots, user-token automation, private endpoint scraping for production) are documented only as platform boundaries, never implemented. See the ToS boundary.