NEXT.JS ENGINE / START HERE

Learn the decisions before the features.

Generation 3 is easier to understand when you start with one question: what work belongs on the server, what genuinely needs the browser, and how does the compiler make that boundary visible? Learn that first; the individual Engine* systems make much more sense afterward.

CURRENT TRACK · 3.0.0-alpha.0 · main-3 · active development

FIRST PASS

A useful order for your first twenty minutes.

01 / WHY

Understand the contract.

Learn why Gen 3 compiles intent early, keeps durable content server-first, exposes runtime cost and treats fallbacks as explicit behavior.

mental model · ~4 min

Open →

02 / COMPILE

Follow schema into an execution plan.

Read the compiler/server-first overview before memorizing component APIs. The important output is the plan: runtime class, capabilities, assets and boundaries.

compiler · server renderer · scheduler · adaptive

Open →

03 / INSPECT

See why the page behaves that way.

EngineDebug turns compiler nodes, runtime boundaries, scheduler state and device simulation into a development workbench instead of hidden framework behavior.

D.5–D.10 · dev-only workbench

Open →

04 / AUTHOR

Use the server-first content systems.

EngineMarkdown and EngineSEO are good examples of the Gen 3 rule: push deterministic work into server output and keep browser ownership deliberate.

Markdown · SEO · semantic output

Open →

05 / NETWORK

Add private runtime work last.

Once the execution model makes sense, move into NENC, EngineCookies, device proof and account/session flows. Those systems assume you understand client/server authority.

NENC · sessions · private commands

Open →

06 / SOURCE

Then explore the full tree.

Use Downloads for main-3 source/package choices and the live docs workspace for the complete reference. Alpha status means current docs matter more than old examples.

main-3 source of truth

Open →

THE SHORT VERSION

Schema is not the destination. The execution plan is.

If you remember only one thing, remember the boundary. Gen 3 is trying to make browser ownership, scheduling and compatibility a consequence of compiled intent rather than a pile of per-component guesses.

author intent ↓ typed schema ↓ compilePage() ↓ execution plan ├─ server/static output ├─ targeted client work ├─ scheduler priority ├─ assets + capabilities ├─ fallback plan └─ debug metadata

DON'T START TOO DEEP

Advanced systems are easier after the boundary model clicks.

NENC before client/server authority

Private command transport makes much more sense once you understand what the compiler leaves in the browser and what remains server-owned.

Compatibility before used-feature manifests

Fallback planning is intentionally based on what a page actually uses. Read the compiler/feature-manifest story before treating compatibility as a browser checklist.

EngineDebug as a production feature

The debugger exists to explain development decisions. Its production guarantee is absence, not a hidden admin screen.

Old Gen 2 examples as the current contract

Generation 3 is an alpha line with a live main-3 source of truth. Historical release docs are useful history, not the current architecture contract.