Carpet reached 26.3 on 18 September. The release changelog is one line – support for Minecraft 26.3 – which is exactly what you want from a mod whose entire job is to sit underneath everything else and not break.
What Carpet is for
Carpet is the technical community’s toolbox. It adds a large set of toggleable rules that change or expose game behaviour: TNT duplication switches, hopper counters, fake players, tick warp, chunk-loading visualisations, movement debug overlays. None of it is content. All of it exists so that someone building a farm can measure what the farm is doing.
The important design choice is that every rule is off by default. A vanilla server with Carpet installed behaves like a vanilla server until an operator turns something on, which is why it shows up on technical SMPs that otherwise refuse to install mods.
The scarpet change worth reading
The 26.3 release itself only mentions version support, but the beta line leading into it carried a scarpet change: the options vein_toggle, vein_ridged and vein_gap were removed from sample_noise(...).
Scarpet is Carpet’s embedded scripting language, and sample_noise is how a script asks the world generator what a noise field looks like at a coordinate. Those three options referred to the ore vein noises in the terrain generator. Their removal means the underlying generator no longer exposes them, which is a worldgen change surfacing in a scripting API rather than a Carpet decision.
If you have scarpet scripts that probe vein noise – most likely something that maps ore distribution – they will need updating. Everyone else will never notice. For the non-scripting version of the same question, the ore distribution by Y level breakdown is the practical reference.
Why technical mods port late
Carpet landed on 26.3 three days after the game version did, which is fast by technical-mod standards. Mods that hook deep into tick order and chunk management usually wait out the first round of hotfixes, because a mixin into a method that gets rewritten in the next patch is wasted work.
The reason Carpet can move quickly is that its rules are mostly independent. If one rule’s hook breaks, that rule is disabled, not the whole mod. Compare that to a rendering mod where everything runs through one pipeline and one broken assumption takes the frame with it.
Carpet on a survival server
Carpet’s reputation is as a creative-mode debugging tool, but plenty of survival servers run it for two or three rules and nothing else. The common ones:
- Hopper counters – turn a wool block into a throughput meter so farm output can be measured in items per hour rather than guessed.
- Fake players – a bot that holds a position for AFK-dependent farms, used carefully and usually restricted.
- Chunk-loading info – visibility into what is actually ticking, which is the first thing you want when the server starts lagging.
That last one overlaps with what a profiler gives you, and the two answer different questions. Carpet tells you what the game thinks is loaded; a profiler tells you where the time went. Running a profiler alongside it is the standard combination.
Rules are policy, not features
The thing server owners underestimate is that Carpet rules are a balance decision disguised as a config file. Turning on TNT duplication changes what a week of play looks like. Turning on fake players changes whether AFK farms are a strategy or an exploit. Neither is wrong, but both are the kind of thing that should be written down somewhere players can read it.
A server that enables three Carpet rules quietly and then argues about them in chat two months later has made a moderation problem out of a config file. This belongs in the same document as your other early server ground rules.
Installing
Carpet is Fabric-only and ships as a single jar for 26.3. It needs no other dependency. Rules are set with /carpet in-game by an operator, or in carpet.conf in the world folder for settings you want to survive restarts.
The default rule set is empty, so installing it and doing nothing is a valid state. That is worth doing before a version migration: having the diagnostic tooling already present when something goes wrong is much easier than adding a mod to a broken server.
What comes next
Carpet’s extensions – the various add-on mods that bundle extra rules – generally follow the base mod by a week or two. If you depend on one of those rather than on core Carpet, check its version tag before you migrate the server, because the base mod being on 26.3 does not mean your extension is.
For 26.4, the same pattern will repeat. The snapshot cycle has already started, and technical mods will wait for the release rather than chase snapshots, which is the correct trade for anyone whose world is measured in months rather than days.
Images: Bliss Shaders project gallery.