MocaccinoOS: the successor, or the estate?

Every write-up of Sabayon Linux ends the same way: the project rebranded to MocaccinoOS, which is still going. Nobody says what that sentence means. A distribution is four separable things — the people, the code, the repository and the idea — and a successor can inherit any subset of them. So here is each one, scored, from the build definitions and the release notes rather than from the claim.

MocaccinoOS MocaccinoOS · shipping · verified 2026-08-04 shipped version 26.08 on 1 August 2026. Sabayon Sabayon · ended · verified 2026-08-04 last shipped 19.03 in March 2019. Between those two dates the name changed, the package manager was replaced, a partnership with a third distribution was announced and abandoned, and one of the developers left to build something else. The shutdown timeline itself belongs to a separate piece; this one is about what came out the other side.

I have not run MocaccinoOS. Nothing below is a report of using it. Everything below is a statement about published artifacts — package definitions, repository indexes, release notes, git history — each of which you can fetch yourself, and most of which I have quoted as records so you can.

First, the base, because every other answer depends on it

This is the question that has been sitting unanswered on this site since launch, and it is worth being blunt about why it was hard. On 20 November 2020 the Sabayon project announced that it was rebranding to MocaccinoOS, replacing Entropy with a new Go package manager called Luet, and moving its base from Gentoo to Funtoo. In 2026, the project's own documentation — first line of the desktop reference, "a Gentoo-based distribution (derived from Sabayon)" — and every piece of press coverage describe MocaccinoOS as Gentoo-based. Both statements cannot describe the same system, and "it moved to Funtoo and then moved back" is the kind of tidy inference that turns out to be wrong.

The build definitions settle it, and they say something more interesting than either version.

MocaccinoOS Desktop is assembled by Luet from package definitions in one repository. The seed of the whole thing is a package called stage3, and its definition file names its category outright:

That is a Gentoo release-engineering autobuild, pulled from Gentoo's own distfiles host, dated 2 August 2026 — the day after 26.08 shipped. The ebuild tree layered on top of it is equally unambiguous:

A pinned commit of gentoo/gentoo, bumped automatically, dated 3 August 2026. There is no Funtoo anywhere in the chain. The answer to the Register's open cell is Gentoo, and it is not a judgement call — it is a URL in a build file.

The reason the 2020 announcement is not simply wrong is that the Funtoo move was real, and it was abandoned in public. The MocaccinoOS team posted about it on 6 December 2021, in a note titled "Updates on Funtoo collaboration". Their account, in summary: a technical obstacle appeared early; one developer was assigned to it full-time as an R&D project while the rest of the team built a Luet repository that, in their own words, still originated on the Gentoo Portage tree; the two projects drifted apart over months; and when the developer leading the Funtoo side left to work on his own fork inside Funtoo, they let the collaboration go. Nine months later, the 1.0 release note thanked Gentoo as upstream and wished that departing developer luck building a Funtoo-based distribution.

The git record matches the note. The Funtoo tree repository, mocaccinoOS/mocaccino-funtoo, took its last commit on 22 November 2021 and is archived. The Gentoo overlay repository, mocaccinoOS/portage-tree, went quiet on 12 October 2021 and is archived too — superseded, not abandoned, by the layer model in desktop, which is still being pushed to daily.

That still leaves the question the note does not answer: the two alpha images that came out while the Funtoo branch was alive. This is checkable, because the build definitions are versioned, so you can read them as they stood on the day an image was tagged rather than as they stand now. The first alpha was tagged at 13:41:45 UTC on 9 October 2021. The tip of mocaccinoOS/desktop two hours earlier was commit ef92d369a1, and it looked like this:

Both alphas were Gentoo. And the seed was Gentoo well before either of them: the commit that made desktop consume a stage3 directly, on 2 June 2021, already named category: "gentoo" and already fetched from distfiles.gentoo.org, and the portage layer it sat under was already pinned to github.owner: gentoo. So for the whole life of the Funtoo branch, the thing that produced images was pointed at Gentoo. The Funtoo repository was real and was being committed to in parallel; it was never what the desktop was built from.

So the accurate sentence is not "the base moved to Funtoo and back". It is: the base was announced as moving to Funtoo, the move was pursued in a separate repository by one developer, every image the project has tagged — including the two alphas released while that repository was live — was built from a Gentoo stage3 and a pinned gentoo/gentoo tree, and the collaboration was formally ended twelve months and sixteen days after it was announced. The 2026 coverage calling MocaccinoOS Gentoo-based is correct. The 2020 announcement is accurate as 2020 news and misleading as a description of anything shipping today, which is exactly the failure mode of reading a rebrand announcement as a permanent fact.

Score one: the people, and this is where succession is strongest

Every MocaccinoOS release note from 1.0 in September 2022 to 26.08 in August 2026 is signed by the same four names: Ettore Di Giacinto, Joost Ruis, Mihai Dobrescu and Jerrod Frost. That is twenty-three consecutive releases over three years and eleven months with an unchanged signature block — I checked every one — and it is not nothing; plenty of larger projects cannot say it. Nothing before 1.0 is signed at all: the nine alpha and beta notes carry no names.

Two of those four came out of Sabayon. Di Giacinto — mudler — describes himself on his own site as formerly a Gentoo developer and Sabayon lead, and wrote Luet's first commit. Joost Ruis's public bio describes him as MocaccinoOS repository maintainer and formerly Sabayon Linux repository maintainer. The lead and the repository maintainer both crossed over. On a project whose entire output was a binary repository, those are the two jobs that matter most.

One name is missing, and it is the founder's. Fabio Erculiani — lxnay, who started the thing on this domain in 2005 — is not on the MocaccinoOS team. His current project is matrixOS matrixOS · shipping · verified 2026-08-04, announced in February 2026, which is immutable, OSTree-based, and carries his own warning that it is a hobby project for homelabs rather than production machines. A third developer, Daniele Rondina — geaaru — is the one who left for Funtoo in 2021; the 1.0 note calls him "our former developer" by name. He now builds Macaroni OS, which its own site describes as created by an "ex Sabayon Linux developer", incubated under the Funtoo umbrella for a year before getting its own domain and infrastructure, and which since August 2024 has been replacing its dependency on the low-level Funtoo stack. Its package manager is Anise, whose README states outright that it is "originally based on luet-0.17.0" — the split took a copy of the tooling with it, and both halves were still taking commits in the first days of August 2026.

Sabayon's people did not move to MocaccinoOS. They split three ways, and MocaccinoOS got the largest share.

Score two: the code, and this is where succession is thinnest

Entropy — the binary package manager Sabayon spent a decade building, and the reason anyone argued about Sabayon at all — is gone. Its GitHub repository took its last push on 7 September 2020, ten weeks before the rebrand was announced. Nothing in MocaccinoOS descends from it.

Luet is not a fork of anything Sabayon wrote, and it is not a post-rebrand project either. Its GitHub repository was created on 5 June 2019, but the first commit in its history is dated 21 September 2018 — root, by Ettore Di Giacinto from a @gentoo.org address, six months before Sabayon shipped its last release. The replacement for Entropy was being written while Entropy was still the product. It describes itself as a zero-dependency container-based package manager using a SAT solver and Q-learning; it is written in Go, where Entropy was Python. Its most recent tagged release is 0.36.5, on 10 February 2026, and it was still taking commits on 22 July 2026.

What did survive, literally, is smaller than a footnote and more interesting than it sounds. Four packages with Sabayon's name on them were still published in MocaccinoOS's live repository index when I checked it:

Read the build files behind those four names and three of them turn out to be thinner than the names suggest. Only one is Sabayon's own code.

sabayon-dracut is the real one. It is a wrapper script plus one configuration file dropped at /etc/dracut.conf.d/99-sabayon.conf, whose first line is a comment reading "Sabayon Dracut Default parameters" and whose body is a list of dracut modules — rescue, resume, btrfs, plymouth, lvm, mdraid, crypt — to force into the initramfs. The script still carries its original header in the desktop repository: an author line reading geaaru@sabayonlinux.org and a copyright line reading Copyright (s) 2018-2020 Sabayon. That is Daniele Rondina — so the one surviving piece of Sabayon-authored code inside its successor was written by the one developer on this page who did not stay. It has been superseded in the same repository by sys-kernel/mocaccino-dracut 1.8.3, which does the same job under the new name.

The other two are branding. Both live in the livecd repository next to packages called lml-init and lml-boot, and the specs are public, so the diff is readable:

So sabayon-live-boot is Ivan Davidov's Minimal Linux Live boot directory with the word "Minimal" replaced by "Sabayon" four times, and sabayon-init is the same upstream with a login banner and a default root password. That is not Sabayon's live-media stack; it is Sabayon's name on somebody else's. The fourth package, luet-migrate-entropy, is the migration shim — a tool for reading Entropy's database, not a piece of it.

That is the inheritance, and it is smaller than the file count implies: one dracut wrapper and one dracut config, both by a developer who has since left, both already replaced under a new name — plus a migration shim and two packages whose Sabayon content is a string. Whatever MocaccinoOS inherited, it was not a codebase.

Score three: the repository, and the answer is no

Sabayon's repositories are gone; a 2019 image cannot update itself, which is the substance of the argument in the piece on Entropy and Gentoo's binhost. MocaccinoOS's repository is a new artifact built from the Gentoo tree described above, and it is much smaller — deliberately.

The public index for mocaccino-desktop-stable listed 349 packages on 2026-08-04; mocaccino-os-commons listed 356. That is not a distribution repository in the Portage sense and is not trying to be. MocaccinoOS groups software into layers: a whole desktop is one installable package.

$ sudo luet repo update
$ sudo luet install layers/plasma
$ sudo luet search --installed
$ sudo luet upgrade
$ sudo luet oscheck

Package names carry their category — apps/neofetch, not neofetch — and the documentation is explicit that this is the point: "All micropackages are abstracted away and are included in a layer and installed as a single package. Think at it like Android apps." Everything outside the index is expected to come from Flatpak, Docker or the community repository, and the release notes say so in the same paragraph every month.

Counting the 349 is worth doing, because it is not 349 pieces of software. I tallied the index by category:

More than half the repository is firmware, split one package per vendor — which is not padding but a deliberate change, announced with the March 2022 beta, to stop shipping the whole of linux-firmware on every ISO. Strip it out and the software you would actually go looking for is the 44 packages under apps, the 19 desktops, and whatever a layer pulls in behind it. "349 packages" and "349 things a user would think of as software" are different sentences, and only the first is true.

There is a front end now. Vajo, installed with sudo luet install apps/vajo, ships as vajo-gui (GTK) and vajo-tui (ncurses, so it works over SSH). Its File menu is a thin wrapper over the commands above: update repositories, full system upgrade, check the system with luet oscheck, clear the cache. Some packages are marked protected so the GUI will not let you remove them. The documentation's own status line, as of 2026-08-04, is that Vajo "is still under development but considered stable" — which is a more honest label than most projects put on a package manager front end.

Whether this is better than Entropy's model is genuinely arguable. Layers make a desktop install atomic and a broken dependency graph impossible, because there is no graph to break. They also mean the thing you want may simply not exist as a package. Entropy exposed thousands of packages and its failure mode was a repository going inconsistent; Luet exposes hundreds, and its failure mode is that you leave the distribution's world for someone else's runtime the first time you need something unusual.

Score four: the idea, and this is what actually carried

Sabayon's proposition was that you should be able to get a Gentoo-shaped system without compiling it, on media that boots, with the awkward parts — codecs, firmware, proprietary graphics drivers — already handled. Strip the branding and MocaccinoOS is making the same proposition with different machinery.

It also says so. The first line of the desktop documentation, on 2026-08-04, is that "Mocaccino Desktop is a Gentoo-based distribution (derived from Sabayon) oriented towards Desktop systems" — the descent is stated by the project itself, in the reference material rather than in an anniversary post, which is where a claim of this kind is cheapest to make and most likely to be true.

The rest of the evidence is in the mundane parts of the changelogs, which is where a project's actual priorities show. 26.06 added Google's CJK font family to every ISO and Blueman — a Bluetooth manager — to the Xfce and MATE images, and gave Calamares an additional-software module. 26.07 added a graphics section to that module and listed Nvidia 595.84 as available in the repository. 25.09 fixed Calamares applying the user's chosen system language, integrated detection code for Intel's Xe and Battlemage graphics, added udev rules for older hard disks, and summarised itself as "fast, minimal, and gaming-ready". layers/codecs is a package. None of that is architecture; all of it is the work of making somebody's first hour uneventful on hardware they already own, which was Sabayon's whole argument for existing.

One difference is worth flagging rather than glossing. Sabayon put the proprietary blob on the disc; MocaccinoOS appears to offer drivers as an installer option and a repository package. Offered at install time is not the same as present on first boot, and I could not find a project page that states plainly which one a fresh MocaccinoOS install gives you. That cell stays open in the Register rather than getting a confident answer here.

There is a documented Sabayon-to-MocaccinoOS migration, and you should read it before you trust it

The project still publishes "Migrating from Sabayon". It states that migrations have been checked for Sabayon 18.05 and newer, which includes 19.03; it warns that this should be done on test systems only; and it recommends at least 30 GB of free disk space, because the process downloads a second complete system alongside the first. The core of it:

$ sudo curl -k https://get.mocaccino.org/luet/get_luet_root.sh | sh
$ sudo luet install repository/mocaccino-os-commons repository/mocaccino-extra repository/mocaccino-desktop-stable
$ sudo luet install system/luet-migrate-entropy
$ sudo luet migrate-entropy

Three observations, in increasing order of how much they should worry you.

First, luet-migrate-entropy still exists — version 0.12.0, published in the luet-official repository, present in the index on 2026-08-04. The path has not been quietly deleted.

Second, the -k in that first command disables TLS certificate verification on a script that is then piped into a root shell. It is also unnecessary: on 2026-08-04 that URL returned HTTP 200 to a plain curl with verification on. Whatever certificate problem prompted the flag has been fixed and the documentation has not caught up. Drop the -k.

Third, and fatally for the advertised path: the documented preparation step is to slim the system down first using Entropy, with commands like sudo equo remove dev-qt/qtcore --ask. Entropy's repositories no longer exist. I have not run this, so I will not tell you it fails — but a procedure whose first stage drives a package manager with no working repository is not a procedure anyone has exercised recently. If you are sitting on a live Sabayon install today, the honest reading is that this document records an intention from 2021, not a supported route in 2026, and the decision tree in the piece on what to do when your distribution dies is a better place to start.

Is it actually alive? Thirty-two releases say yes, and one sentence says be careful

The release record is public and easy to check.

Thirty-two tags between October 2021 and August 2026 — eleven of them in the twelve months to 4 August 2026, on current kernels and a current Mesa, with two months skipped and one month doubled. By any reasonable test, this is a live distribution and not a parked name.

The sentence that should temper that appears in the same release notes, in a section the team calls the project sustainability note:

The same sentence is printed under each figure, word for word: this level of funding "is not sufficient to sustainably support long-term work". That does not mean the distribution is about to disappear — volunteer work does not stop when the donations do. It does mean the thing keeping it alive is four people's evenings, and that publishing the number instead of hiding it is straighter behaviour than most projects manage. Sabayon died of a bus factor of one. MocaccinoOS has a bus factor of about four and a budget of a dollar a month, and it says so on its own release page.

A related signal, offered as observation rather than verdict. The fifty most recent commits to Luet span 19 April 2024 to 22 July 2026, and break down by author like this:

Thirty of the fifty are machine accounts and twenty are people, sixteen of those twenty being the project lead. localai-bot is not an anonymous automation: the lead published a post in February 2026 describing how he built a local autonomous agent team to maintain LocalAI, his main project, and it is the same account. Read charitably that is a maintainer scaling himself; read plainly it is fifty commits in twenty-seven months, which is maintenance velocity, not development velocity. Luet works. It is not where the energy is, and the person whose energy it would be has said in public where his is going instead.

The verdict: it is the successor, and there was never an estate

Score the four axes. People: yes, substantially — the lead and the repository maintainer both came from Sabayon and have signed twenty-three consecutive release notes together. Idea: yes, almost completely — a binary Gentoo desktop that boots and works, which is the entire premise of the 2005 DVD, and the project states the descent itself in its own documentation. Code: no — one dracut wrapper by a developer who has left, already superseded. Repository: no, and not comparably — an index of 349 entries, more than half of it firmware, with a layer model in place of a general-purpose binary tree.

Two out of four sounds like a hedge, but the two it gets are the two that a distribution actually is. A distribution is not its package manager; Sabayon replaced its installer, its init system and its desktop defaults repeatedly and remained Sabayon. The claim "MocaccinoOS is Sabayon's successor" survives the audit, with one amendment: it is a successor that kept the goal and rebuilt the machinery from scratch, and it took three and a half years of shipping nothing to do it.

Which makes the title of this piece a false choice, and I would rather say so than pretend the framing worked. There is no estate. Sabayon's repositories are gone, its forum and wiki are gone, and sabayon.org answered HTTP 521 from Cloudflare — an origin server that is not there — when I requested it on 2026-08-04, while wiki.sabayon.org and forum.sabayon.org did not answer at all. Nobody inherited that, because by 2020 there was nothing left to inherit but a name and four people's willingness to keep going. What they took was the only asset that had survived, and they were right that it was the valuable one.

What would change my mind. The base finding rests on the build definitions as they stood at the moment each image was tagged, so it fails if the ISOs were not built from that repository at all — a MocaccinoOS ISO from October 2021 or February 2022 with Funtoo package metadata inside it would do it, and I would write the correction. If the signature block on the release notes loses Di Giacinto and the cadence holds anyway, the "people" score gets stronger, not weaker; if it loses two of four and the cadence breaks, this becomes a different piece. And if someone produces a completed Sabayon 19.03 to MocaccinoOS 26.x in-place migration with a transcript, I want to see it, because I have found the documentation and not the evidence.

Who this is for, and who should be reading something else

On the documented properties, MocaccinoOS suits someone who wants a Gentoo-derived desktop without compiling one, who is comfortable with a small curated repository plus Flatpak or Docker for anything unusual, and who understands that they are trusting a four-person volunteer project. The immutable-adjacent layer model is a real design idea and not marketing; if it appeals, it is the reason to look.

It is the wrong answer if you want Portage itself — emerge, USE flags, the full tree — with binaries to skip the compile. That is now Gentoo's own binary host, and the argument about how it got there is a piece of its own. It is also the wrong answer if you are choosing between several options rather than asking about this one specifically; the side-by-side comparison, with the cells I could and could not source, is in the five candidates for "Sabayon in 2026". This piece deliberately does not repeat that table.

Where to go next

If you arrived asking what happened to Sabayon rather than what MocaccinoOS is, the timeline with a date on every step is here. Every distribution named on this page has a dated row in The Register, where this piece closed the base cell and left one open: whether a fresh install preselects proprietary drivers and codecs, or merely offers them.