FoRecoML
The machine-learning arm of a forecast reconciliation toolkit, four months old and already sharing its sibling's plumbing.
A side-by-side editorial comparison of Nebula and projoint — release velocity, themes, recent moves, and the top alternatives to consider.
Overlay network that rewrote its certificate format, then spent a year fixing what it exposed.
Nebula is a peer-to-peer mesh VPN built around a certificate authority model. The v1.10.0 release was the pivot: IPv6 and multiple addresses in the overlay, plus a new v2 ASN.1 certificate format with a unified interface for external implementations. Everything since has been consequence management — a P256 signature malleability issue that allowed blocklist bypass, a source-IP acceptance flaw tied to the new multi-address certificates, and a run of fixes around route tables and Windows listeners.
projoint spent a year on CRAN paperwork, then shipped a correctness fix it flagged itself.
projoint is an R package for analysing conjoint survey experiments, covering Qualtrics import, reshaping, and quantity-of-interest estimation with inter-rater reliability correction. Most of its release history is CRAN admission work — citation formats, DESCRIPTION fields, \value{} tags, vignette cleanups — with four tags backfilled within ninety seconds of each other on 15 July in non-monotonic version order, so neither tag order nor timestamps in this feed track the real sequence. The substantive releases are the ones fixing data-preparation bugs that silently corrupt estimates.
Nebula is a peer-to-peer mesh VPN built around a certificate authority model. The v1.10.0 release was the pivot: IPv6 and multiple addresses in the overlay, plus a new v2 ASN.1 certificate format with a unified interface for external implementations. Everything since has been consequence management — a P256 signature malleability issue that allowed blocklist bypass, a source-IP acceptance flaw tied to the new multi-address certificates, and a run of fixes around route tables and Windows listeners.
The project has moved from single-IPv4-per-node assumptions toward a genuinely flexible addressing model, and the security fixes since v1.10.0 map directly onto that change — the new certificate features widened what the code has to validate. v1.11.0 shifts attention to the operational surface instead: structured logging, corrected firewall reject-versus-drop semantics, and Windows WFP filters installed by default.
The stated plan to assert low-s signature form when validating certificates is the concrete next step visible in these entries; expect it to land as a breaking validation change in a future release.
projoint is an R package for analysing conjoint survey experiments, covering Qualtrics import, reshaping, and quantity-of-interest estimation with inter-rater reliability correction. Most of its release history is CRAN admission work — citation formats, DESCRIPTION fields, \value{} tags, vignette cleanups — with four tags backfilled within ninety seconds of each other on 15 July in non-monotonic version order, so neither tag order nor timestamps in this feed track the real sequence. The substantive releases are the ones fixing data-preparation bugs that silently corrupt estimates.
The maintainer is hardening the path from raw Qualtrics export to estimate, which is where conjoint analysis quietly goes wrong. Three separate releases fix that path: dropped respondent-level weights in organize_data(), repeated-task reshaping in reshape_projoint(), and choice-to-profile mapping in 1.1.3. Each fix now arrives with regression tests and stricter validation rather than just a patch, and 1.1.3 adds an explicit .choice_map so the mapping is auditable instead of inferred.
Expect the validation-and-regression-test pattern to keep extending across the import path, with releases continuing to arrive in bursts around CRAN submission rather than on a cadence.
Other Infra & APIs products tracked by Sparkpulse, ranked by recent ship velocity. Each card links to a full editorial trajectory and lets you pivot into a head-to-head comparison with either Nebula or projoint.
The machine-learning arm of a forecast reconciliation toolkit, four months old and already sharing its sibling's plumbing.
Forecast reconciliation with a real object model, five years after it started returning bare matrices.
A textbook data package whose whole job is to stay installable, and whose releases prove how much work that is.
A graph-centrality package that spent 2026 making its existing measures usable at scale, then went quiet.
A test-theory package that grew into a graphical-model toolkit, now spending its releases paying down the API debt that growth created.
nuggets keeps compounding on the 2.0 rewrite — more pattern families, lighter install.
See all Nebula alternatives → · See all projoint alternatives →
Latest ship moves from both products, interleaved chronologically. ⚡ = editorial spark.
They serve adjacent needs but don't currently overlap on shipped themes. Nebula and projoint are shipping at a similar cadence (velocity 2.5 vs 2.5, both within Sparkpulse's "active" band). See the at-a-glance table above for a side-by-side breakdown of velocity, recent sparks, and editorial themes.
Sparkpulse doesn't pick a winner — we score release velocity, not feature parity. Nebula and projoint are shipping at a similar cadence (velocity 2.5 vs 2.5, both within Sparkpulse's "active" band). For your specific use case, the alternatives sections above list other Infra & APIs products to evaluate alongside.
Top Nebula alternatives in Infra & APIs are ranked by recent ship velocity. Browse the "Nebula alternatives" section above for the current picks, or visit /alternatives/nebula-networking for the full list with editorial commentary on each.
Top projoint alternatives in Infra & APIs are ranked by recent ship velocity. Browse the "projoint alternatives" section above for the current picks, or visit /alternatives/projoint for the full list with editorial commentary on each.