Roadmap review — 2026-Q3

[marius-bughiu]

Quarterly roadmap reflection for 2026-Q3 (generated 2026-07-18). This is the big-picture pass: which milestones to close, ROADMAP-vs-evidence drift, and features that shipped without ever being rostered. It does not edit ROADMAP.md or touch milestones — it surfaces signal for the maintainer to act on.

Headline: 5 open milestones reviewed, all at 100% (0 open issues repo-wide). 17 items flagged (1 drift, 0 stale, 16 unrostered features). 6 suggestions below.

Milestone status

Milestone	Open	Closed	% complete	Suggestion
1.1.0	0	14	100% (14/14)	close
1.2.0	0	14	100% (14/14)	close
1.6.0	0	4	100% (4/4)	close
2.0.0	0	11	100% (11/11)	close
2.1.0	0	11	100% (11/11)	close

Every open milestone is fully closed out — there are no open issues anywhere in the repo. All five are closure candidates with nothing to migrate.

Drift

• 2.0.0 milestone prose is stale — ROADMAP.md (milestone 2.0.0 intro) still frames the release as human-gated: "What remains is the human-gated go/no-go on the actual v2.0.0 tag (Ready to release v2.0.0 — needs human review for breaking changes #213)." Evidence contradicts this: v2.0.0 was tagged and released 2026-06-21 (v2.1.0 and v2.2.0 have since shipped), and #213 is closed. The item is done; the prose describes it as pending. (Opposite polarity to classic drift, but a ROADMAP-vs-evidence mismatch worth correcting.)
• Otherwise clean — every other Status: done item traces to a closed GH issue and/or a CHANGELOG entry.

Stale items (no activity in 90+ days)

• None. There are zero open issues in the repository.

Suggestions

• Close all five open milestones (1.1.0, 1.2.0, 1.6.0, 2.0.0, 2.1.0) — each is 100% complete with zero open issues; nothing to migrate to a next milestone.
• Update the 2.0.0 milestone prose — drop the "what remains is the go/no-go on the v2.0.0 tag (Ready to release v2.0.0 — needs human review for breaking changes #213)" framing; v2.0.0 released 2026-06-21 and Ready to release v2.0.0 — needs human review for breaking changes #213 is closed.
• Add to ROADMAP retroactively — post-2.1.0 collection crop (shipped in v2.2.0 / [Unreleased], no milestone, 0 ROADMAP mentions): DisjointSet<T>, Deque<T>, LruCache<TKey,TValue,THasher>, XorFilter<T,THasher>, EnumMap<TEnum,TValue>, EnumSet<TEnum>, SmallSet<T>, and the specialized-set parity family RobinHoodSet / HashCachingSet / PooledCeleritySet. These follow the established "tier-(c), roadmap otherwise exhausted" convention but none are captured in ROADMAP — roster them or note the convention explicitly.
• Add to the Celerity.Primitives roster — BitWriter / BitReader (span bit-cursor pair, [Unreleased]) is absent from the milestone 2.1.0 Primitives list it naturally extends.
• Roster the "Built with Celerity" showcase tier — three new packages (Celerity.Ring, Celerity.Sentinel, Celerity.Cardinality) shipped with 0 ROADMAP mentions. The ROADMAP Vision section lists only Celerity.Collections / .Hashing / .Primitives; this downstream-showcase strategy is a real directional expansion that belongs in Vision (or noted as out-of-scope for the core roadmap).
• Open a rolling post-roadmap milestone (e.g. 2.3.0) to house ongoing "tier-(c)" enhancements. The roadmap is 100% done through 2.1.0 yet shipping continues every cycle; a standing milestone gives that work a home instead of it landing unrostered.

[marius-bughiu]

Acted on — with two corrections to the review above

All six suggestions are done, via #302 and the milestone changes below. Two things in the original report need correcting:

1. v2.3.0 shipped mid-review, so the rolling milestone is 2.4.0, not 2.3.0.
The review was written against a snapshot where v2.2.0 was the latest release. v2.3.0 released 2026-07-19 while it was in progress. The standing milestone is therefore 2.4.0, and 2.3.0 is rostered as a shipped milestone in ROADMAP.md.

2. The unrostered list was incomplete — it is 14 types, not 10.
The review checked the surface as of the snapshot. Re-running the audit against current main adds four more that shipped or merged with no roadmap entry: IndexedPriorityQueue (v2.3.0), and Trie, SparseSet, FenwickTree (merged, awaiting a tag). All are now rostered.

What changed

Suggestion	Status
Close the five completed milestones	Done — 1.1.0, 1.2.0, 1.6.0, 2.0.0, 2.1.0 all closed at 0 open / 100%
Update the stale 2.0.0 release-gate prose	#302
Roster the post-2.1.0 collection crop	#302 — new milestone 2.3.0 section, 14 types grouped by why each exists
Roster BitWriter / BitReader	#302 — added to milestone 2.1.0's Primitives list
Roster the "Built with Celerity" showcase tier	#302 — described in Vision as a downstream tier, explicitly not a fourth core package
Open a rolling post-roadmap milestone	Done — 2.4.0, with the four open issues rostered onto it

A third drift item turned up while editing and is fixed in #302: the Vision section still described Celerity as shipping "a single NuGet package", which the v2.0.0 split superseded.

Closing this once #302 merges.