Repository navigation
Roadmap
As of 2026-09-29. This page is a map of open projects and open questions, organized by area, each with a one-line status. It is not the list of missing heads, statistics or examples — those are generated and committed in the repo, and they are the source of truth for "what's missing" (see Where to look for work below). If an item here is done, it should come off this page; a roadmap with finished items on it is how a roadmap stops being read.
Status values: idea (a shape, no plan), designing (a doc or proposal exists, undecided or unbuilt), in flight (a lane, PR or phased build under way), done (landed — kept here only when the area has more to do).
-
FullSimplify— idea. The most-used head in the Wolfram frontier with no answer here. -
The
Function*property family (FunctionDomain,FunctionRange,FunctionInjective,FunctionMonotonicity,FunctionSingularities, ...) — idea. Individually small, collectively the largest coherent block on the Wolfram frontier; none of it exists yet. -
HarmonicNumber-shape audit — in flight. A recurring check (census/src/head-map-audit-data.tsHEAD_MAP_AUDIT,vp node packages/census/scripts/audit-head-map.ts) for heads the head map claims and the transpiler emits but the engine doesn't actually evaluate; a guard test refuses new undeclared rows. - Upstream candidates ready to send — designing. See Upstreaming §8 for what's ready today, each with its reproduction; each draft needs the maintainer's sign-off before posting (see Lanes).
-
Equality and
SameTest— idea. compute-engine hasIsSame(structural, doesn't evaluate its arguments),Same(evaluates, variadic) andEqual(compares lists whole, threads a scalar against a list). Wolfram hasSameQ,Equal, and aSameTestoption (VerificationTest,Union,DeleteDuplicates,Gather) taking a binary predicate, used for "within tolerance" and, with a canonical form, "equal up to". Align ours with that, and decide whether a named equivalence or tolerance head (Almost?) earns its place. See Speculative Options Restrictions and Orders §6. -
Declaration order is load-bearing (BL-13) — designing. Which head-overload wins
currently depends on package declaration order (
declareMapsvs.declareDomainPlurals); the fix makes winners explicit via a golden test now, and specificity-based overload insertion later. Owner: coordinator A with the Protocols session (engine and structures); proposed, scheduled before the combinatorics migration's step 6.
- Numerals as polynomial-like domains — idea. Represent a numeral as a structured object rather than a digit list, so factoradic (the Lehmer code) and the combinatorial system (subset unranking) become mappable onto the structures they secretly are. See Speculative Symbols.
-
modularpackage normalization — idea. 34 heads, heavily prefixed (Form*,Modular*), several named for their implementation rather than the mathematics, and no Wolfram counterpart to align against. See Speculative Symbols. - Numeral system renames — in flight. Decided (option A, 2026-09-25); renames are being executed. See Speculative Numeral Naming.
-
The enumeratio/notatio name split — in flight.
enumeratiois the symbol definitions and their evaluation;notatiois the notebook, explorer and tools around them. Package and folder names don't all say so yet; renaming waits for a quiet moment, one at a time (see Component Naming for how the last one went).
-
A canonical form for the
<Head>syntax — designing. Which arguments lift to named attributes, how attributes are ordered, and when atoms take the short spelling, so markup and MathJSON map one to one. See Speculative Vdom Canonical Form. -
Writing more of the docs in enumeratio's own terms — idea. Builds on the canonical
<Head>form above. See Speculative Docs in Enumeratio. -
Prose through the engine (BL-8, folds in "notatio prose islands") — in flight. Record prose reads into prose boxes and renders with KaTeX at build (#409); every typeset formula, example cells included, is KaTeX, with MathLive only to edit (#416). Next: guides onto the same renderer, holes filled by the page's environment,
<Head>tags evaluated in prose. See Speculative Prose Pipeline. -
MakeExpression: boxes read back as expressions — idea. MathML and AsciiMath input reach an expression through the boxes' LaTeX and compute-engine's parser (
boxesToExpressionin@enumeratio/formats), which loses what LaTeX can't say and trips on notation CE reads differently. A reader from boxes straight to MathJSON (Wolfram'sMakeExpression, the inverse ofmakeBoxes) replaces that seam, and is exact for our own output wherever the boxes carry their expression. See Boxes. -
AsciiMath through boxes — idea, after MakeExpression. AsciiMath is read by asciimath2tex into TeX today; a reader onto boxes (the symbol table is linear.ts's) would put it on the same path as MathML and give us the dialect to control.
-
StandardForm in the terminal, and forms per environment — idea. The CLI maps
standardformto InputForm because a terminal can't typeset boxes, but it can do a lot: Unicode, script characters, colour, weight, sometimes a second line. StandardForm there should be an enriched text rendering of the same boxes (OutputForm, plus styling and 2D layout where the terminal allows), and the alias lifted. More generally each form should say what it becomes in each Rendering Environments environment (web, print, tty, pipe, compact); the forms table on Syntax and Formats is the start of that.
-
Controlled evaluation for our own test suites — idea. An isolated evaluation process
with per-test
TimeConstraint/MemoryConstraint, the way the oracle scans are already capped today. See Speculative Evaluation (note: this doc's own provenance line points at adesign/evaluation.mdthat no longer exists at the top level — possibly folded into Computation).
-
Scenes — idea. Wolfram's
GeometricScene(a scene solved from its constraints) as the right generalization for more than synthetic geometry: every display we draw into (plot, worksheet, 3-D view) as a scene with its own coordinate system and camera, and the projective geometry over Cl(p,q,1) asProjectiveScene. The unified-rendering speculation (Speculative Graphics and Space, Speculative Rendering Environments) should be rewritten in those terms. -
Site build: heavy packages can't be split into lazy chunks (BL-12) — idea, open.
frontend,compute-engine,boxesandformatsare imported both dynamically and statically acrosscomponents/src, so Rollup keeps them eager; real gains need narrowing the eager surfaces (type-only/lightweight entry points), which would cut shipped page JS more than build time. Owner: B (web), with the package owners.
-
Each area owns its carriers; Plausible as symbols (#401 / #402) — designing, in
flight (lane A-80). Re-layers the combinatorial packages so a carrier sits beside its
families, statistics and maps, then re-expresses Plausible as a composition of heads over
those typed values. Nothing built yet; sequenced after #401 and after A-78 finishes typing
the remaining families. This is the design behind the #402 migration (collections and
domains merging into
@enumeratio/combinatorics, retiring the standalonedomainspackage). See Speculative Combinatorics Layering and Plausible. -
CombinatorialStat/CombinatorialMap: statistics and maps live on their collection — in flight. Signed off; phases 1 and 2 built. See Speculative Statistics and Maps. - Carriers, collections, orders, restrictions and options — designing. The vocabulary (carrier set, carrier, collection, representation, restriction), orders as first-class (implied or chosen; siblings for distinct orders), grading and filtration, species, and when an option is an option vs a separate carrier. See Speculative Options Restrictions and Orders.
-
A function repository — idea. Symbol definitions should keep growing without the core
packages growing with them — the model is Wolfram's Function Repository
(
ResourceFunction["Name"]resolving through a registry to a lazily-loaded definition carrying its own examples and tests). Namespaces §3 has theResourceFunctionanalysis; Speculative Namespaces has the lazy-context idea it leads to. Nearly every catalog addition, including the combinatorics areas above, is a candidate for the repository rather than core. -
Collections as pure data — idea (the maintainer, 2026-09-29). As much of a collection as
possible — carrier, family, statistics, glyphs, rendering — becomes data in its symbol
metadata rather than bespoke code per collection. Glyphs and renderers become expressions
that return boxes (e.g. a vector-graphics box), with a minimal, nailed-down set of
rendering primitives and fewer bespoke pieces living in the core packages. A function
repository (above), whose entries go into a registry, is what makes each piece
independently and efficiently testable rather than a monolith. Related: the #402
combinatorics-layering migration above (carriers, families, statistics and maps moving
into
@enumeratio/combinatorics); BL-8 / #398, prose through the engine (Speculative Prose Pipeline, onmain), since a glyph/renderer-as-expression needs the same "expression projects into a form" machinery prose does; and the boxes design (Boxes) for what a returned box actually is.
-
A unit belongs to its algebra — idea. A unit is a value of an open family,
Unit(family, k), not a symbol; what it squares to and commutes with are relations of the family. See Speculative Algebras Units and Places. -
Quadratic and algebraic integers — idea. number-theory has the Gaussian integers ℤ[i];
the same treatment (norm, units, class number, factorisation where unique) belongs to
every quadratic integer ring ℤ[(1+√d)/2], and past that to rings of integers of number
fields generally. The number-field layer the adeles' next phase and the quaternion places
above both wait on. Precedents: Sage's
QuadraticField/NumberField, Mathlib'sNumberField.RingOfIntegers. -
SqrtSpace— idea. The Function Repository'sSqrtSpacemaps Cartesian points to algebraic numbers in a quadratic field and back, tracking signs — a candidate head and a Wolfram counterpart for number-theory's quadratic-field work. -
Fixed-precision decimals and money — idea. An exact decimal type with a fixed scale,
rounding to its last place by a stated rule. Not a ring (money × money isn't money), but
an ordered additive group rationals can scale; its ticks (whole units or cents) floor and
round through structures'
FloorOrderextension without needing one of its own.
The Postgres-era predecessor (~/Playground/@enumeratio/archive/) did things this repo
doesn't yet; the catalogue itself came across whole, the gap is behaviour. Full detail in
Legacy Port.
- Representations and table control — idea. Several glyph kinds per carrier, glyph families not yet drawn, alternate text notations that round-trip, printer policy (GRANT/REVOKE-style), typed columns, column controls, a statement bar, element/identity panes. Legacy Port §1.
- Polytopes — idea. Cyclohedron W_n (issue #330 open: an explicit unrank), vertex-only families (hypersimplex, Birkhoff, type-B permutahedron), multi-polytope overlay, guide content. Legacy Port §2.
-
Combinatorial species — idea; open decision on whether species become the backbone
here or stay an annotation. Every collection's record should name its species, and ideally
species are values in Epsil (+, ·, ∘, derivative, evaluation at a label set, generating
functions, cycle indices), with L-species as where a collection's order comes from. See
Speculative Options Restrictions and Orders §5. About thirty readings on
main-equivalent branches, another thirty on an unmerged branch, plus the cycle-index/plethysm machinery and a slow definitional-enumerator oracle. Legacy Port §3. - Collections without kernels — idea. 184 of 282 catalogued collections have no declared head; 89 are numeric sets. Legacy Port §4.
-
Docs & design ports — idea. Counting guides, explorations, and wiki design pages with
no
design/counterpart (Grading, Triangular Grading, Number Gradings, Sequence Transforms, Maps and Bijections, Function Properties, Family Parameters and Numerals, Dimensions and Measures, Adaptive Enumeration Windows, Self-Certification, Element Relations, Parameterized Collections) and the four system surveys — see see the sidebar's "Archive & surveys" group (Home) for which of those the wiki now carries directly. Legacy Port §5.
-
knots.md's table — idea, parked. The value landed; the table did not. See Knots.
Eight lists, all committed, all regenerable, each saying what we do not do and, where the answer is interesting, why not. These lists — not this page — are the source of truth for what's missing.
| list | what it holds |
|---|---|
census/src/wolfram-frontier-data.ts FRONTIER
|
heads Wolfram's own documentation examples reach for that we cannot answer, ranked by how often |
census/src/wolfram-frontier-data.ts CALL_FORMS
|
argument shapes a symbol we DO map is shown taking, next to ours |
statistics/src/frontier.ts FRONTIER
|
statistics with no definition, each with a stated reason |
domains/src/frontier-maps.ts UNDEFINED_MAPS
|
catalog maps not defined yet, same discipline |
reference/src/fungrim-verified-data.ts fungrimFrontier
|
heads whose Fungrim identities our engine cannot evaluate, ranked by how many identities each holds up |
census/src/head-map-audit-data.ts HEAD_MAP_AUDIT
|
the HarmonicNumber shape, generalized: every mapped head the engine doesn't declare or doesn't actually evaluate |
census/src/rename-queue.ts RENAME_QUEUE
|
heads still declared under a spelling we have decided against, with what blocks each |
<package>/reference/*.yaml aspirational: true
|
examples we document but don't yet produce ("not yet implemented"), triaged in issue #92 |
The Wolfram frontier is the one to open first when the question is "what should this system be able to do that it can't" — derived from ~10,000 documentation examples, ranked by what the people who built those functions thought worth demonstrating. See Symbols §4 for how it's collected; regenerate with a Wolfram kernel on PATH:
vp node packages/census/scripts/collect-wolfram-frontier.tsThe Fungrim frontier is the newest and most mechanical to work through (1444 identities as
rewrite rules, vp run crosswalk:verify to regenerate, ~45s). Every disagreement is
classified in KNOWN_CAUSES (packages/reference/src/crosswalk/fungrim.ts).
The two statistics/maps frontiers are small, well-specified, and each already carries the argument for why it was left. The statistics one is empty as of September 2026 — every entry it ever held was pessimistic and moved off once checked; treat anything landing there next as unproven, not impossible.