Skip to content

Roadmap Stability Families

Chris Dare edited this page Nov 15, 2026 · 1 revision

Stability Families

schema: roadmap/1
slug: stability-families
project: derived-alg-geo-lean
title: "Stability conditions in families — from absolute foundations to relative moduli"
status: active
phase: sequenced
brief: |
  Treat Bayer–Lahoz–Macrì–Nuer–Perry–Stellari, Stability conditions in
  families, arXiv:1902.08184v4, as a program-level north star rather than a
  single theorem milestone. The current repository has strong absolute and
  numerical foundations: group actions, the stability distance, an
  independently reviewed HRS t-structure, an abstract support-property lane,
  and numerical wall nesting away from the charge-zero locus. The repository
  now also owns the coherent-sheaf and derived-category substrate that had been
  tracked separately when this roadmap was first written.

  The repository still does not have all relative derived-category,
  boundedness, Quot, stack, or moduli infrastructure required by the paper.
  This roadmap lands the recovered absolute layer while advancing the paper's
  weak-stability/tilting spine, and keeps every remaining geometric dependency
  visible with an explicit in-repository or external owner.

goal:
  hmw: "How might we advance from an honest absolute Bridgeland foundation to stability conditions in families while making every geometric and moduli dependency explicit?"
  objective: "Formalize the reusable abstract spine of arXiv:1902.08184v4, recover the unfinished absolute proof chain, and establish typed cross-repository boundaries for the geometric and moduli program."
  key_results:
    - "The absolute stability topology no longer accepts HN-mass subadditivity as an external premise, and the recovered proof has current CI plus independent mathematical review."
    - "The HRS t-structure is reviewed, its heart is related to the textbook extension closure under explicit hypotheses, and §14 weak-stability tilting is formalized without geometric assumptions."
    - "Support is expressed by a genuine bundled quadratic form on semistable classes and has an explicitly separate uniform-in-families layer."
    - "The five conditions in Definitions 20.5 and 21.15 are represented separately; stable declarations bind useful categorical or geometric data, while vacuous-probe inhabitance checks (constant-true, full-locus, arbitrary-Prop) remain development-only scaffolding; genuine-input constant-family reindexings remain stable API."
    - "Every result mapped to this program is pinned to 1902.08184v4 and defaults to target/no-claim until its statement and proof have been reviewed."
    - "Coherent-sheaf and derived work is integrated through the monorepo's explicit package boundaries; Quot, boundedness, stack, and good-moduli work remains assigned to a named owner and is never hidden behind an axiom."
  assumptions:
    - tier: must
      text: "The dormant HN mass-triangle branch contains reusable correct mathematics after it is ported onto current main."
      validation: "The source history is preserved under archive branches. Destination migration PRs #218 and #219 carry the remaining reviewed mass/topology and H⁰-cokernel slices on v4.32.1."
    - tier: must
      text: "The Hom-orthogonal HRS aisles agree with the textbook cohomological formulation under hypotheses expressible at the pinned anchor."
      validation: "Validated in the source repository by PRs #76/#93/#97 and transferred with their history. The destination tracker records the tilted-heart extension work as closed issue #204."
    - tier: should
      text: "Useful relative-HN interfaces can be stated abstractly before the full geometric substrate exists."
      validation: "Require explicit type-level premises and concrete categorical or geometric realization for stable results. Logical inhabitance checks may live under Development but do not count as mathematical progress."
    - tier: should
      text: "Quot, algebraic stacks, and good-moduli foundations should be reusable beyond this paper."
      validation: "The parked cross-repo epic must select an upstream or separate-package owner before implementation begins."
  wont:
    - "No wholesale merge of origin/agent/hn-mass-triangle-chain; only reviewed current-main slices."
    - "No implementation may bypass the monorepo's subject boundaries under DerivedAlgGeo or introduce a hidden trust assumption."
    - "No placeholder axiom, sorry-backed instance, or vacuous structure to make a paper definition typecheck."
    - "No claim that ingestion in arXMCP establishes paper coverage or source-version identity."
    - "No claim of Theorems 22.2, 24.1, or 29.2 until their geometric, boundedness, and moduli prerequisites are formalized."
  evidence:
    - "arXiv:1902.08184v4 — authoritative paper version, 146 pages, Parts I–VI."
    - "PR #93 — independent declaration-by-declaration review of PR #76's HRS t-structure; owner dispositions recorded in issue #86."
    - "PR #97 — both Hom-orthogonal aisles connected to the textbook truncation formulation; issue #94 closed."
    - "Destination PRs #218–#223 — v4.32.1 ports of the remaining source PR stacks, with source PR/head provenance and fresh validation recorded in each PR."
    - "PR #101 — 1902.08184v4 source pin and zero-claim coverage map; all theorem promotions remain review-gated."
    - "The AlgebraicGeometry subject and destination B1–B5 milestones — geometric substrate and explicit HRR trust boundary."
    - "arXMCP bridgeland-stability notebook — paper present with 548 chunks; ingestion alone is not a source-binding claim."

generated_by: { agent: codex, at: "2026-08-08" }
generations:
  - { rev: 1, at: "2026-08-07", note: "Initial program materialized as GitHub milestones #6–#9, epics #79–#84, and active SF0 children #85–#90." }
  - { rev: 2, at: "2026-08-08", note: "Refreshed after #85–#87 and #94 completed, the #88–#90 recovery stack reached reviewed green drafts, and SF1 children #106–#111 were materialized." }
  - { rev: 3, at: "2026-08-12", note: "Migrated to derived-alg-geo-lean: remapped milestones and issues, updated monorepo ownership boundaries, and linked destination migration PRs #218–#223." }
  - { rev: 4, at: "2026-08-14", note: "Unified source layout recorded. SF0 is complete and milestone #14 is closed; current modules live under DerivedAlgGeo/CategoryTheory/Triangulated/StabilityCondition. Active geometry links now point to the in-repository AlgebraicGeometry subject, while the reusable stacks/moduli boundary remains an owner decision in #214." }
  - { rev: 5, at: "2026-08-14", note: "SF1 source-faithfulness closeout kept milestone #17 open and reopened #208: the current Lemma 14.17 biconditional stores an extra Hom-vanishing implication in its type-two branch. The strengthened candidate remains mapped but is not literal source coverage." }
  - { rev: 6, at: "2026-08-15", note: "SF2 implementation started with native child #380 under epic #193: bind ordinary slicing semistability to genuine and indexed-uniform quadratic support before tackling the Definition 21.9 quotient and geometric fibers." }
  - { rev: 7, at: "2026-08-15", note: "Native child #387 under epic #193 now connects the saturated integral quotient to the real support quotient and its scalar extension. Bijectivity is an explicit model-specific premise, not an automatic consequence of saturation." }
  - { rev: 8, at: "2026-08-15", note: "Native child #389 removes the one-fixed-category restriction from the ordinary numerical adapter: actual slicings in varying fiber categories map to one common class space with one charge and quadratic form. Geometric base change remains separate." }
  - { rev: 9, at: "2026-08-15", note: "Native child #391 under epic #194 binds the five-clause ordinary families package and the Theorem 22.2 source-clause projection to actual slicing-defined loci on varying abstract fiber categories. The geometric probes remain explicit premises." }
  - { rev: 10, at: "2026-08-15", note: "Native child #393 under epic #194 replaces the discrete collection of unrelated fibers with a contravariant category-valued family, triangulated pullbacks, functorial K₀ maps, and compatible common class maps. It remains an abstract categorical base, not a scheme-derived family." }
  - { rev: 11, at: "2026-08-15", note: "Native child #395 under epic #194 binds actual fiber pre-stability conditions to the categorical pullbacks: explicit preimage-slicing witnesses give phase transport, while common class maps and central charges give class, charge, and semistable-locus transport. Construction of those witnesses from geometry remains outside the abstract package." }
  - { rev: 12, at: "2026-08-15", note: "Native child #397 under epic #194 combines the five-clause ordinary fiber package with #395's categorical base-change compatibility. The adapter projects the Theorem 22.2 source clauses and transports actual semistable classes and their common charge; it still asserts no geometric conclusion." }
  - { rev: 13, at: "2026-08-15", note: "Native child #399 under epic #194 introduces the first scheme-facing layer: base changes are actual objects of Over S and charge, open-locus, and generic-semistability probes use schemes with their Zariski topology. Derived categories, pullback, and every geometric theorem remain explicit obligations." }
  - { rev: 14, at: "2026-08-15", note: "Native child #401 under epic #194 extracts actual residue-field point fibers from a scheme base change T -> S. The Spec kappa(x) geometry and canonical map to T are concrete; the triangulated residue category and its pullback remain evaluations of the client-supplied family." }
  - { rev: 15, at: "2026-08-15", note: "Native child #403 constructs the unbounded and bounded derived categories of O_X-module sheaves for every scheme X and its residue-field schemes, then supplies objectwise realization interfaces. Derived pullback and bounded coherent or perfect subcategories remain explicit follow-ons." }
  - { rev: 16, at: "2026-08-15", note: "Native child #405 connects scheme-derived fibers to Mathlib's actual additive pullback on module sheaves. Exact pullbacks descend to triangulated derived functors with localization and K0 compatibility; flatness criteria, nonexact left-derived pullback, and derived pseudofunctor coherence remain follow-ons." }
  - { rev: 17, at: "2026-08-15", note: "Native child #407 proves that exact scheme pullback is closed under identities and composition, and lifts Mathlib's coherent identity and composition isomorphisms to cochain complexes. Derived-level pseudofunctor isomorphisms and coherence remain explicit follow-ons." }
  - { rev: 18, at: "2026-08-15", note: "Native child #414 descends the exact-pullback unit and compositor through derived localization and proves identity and composition for the induced K0 maps. Triangle, pentagon, and shift-compatibility laws remain explicit follow-ons." }
  - { rev: 19, at: "2026-08-15", note: "Native child #417 proves the complex- and derived-level left-unit, right-unit, and associativity laws for exact pullback, with canonical equality congruence lifted through localization. Shift compatibility remains an explicit follow-on." }
  - { rev: 20, at: "2026-08-15", note: "Native child #425 proves that the localization comparison, complex and localized unit/compositor, and exact derived pullback unit/compositor commute with integer shift. Flatness criteria, nonexact left-derived pullback, and bounded coherent or perfect fibers remain explicit follow-ons." }
  - { rev: 21, at: "2026-08-15", note: "Native child #429 starts the flatness criterion without adding a trust assumption: ordinary module pullback is automatically right exact, exactness reduces to finite-limit preservation, the local-ring-valued module-stalk functor is constructed and proved left exact, and the flat stalkwise scalar-extension model is left exact. The natural pullback-to-stalk comparison needed to install exact pullback from geometric flatness remains the next explicit slice." }
  - { rev: 22, at: "2026-08-15", note: "Child #429 now proves that module stalks jointly detect isomorphisms and promotes pointwise finite-limit preservation to a global module-sheaf theorem. The exact comparison datum for arbitrary pullback is typed, flatness plus that datum implies exact pullback, and the comparison is discharged unconditionally for open immersions via Mathlib restriction, giving an exact derived pullback instance there. Constructing the arbitrary pullback-to-stalk comparison remains explicit." }
  - { rev: 23, at: "2026-08-15", note: "Child #429 factors out the presheaf-module stalk functor and proves that module sheafification preserves every module stalk. It also makes the pullback-to-stalk comparison contract canonical for identity morphisms and closed under composition, using the corresponding laws for module pullback, scheme stalk maps, and extension of scalars. The remaining task is therefore the presheaf-pullback comparison for an arbitrary scheme morphism, not sheafification or separate coherence work for composites." }
  - { rev: 24, at: "2026-08-15", note: "Child #429 now exposes the presheaf-module pullback beneath scheme module pullback and proves that its stalk controls the sheaf-level stalk comparison through Mathlib's pullback construction and stalkwise invariance of sheafification. Consequently, flat exactness reduces to one typed PresheafPullbackStalkComparison; no sheafification or sheaf-level comparison premise remains. Constructing that presheaf comparison for an arbitrary scheme morphism is the remaining theorem." }
  - { rev: 25, at: "2026-08-15", note: "Child #429 now proves the local comparison over neighborhood categories: pullback along the induced neighborhood ring diagram followed by its module colimit is naturally isomorphic to the source stalk followed by extension of scalars along the scheme stalk map. The remaining presheaf comparison is therefore isolated to global-to-neighborhood pullback base change." }
  - { rev: 26, at: "2026-08-15", note: "Child #429 completed the arbitrary presheaf pullback-to-stalk comparison and installed exact derived pullback for every flat scheme morphism. Epics #193 and #194 and milestone SF2 are closed; the source coordinate remains target because geometric slicing, openness, relative HN, and moduli theorems are still open." }
  - { rev: 27, at: "2026-08-15", note: "Milestone SF4 (#26) begins the next families phase with epics #441 and #442. Native child #443 first makes preimage-slicing witnesses compositional before scheme inducing data, genuine Zariski openness, and relative HN over curves are attempted." }
  - { rev: 28, at: "2026-08-15", note: "Native child #444 specializes the honest left-adjoint inducing boundary to concrete exact scheme-derived pullback. The adapter records the adjoint, zero-reflection, and phase-monad premise, requires the presentable/Ind theorem explicitly, uses flatness only for exactness, and transports identity/composition through derived pullback coherence." }
  - { rev: 29, at: "2026-08-15", note: "Native child #445 removes the arbitrary-set shortcut from scheme-facing openness: the locus consists exactly of points where the actual residue-fiber pullback lies in the chosen slicing phase. Universal and generic probes reduce to Zariski openness of that locus; the named FiberPreStabilityBaseChangeData witness makes it full or empty and supplies both openness clauses, with a constant categorical model. No unconditional geometric witness or relative HN theorem is asserted." }
  - { rev: 30, at: "2026-08-15", note: "Native child #446 replaces the proposition-only Dedekind HN probe with SchemeRelativeHNFiltration: an actual HN tower over a scheme base change together with ordinary HN filtrations in every residue fiber, equality of lengths and phases, and factor isomorphisms to categorical pullback. A named geometric base-change witness constructs and pulls back the data; regular Noetherian one-dimensional curve bases and a nonempty constant-family model are explicit. General geometric existence remains a premise." }
  - { rev: 31, at: "2026-08-16", note: "SF3 is active again under epic #195 with native children #460--#462. Child #460 starts the ordered cutover by constructing Dᵇ(Coh X), the perfect thick envelope of finite locally free coherent sheaves, scheme and residue fibers, the proved inclusion Perf(X) → Dᵇ(Coh X), and typed coherent/perfect pullback restriction and coherence contracts. The geometric base-change witness and unconditional finite-type openness/relative-HN existence remain the next two children." }
  - { rev: 32, at: "2026-08-16", note: "Child #461 connects the genuine bounded coherent pullback to the family-level pre-stability witness. Triangulated equivalences now transport the non-formal preimage axioms; bounded coherent identity and composition use the geometric unit/compositor; a bounded coherent realization constructs FiberPreStabilityBaseChangeData from left-adjoint inducing premises. The Appendix-A presentable/Ind theorem remains the precisely named upstream input rather than an assumed PreimageData witness, and #462 remains the next finite-type openness/relative-HN task." }
  - { rev: 33, at: "2026-08-16", note: "Child #462 restricts the geometric witness to actual finite-type objects of Over S, using Mathlib's locally-of-finite-type plus quasi-compact predicates. Residue-fiber phase equivalence proves each semistable locus Zariski open and yields generic openness from the same locus. The slicing HN axiom constructs SchemeRelativeHNFiltration objectwise over every regular finite-type curve, with its existing residue-fiber and pullback compatibility; no openness or relative-HN conclusion is supplied as an input. These are consequences of the repository's strong strict-family interface, not a semistable-reduction or moduli theorem." }
  - { rev: 34, at: "2026-08-16", note: "Materialized SF5 (#29), SF6 (#28), and SF7 (#27). The closeout audit found that #127/#216 and #211/#215 were already implemented and audited on main; §3.4 is proposed as mapped with direct v1 PDF evidence and no coverage claim. SF7 is decomposed under #475 into bounded Corollary-A.23 truncation (#478), compact generation/Ind/A.17 (#477), scheme realization (#476), and downstream parameter removal (#479)." }
  - { rev: 35, at: "2026-08-16", note: "SF7.1 begins by representing formula (A.8) as Slicing.InducedTStructures. T-structure orthogonality now proves the Hom-vanishing half of PreimageData, with an inhabited identity model; finite phase-truncation HN existence remains the next theorem rather than a structure field." }
  - { rev: 36, at: "2026-08-16", note: "SF7.1 (#478) is complete. A charge-free head/tail theorem splits any target HN filtration with exact tail phase indexing. Formula-(A.8) truncation triangles are identified with those target splits by t-structure uniqueness, and induction on target HN length constructs source HN filtrations with strict phase windows. InducedTStructures.preimageData now has no HN-existence input; compact generation/Ind/A.17 (#477) is next." }
  - { rev: 37, at: "2026-08-16", note: "SF7.2 (#477) now owns the categorical substrate and Steps 1--4 of A.17: small-coproduct compactness, the inductive Coprod closure and universal property, adjoint preservation of compact objects, generated-aisle right t-exactness, the monad-to-t-exact Step 1, truncation comparison, bounded-subcategory restriction, exact A.3/A.4 recognition, and boundedness. IndExtensionData records the precise A.14 output without asserting existence. The remaining #477 theorem is the deep A.13/A.14 construction itself; #476 remains the Dqc/Db(Coh) scheme realization and S-locality layer." }
  - { rev: 38, at: "2026-08-16", note: "SF7.2 now owns the formal existence assembly behind A.13. AisleData constructs a t-structure from a shift-closed aisle and honest approximation triangles, proves its coaisle is the right orthogonal, and CompactGeneratorApproximation propagates G[1] ⊆ G through Coprod(G) to produce the compactly generated t-structure and formula (A.2). Polishchuk.induceOfApproximation then consumes that output directly and derives source compactness by adjunction. No compactness-to-approximation instance is asserted: Brown representability remains the exact #477 boundary, followed by A.14 and the #476 scheme realization." }
  - { rev: 39, at: "2026-08-16", note: "SF7.2 pushes the A.13 boundary from preconstructed triangles to Brown-style universal maps. ApproximationMap records the surjective Hom control and shifted injective Hom control; AisleData.ofApproximationMaps completes each map to a distinguished triangle and proves the cone right orthogonal by exactness. The compact-generator and Polishchuk constructors consume this map-level output directly. IndExtensionData.ofApproximation derives the large A.14 aisle and compact generation, leaving universal-map existence and the geometric restriction equivalences as the honest #477 boundary." }
  - { rev: 40, at: "2026-08-16", note: "SF7.2 begins the missing Brown construction with a native triangulated mapping-telescope API. For a sequence with a countable coproduct, MappingTelescope.map is 1 minus the transition shift, MappingTelescope.chosen completes it to a distinguished triangle, and Data.exists_desc proves compatible cocones factor through the telescope by triangulated exactness. This removes the first Mathlib API gap without asserting Brown representability; construction of the approximation tower and compact-source Hom computation remain next, followed by the geometric A.14 restriction equivalences." }
  - { rev: 41, at: "2026-08-16", note: "SF7.2 corrects the compact-object substrate before using it in the Brown tower. IsCompactObject now asks the additive coyoneda Hom-group functor to preserve coproducts, so the comparison target is the direct sum of Hom groups rather than the disjoint union of underlying Hom sets. The explicit coproductComparisonIso and injection compatibility theorem are proved, adjunction transport is rebuilt through the additive Hom equivalence, and all Polishchuk consumers compile. The telescope colimit computation is next." }
  - { rev: 42, at: "2026-08-17", note: "SF7.2 is complete through the owned Brown tower, A.13, A.14 restriction, and Brown-backed A.17 constructors (#507, #510, #512); #477 is closed. SF7.3 now defines the honest quasi-coherent-cohomology locus inside the all-sheaf derived category, proves exact derived coherent inclusion lands there, and constructs bounded coherent and perfect inclusions. The general-scheme bounded-coherent equivalence, inherited triangulated/coproduct structure, compact-equals-perfect theorem, and compact-generator closure equality remain explicit uninhabited #476 obligations." }
  - { rev: 43, at: "2026-08-17", note: "SF7.4 deletes the legacy global inducing-theorem switch and its bounded left-adjoint surrogate. Scheme pullback inducing packages now carry the actual phase-indexed A.17/A.8 output, and the owned Corollary-A.23 finite truncation theorem constructs every downstream preimage witness. Natural-isomorphism transport supplies concrete unbounded and bounded-coherent identity pullback models. Geometric base change, finite-type openness, and relative-HN consumers no longer accept a global theorem parameter or caller-supplied PreimageData." }
  - { rev: 44, at: "2026-08-17", note: "SF7 is closed at its supported boundary after green merged #514 and #515; the exact unproved general-scheme Dqc identifications continue as SF8.0 #528. SF8 (#31) and SF9 (#30) are materialized under epics #517 and #516. SF8.1 begins with honest Dqc total objects, explicit pseudo-coherence, local bounded stalkwise-flat Tor models, geometric flat-model fibers, and fiberwise negative-Ext vanishing; no arbitrary Tor predicate or general derived-pullback instance is introduced." }
  - { rev: 45, at: "2026-08-17", note: "SF3 and its ownership gate are closed: #460--#462 delivered the concrete derived fibers, geometric base-change witness, finite-type openness, and relative-HN package, while ADR-0013 selects the monorepo and unblocks SF8/SF9. SF8.1 (#521) delivers the relative-perfect and universally-gluable object layer with a concrete zero model; algebraicity, boundedness, and arbitrary derived pullback remain unclaimed." }
  - { rev: 46, at: "2026-08-17", note: "SF8.2 (#518) constructs a groupoid-valued relative-perfect moduli problem on diagrams of actual scheme base changes. Mathlib's Pseudofunctor owns the coherence laws, every abstract fiber is equivalent to the core of the universally-gluable relative-perfect category, and each transition is compared naturally with the repository's exact derived pullback. Isomorphism classes, geometric fibers, and a nonempty identity model are concrete; preservation along arbitrary base changes, descent, and algebraicity remain unclaimed." }
  - { rev: 47, at: "2026-08-17", note: "SF9.1 (#523) builds the reusable stack-in-groupoids substrate directly on Mathlib's Pseudofunctor, DescentData, IsPrestack, and IsStack APIs. Cover restriction is proved fully faithful and essentially surjective, the stack condition transports across explicit fiber/descent equivalences, and ordinary sheaves of types yield discrete stacks. Subcanonicity supplies a nonconstant big-Zariski stack represented by every scheme. Pullback-compatible and fiber-representability interfaces are separate from the stack condition; no algebraicity theorem is bundled or claimed." }
  - { rev: 48, at: "2026-08-17", note: "SF8.3 (#519) replaces the caller-selected boundedness switch by an actual geometric predicate: a locally-finite-type, quasi-compact scheme over the base, a universally-gluable relative-perfect universal family in the selected subfunctor, explicit geometric fibers, and coverage up to isomorphism after equality of the actual residue-field base changes. The predicate is monotone under enlarging admitted families and shrinking the target, and adapts directly to Definition 21.15(5). The identity parameter scheme and zero complex give a concrete inhabited model after every locally Noetherian base change. General residue pullback for nonzero families and semistable boundedness remain named inputs to SF8.4 rather than hidden propositions." }
  - { rev: 49, at: "2026-08-17", note: "SF8.4 (#520) introduces actual epimorphic module-sheaf quotient presentations, their commuting morphisms, target functor, and pullback functor; preservation of epimorphisms makes base change a theorem. Representability is never a name: RepresentableQuotientProblem requires a Yoneda isomorphism, and finite-type parameter spaces additionally require the actual structure morphism to be locally of finite type and quasi-compact. Since Mathlib has no general Quot scheme or Grassmannian at this pin, the supported zero-quotient functor is proved represented by the terminal identity scheme, maps to the concrete zero relative-perfect moduli problem, and inhabits SF8.3 boundedness. No nonzero Quot, semistable boundedness, or algebraicity claim is made." }
  - { rev: 50, at: "2026-08-17", note: "SF8.0 (#528) gains its first honest affine realization layer. Mathlib's tilde equivalence transports finite products and an abelian structure to the actual full category of quasi-coherent sheaves on Spec R; its derived category is therefore genuinely triangulated. Derived tilde and Gamma functors and their degreewise cohomology comparisons are constructed. The functors are not called an equivalence until localization-level unit/counit data exist, and no unbounded essential-surjectivity, general-scheme coproduct, bounded-coherent, compact-equals-perfect, generator-closure, or S-linearity statement is claimed." }
  - { rev: 51, at: "2026-08-17", note: "SF9.2 (#522) begins with an actual algebraic-stack presentation layer. Scheme maps induce pullback-compatible maps of representable big-Zariski stacks whose fibers are represented by scheme fiber products. Representable properties are checked on those actual structure morphisms; representable diagonals use scheme equalizers; and algebraic stacks carry an actual smooth-surjective scheme atlas. Representable stacks, stacks of locally finitely presented schemes, and the positive-dimensional affine-line stack inhabit the interface. The relative-perfect moduli stack, its atlas and diagonal, and its semistable open substack remain to be constructed, so #522 stays open." }
  - { rev: 52, at: "2026-08-17", note: "SF9.2 (#522) now connects the SF8 relative-perfect layer to actual algebraic stacks. A relative-perfect algebraic presentation identifies every supported locally Noetherian fiber of its structure morphism with the selected full M_pug subgroupoid; the bounded refinement lifts the finite-type witness's universal family to an actual stack-fiber object. Open presentations carry an actual representable open immersion and structural triangle. The zero subproblem is proved contractible and presented by the identity stack, with its finite-type witness and a positive-dimensional Spec Z case. The full M_pug stack still needs arbitrary pullback/descent, diagonal and atlas constructions, so #522 remains open." }
  - { rev: 53, at: "2026-08-18", note: "Architecture review #598 corrects the rev-51/rev-52 terminology. The implemented data is now named ZariskiStackPresentation: it has big-Zariski descent, a scheme-representable diagonal, and a smooth-surjective scheme atlas, but no fppf or etale descent theorem. Relative-perfect and bounded interfaces are correspondingly named big-Zariski presentations. No declaration named AlgebraicStack remains until the standard stack condition and representability level are formalized." }
  - { rev: 54, at: "2026-08-18", note: "Architecture review #599 corrects the early logical-nonvacuity strategy. Constant, full-locus, constant-true, and PUnit constructors are removed from the stable family API and retained only as compile-only Development scaffolding. The theorem-numbered arbitrary-Prop dependency contract is removed from Lean; Theorem 22.2 gaps now live in theorem22-2-dependencies.yaml as explicitly non-mathematical planning metadata. Stable adapters project theorem-neutral OrdinaryDeformationInputConditions instead." }
  - { rev: 55, at: "2026-09-03", note: "Spike: the projective-families lane (#1033) reached Slicing.IndExtensions.largePreimage (#1057), the slicing of Theorem 2.8(1) of arXiv:2607.28411v1 modulo Harder-Narasimhan existence, and could go no further. Reading that paper's appendix established why, and it is a citation chain rather than a gap in the lane. Corollary A.23 carries assumption (iv) of Theorem A.17 -- Phi(F) is bounded coherent exactly when F is -- which fails for pi_* along an infinite field extension, since pi_* of the structure sheaf is not coherent; Example A.18(1) verifies (i)-(iv) only for a FINITE morphism, which is the intended scope and is the case the lane already owns. A.23's own HN step takes its phases and its termination measure from an HN filtration of the TARGET, which is the argument Phase/Transfer/HN.lean already implements, so there is no large-target HN argument in that appendix at all. And Theorem 2.8 is not proved there: page 12 cites it as a combination of [7, Theorem 5.3, Proposition 5.9, Theorem 12.17], where [7] is arXiv:1902.08184v4 -- this roadmap's own pin. Reading [7]: the slicing P_l is built from base-changed HEARTS by (12.3), not from the Ind-extension, and the finiteness that gives HN existence comes from noetherian hearts (Theorem 5.7(2)) plus, for a non-finitely-generated extension, the descent of Proposition 5.9(3) to a finitely generated subfield. That is why an Ind-extension route reaches the phase-indexed t-structures and Hom-vanishing but stops short of HN: largePreimage never mentions a heart. Milestones SF11 (sections 3-6, #1060-#1066) and SF12 (section 12, #1067) are the scoped path, eight items, four XL. The costly one is #1060: base change categories D_T do not exist in the tree and every result in section 5 is stated relative to them. Note the reusable half -- Lemma 5.1 of [7] is close to Lemma A.14 of 2607.28411, which SF7 (#475) already owns as TStructure.IndExtensionData, so #1062 reconciles presentations rather than rebuilding. Nearer-term alternative, not on this path: #722 and #723 close the Dqc realization and unblock projective-families-e5 and milestone PF2 directly." }
  - { rev: 56, at: "2026-09-11", note: "SF11.1 (#1060) begins with the source-faithful category and external-product layers. BaseChangeCategory.lean defines the compact part of a source Dqc component, the thick external-product envelope, its coproduct-and-extension closure (Dqc)_T, and the intrinsic bounded-coherent intersection D_T without imposing local Noetherianity on T. DqcLeftDerivedPullback refines the existing universal-property LeftDerivedPullback by the single quasicoherence-preservation obligation; the two projections from X times_S T then assemble (F,G) maps to Lphi'^*F tensor Lg'^*G from those honest lifts and an explicit Dqc tensor bifunctor. Still open: construction of that unbounded derived tensor, compact preservation and Proposition 3.15, strong-SOD extension and finite-amplitude Theorem 3.17, the induced pullback/pushforward functors, and Lemma 3.18 essential surjectivity." }
  - { rev: 57, at: "2026-09-11", note: "SF11.1 (#1060) reaches the compactness seam of Proposition 3.15. BaseChangeCategory.lean now names the exact preservation obligations on the honest Dqc left-derived pullbacks and supplied tensor bifunctor, proves every assembled external product Lphi'^*F tensor Lg'^*G compact, proves the resulting generator property lies in the compact objects of Dqc(X_T), and puts the full thick envelope there once compact objects are supplied as a thick triangulated property. This does not pretend to construct the missing unbounded derived tensor: an actual K-flat resolution/descent construction and the generic theorem that the repository's coproduct-defined compact objects form a thick subcategory remain open, as do the strong-SOD and functoriality parts of Theorem 3.17 and Lemma 3.18." }
  - { rev: 58, at: "2026-09-11", note: "SF11.1 (#1060) discharges the generic categorical half of the compactness seam. Triangulated/CompactObject.lean proves that shifts, zero objects, extensions in distinguished triangles (by the five lemma on compactness comparison maps), and retracts preserve the repository's coproduct-defined compactness; compactObjects is therefore registered as a thick triangulated property. The perfect external-product envelope in BaseChangeCategory.lean now lies in compact objects without extra typeclass hypotheses. The geometric half remains open: construct the actual unbounded Dqc derived tensor through K-flat resolutions and descent, then prove its compact-preservation theorem rather than supplying a tensor bifunctor and preservation witness." }
  - { rev: 59, at: "2026-09-12", note: "SF11.1 (#1060) Tranche C assembles the sequence-level base-change layer after the K-flat tensor, pullback, pushforward, adjunction, and component restrictions. The generic semiorthogonal root now distinguishes paper-strongness (right admissibility) from strong fullness, carries chosen right projections and cohomological amplitude, restricts sequences and projections along fully faithful inclusions, and propagates orthogonality through coproduct closures generated by compact objects. Families/BaseChangeSequence.lean constructs the perfect sequence on the compact-object model of Perf(X_T), the quasicoherent sequence, and its bounded-coherent inverse image from an explicit perfect-envelope semiorthogonality hypothesis. DecompositionData packages, without manufacturing, the remaining fullness, triangulatedness, projection, cocontinuity, and bounded-preservation inputs; BaseChangeSequenceFunctoriality.lean lifts pullback, pushforward, tensor linearity, and the projection formula from every component to the ordered sequences. The item remains in progress: the geometric theorems supplying perfect-envelope semiorthogonality and fullness, right adjoints with finite amplitude, cocontinuity, bounded-locus preservation, and the general hypotheses of Lemma 3.18 are still obligations rather than constructed witnesses." }
  - { rev: 60, at: "2026-09-12", note: "SF11.1 (#1060) Tranche D sharpens the geometric boundary of Proposition 3.15. Generic semiorthogonality of two shift-stable generator properties now extends formally to their triangulated envelopes, and adding shifts before taking the envelope is proved harmless. The K-flat family layer exposes the concrete external-product generators and their shift closure; PerfectExternalProductsSemiorthogonal states literal Hom vanishing between arbitrary shifts of earlier and later external products, is proved equivalent to shifted-generator orthogonality, and implies perfect-envelope semiorthogonality. Triangulatedness of the perfect, quasicoherent, and bounded target sequences is now derived from source triangulatedness (plus triangulatedness of the intrinsic bounded-coherent locus), rather than stored in decomposition data. The item remains in progress at genuinely geometric seams: prove the concrete external-product Hom vanishing and fullness, construct the right projections with cocontinuity and finite amplitude, prove bounded-locus preservation, and discharge the general detection hypotheses of Lemma 3.18." }
  - { rev: 61, at: "2026-09-12", note: "SF11.1 (#1060) Tranche E factors the concrete external-product Hom vanishing through paper-facing geometric interfaces. Tensor duality for compact fibre factors and the pullback-pushforward adjunction produce the Hom reduction; a first-projection formula reduces component membership to source tensor closure; compact Perf(S)-linearity and its presentable Dqc(S)-linear extension act through genuine derived pullback X -> S; and the latter handles arbitrary quasicoherent base coefficients, avoiding an unjustified compactness assumption on T -> S pushforward. The X -> S pullback and all three tensors are pinned to explicit K-flat resolutions. DqcFaithfulBaseChange names the derived base-change isomorphism Lg* Rphi* ~= Rphi'* Lg'*, and the final theorem uses it on the explicit fibre coefficient (Gj tensor Gi-dual)[b-a]. The item remains in progress: construct the compact-duality and monoidal coherence witnesses, prove the projection formula and faithful-base-change isomorphism from the stated scheme hypotheses, derive the presentable linearity/fullness statements, construct cocontinuous finite-amplitude projections, preserve the bounded locus, and discharge Lemma 3.18 detection." }
  - { rev: 62, at: "2026-09-12", note: "SF11.1 (#1060) Tranches F and G assemble the formal core and projection pipeline of Proposition 3.15. Atomic tensor/coefficient comparisons, faithful base change, presentable Dqc(S)-linearity, and external-product generation feed Proposition315CoreData, yielding perfect and quasicoherent semiorthogonality and fullness from explicit geometric witnesses. Every t-structure aisle now has canonical chosen right-projection data; a compact-generator approximation therefore projects onto Coprod(G), with cocontinuity transferred from zero truncation. Componentwise approximations assemble the quasicoherent projectors, and compact-preserving projectors restrict to the perfect category after an explicit compact-intersection identification. Proposition315ProjectionConstructionData feeds these directly into DecompositionData. The item remains in progress at geometric existence seams: construct the compact-duality/coherence, projection-formula, faithful-base-change, generation, approximation, truncation-cocontinuity, compact-intersection and compact-preservation witnesses; then prove finite cohomological amplitude, bounded-locus preservation, and Lemma 3.18 detection." }
  - { rev: 63, at: "2026-09-13", note: "SF11.1 (#1060) Tranche H discharges the boundedness half of the bounded-coherent restriction and isolates the half that is geometric. The generic layer adds finite cohomological amplitude for a functor and for chosen projection data, with the interval existentially quantified and allowed to depend on the component; amplitudes compose, a t-exact functor has finite amplitude, and a functor of finite amplitude carries the t-bounded locus into itself. Chosen projections of finite amplitude therefore preserve that locus, restrict along its inclusion, and leave the restricted sequence strong. The geometric layer splits Dqc(X)'s intrinsic bounded-coherent locus into ambient boundedness and finite presentation of every cohomology sheaf, so the single preservation obligation of DecompositionData.BoundedData splits in two: boundedness follows formally from finite amplitude measured against a supplied t-structure on Dqc(X_T) whose bounded objects are the ambiently bounded ones, and coherence preservation stays an explicit input because no amplitude bound implies it. Proposition315CoreData now reaches BoundedData and strongness of the bounded-coherent sequence. The item remains in progress: supply the presenting t-structure and prove finite amplitude for the constructed projectors, prove coherence preservation and bounded-locus generation, construct the outstanding Tranche F/G witnesses, and discharge Lemma 3.18 detection. Base-changed t-structures are deliberately out of scope here and belong to #1062." }
  - { rev: 64, at: "2026-09-14", note: "SF11.1 (#1060) Tranche I builds the source sequence that every earlier tranche assumed. Closing each component of a semiorthogonal sequence of compact objects under coproducts and extensions is now a sequence-level constructor: semiorthogonality survives because compactness carries Hom-vanishing across the adjoined coproducts, and the closure's components are triangulated whenever the compact ones are, replete unconditionally, and closed under coproducts in the indexing universe unconditionally. Those are exactly the three standing hypotheses -- HasTriangulatedComponents, IsClosedUnderIsomorphisms, IsClosedUnderColimitsOfShape -- that tranches E through H carried on their source sequence A, and they are now theorems about the sequence the paper actually uses. On the geometric side the quasicoherent source sequence 𝒟qc is that closure of the compact (perfect) source components, which BaseChangeCategory.lean had named as a remaining layer of Theorem 3.17. The converse stays geometric and is not proved: a compact component lies in the perfect part of its own closure, but recovering the whole perfect part is compact generation, and the repository keeps that as the uninhabited comparison Dqc.PerfectObjectsAreCompactInDqc. The item remains in progress: the Tranche F/G existence witnesses, tranche H's presenting t-structure, amplitude and coherence preservation, compact generation on the source, and Lemma 3.18 detection." }
  - { rev: 65, at: "2026-09-14", note: "SF11.1 (#1060) Tranche J closes the categorical half of Lemma 3.18 and retires the last stale claim about Theorem 3.17. ObjectProperty.preimageLift now carries Full and Faithful, transported from Mathlib's instances on lift through the full faithfulness of a full-subcategory inclusion, so full faithfulness restricts along a detected property for free; combined with the essential surjectivity already proved from the detection, preimageLiftEquivalence turns an ambient equivalence into an equivalence of the two selected subcategories. The base-change layer instantiates that twice: the restricted quasicoherent and bounded pullbacks inherit full faithfulness from the ambient Dqc pullback, and an ambient equivalence restricts to (Dqc)_U ≌ (Dqc)_T and D_U ≌ D_T. The hypothesis stays on the ambient functor throughout -- nothing here proves that any base-change pullback is an equivalence, which remains the geometric content of 3.18. BaseChangeCategory.lean's docstring named three remaining layers of Theorem 3.17 since tranche A: the unbounded derived tensor, the source Dqc component, and pullback/pushforward functoriality. All three have been built (tranches B, I and C respectively), so the paragraph now records that what remains of 3.17 is inhabitation rather than construction -- the compactness preservation statements and compact generation on the source are still supplied, not proved." }
  - { rev: 66, at: "2026-09-14", note: "SF11.2 (#1061) opens with the categorical core of S-locality. A t-structure carries two object properties, but they determine each other: the connective half is the right orthogonal of the coconnective one, by Mathlib isGE_iff_orthogonal. So TStructure.ge_eq_of_le_eq derives one from the other and TStructure.ext_le concludes that a t-structure is determined by its aisle -- which is exactly the engine Remark 4.6(1) of arXiv:1902.08184 runs on, since pinning the aisle of tau_U pins tau_U. TStructure.Restriction is one clause of S-locality at one functor: a t-structure on the target making the functor t-exact, with RestrictsAlong as its propositional shadow and the identity as the trivial inhabitant. The quantifier over the quasi-compact opens of the base is deliberately NOT taken here, because the categories D_U vary with U and the #1060 base-change layer owns them; that is the next slice, and it can already reuse Families/OpenImmersionPullback.lean, which proves pullback along an open immersion exact and essentially surjective on derived fibres. Nothing here constructs a t-structure: Restriction is inhabited only by someone who already has tau_U, and the uniqueness statement says which tau_U it must be, not that one exists. Phase/Transfer/Inducing.lean still records S-locality as outside the categorical layer; retiring that sentence needs the geometric quantifier." }
  - { rev: 67, at: "2026-09-14", note: "SF11.2 (#1061) takes the quantifier. Three facts made it cheap: the source category is itself a base-change category, namely the one at identityBaseChange S, so restriction is an ordinary base-change pullback and no separate source is needed; an open of the base is already an object of Over S, which opensBaseChange names without constructing anything; and base change is already functorial in the index through quasicoherentPullback. What is not cheap is the base-change data over each open -- the repository consumes bundled K-flat resolutions rather than a bare morphism -- so OpenRestrictionFamily carries a family of them together with the restriction functors, and IsSLocal quantifies TStructure.RestrictsAlong over the quasi-compact opens. SLocalData is the chosen form, and SLocalData.tStructure_eq_of_le_eq is Remark 4.6(1) with its geometric half visible: TStructure.ext_le settles uniqueness once the aisle of the restricted t-structure is pinned, and pinning it is a generation theorem about base change that is not proved. Nothing is inhabited: neither the family, nor the restricted t-structures, nor the definition. The two inhabitants #1061 asks for -- Remark A.20 over an affine base is the cheap one -- and Lemma 4.15, Lemma 4.16(3) and the slicing analogue all remain. Phase/Transfer/Inducing.lean still calls S-locality outside the categorical layer; that sentence is now false in letter but its retirement waits on an inhabitant." }
  - { rev: 68, at: "2026-09-15", note: "SF8.0 (#528) closes the coproduct clause of its open list, on both sides of the affine realization. Quasi-coherent sheaves on an ARBITRARY scheme now have their own i-indexed coproducts, created by the inclusion into X.Modules, so a coproduct formed among them is the ambient one; no hypothesis on X is used and the subcategory is still not claimed abelian. Exactness of those coproducts is separate and affine-only: AB4 for AffineQuasicoherentSheaves R is transported across the tilde equivalence from ModuleCat R, which is the only place an equivalence enters, and it is what DerivedCategory.hasCoproductsOfShape asks for beyond the coproducts themselves. A new generic theorem in the Mathlib-owned ExactFunctor directory -- an exact, coproduct-preserving functor between AB4-at-that-shape abelian categories has a mapDerivedCategory that preserves coproducts, by the localization square and essential surjectivity of Q -- then carries this to the exact derived inclusion, and reflection of coproducts by the honest Dqc(X) inclusion (creation, from the #721 closure) carries it to affineQuasicoherentDerivedToDqc itself. Preservation of coproducts is a standing hypothesis of compact-generation arguments and is not one: unbounded essential surjectivity remains uninhabited (AffineQuasicoherentDqcIdentification), and the bounded-coherent equivalence (#722, #1069--#1071), compact-equals-perfect, generator closure and S-linearity (#723) are untouched." }
  - { rev: 69, at: "2026-09-19", note: "SF1 closes. The 2026-08-14 closeout had folded two unrelated defects in IsPhaseTiltTypeTwo into one finding and withheld promotion on both. Re-reading v4 pages 71--75 separates them and they resolve opposite ways. The phase-free field was a real over-strengthening and is now a theorem: phaseFree_of_tiltHeart_triangle derives it from the phase interval alone, needing neither the semistability of U nor the zero charge of V, only that both lie in the original heart, so the field is gone from the predicate and the constructive direction derives it (#1383, squash 2d66007c). The positive-imaginary Hom implication is the opposite case: the printed case (2) is FALSE without it, since a split U[1] + V with V a nonzero zero-charge object satisfies every other clause and is destabilised by V, whose slope is +infinity under Definition 14.2. The printed moreover therefore belongs to the right-hand side of the biconditional and the Lean field is a faithful minimal rendering of it -- it carries only the positive-imaginary disjunct, not the Z-stable one. Owner accepted that reading on 2026-09-19. sec-14-weak-stability-tilting moves mapped -> reviewed on the dated review document; it is NOT formalized, evidence.owner_accepted is unset and nothing downstream may treat §14 as covered. §14.3 base change stays out of scope. Also recorded: v4 omits the proof of Lemma 14.17, citing [PT19, Lemma 2.19], so any further obligation is sourced from Piyaratne--Toda, not from this paper." }
  - { rev: 70, at: "2026-09-28", note: "SF9 re-scoped to the stability consumers of the stacks-and-moduli plan (stacks-and-moduli.yaml, ADR-0014 draft on #1624). Milestone #30 is renamed SF9 — Moduli of σ-semistable objects (BLMNPS §§9, 21); #516, #522, #525 and #524 are rewritten as consumers of the general STK/GM/MOD milestones and gain native blocked-by links, mirrored here; SF9.5–SF9.8 (#1757, #1758, #1774, #1759) join milestone #30 as milestone-direct tasks listed in stacks-and-moduli.yaml. General stack, descent and good-moduli theory now lives in STK1–STK4, ASP1, GM1–GM2 and MOD1–MOD2, not in this lane." }
retired: []

# GitHub milestones are roadmap phases. Epic issues live inside them, and
# gh_sub_issues records native GitHub sub-issue relationships. IDs are
# write-once; retire an item instead of reusing its id.
items:
  - id: stability-families-m0
    gh_milestone: 14
    kind: milestone
    title: "SF0 — Stabilize and recover"
    status: done
    lane: now
    acceptance:
      - "Current trust prose agrees with merged HRS, wall, support, and census declarations."
      - "The HRS tilt has an independent, commit-pinned mathematical review."
      - "The paper target is pinned as 1902.08184v4 without altering the Bridgeland 2007 source record."
      - "The mass-triangle chain is recovered as reviewed slices, merged in dependency order, and the global proposition is inhabited on main."

  - id: stability-families-e1
    gh_issue: 190
    gh_sub_issues: [196, 197, 198]
    kind: epic
    parent: stability-families-m0
    title: "Stabilize the current abstract foundation"
    status: done
    summary: "Destination issues #196, #197, and #198 are closed; the independent HRS review, aisle-agreement theorems, and versioned zero-claim paper map were transferred with the merged history."
    priority: must
    size: M
    tags: [trust, review, source-pin]
    links:
      source_issues: [6, 44, 51, 60, 94]

  - id: stability-families-e2
    gh_issue: 191
    gh_sub_issues: [199, 200, 201]
    kind: epic
    parent: stability-families-m0
    title: "Recover the mass-triangle chain and uncondition the topology"
    status: done
    summary: "The destination epic is closed. Issues #199 and #200 are closed; #201 remains a separately tracked follow-up. Remaining source-branch closeout is represented by destination migration PRs #218 and #219."
    priority: must
    size: XL
    tags: [lean, hn, topology, recovery]
    links:
      source_branch: ["archive/bridgeland-stab-lean/remote/agent/hn-mass-triangle-chain@716c0d1"]
      pull_requests: [218, 219]

  - id: stability-families-m1
    gh_milestone: 17
    kind: milestone
    title: "SF1 — Weak stability and tilting (§14)"
    status: in_progress
    lane: now
    acceptance:
      - "Weak stability, zero-charge subcategories, and the tilting property are formalized."
      - "The current HRS heart is connected to the textbook extension-closure formulation under explicit hypotheses."
      - "Proposition 14.16 and Lemma 14.17 have reviewed v4 source bindings."

  - id: stability-families-e3
    gh_issue: 192
    gh_sub_issues: [204, 205, 206, 207, 208, 209]
    kind: epic
    parent: stability-families-m1
    title: "Formalize weak stability and HRS tilting from §14"
    status: done
    summary: "The §14 program is decomposed into destination issues #204–#209. All six children are closed. The 2026-08-14 source-faithfulness audit reopened #208 over two fields of the Lemma 14.17 type-two predicate; the phase-cut field was derived away in #1383 and the Hom implication was accepted as load-bearing on 2026-09-19, and sec-14-weak-stability-tilting is reviewed."
    priority: must
    size: XL
    tags: [stability, tilting, paper-spine]
    links:
      issues: [204, 205, 206, 207, 208, 209]

  - id: stability-families-m2
    gh_milestone: 18
    kind: milestone
    title: "SF2 — Uniform support and families (§§18–23)"
    status: done
    lane: now
    acceptance:
      - "A genuine quadratic form controls semistable classes and is packaged uniformly over an indexed family."
      - "Definitions 20.5 and 21.15 are decomposed into separately auditable fields."
      - "Logical constant-family models are quarantined under Development and are not exposed as stable mathematical constructors."
      - "Theorem 22.2 remains absent from Lean; its geometric and moduli prerequisites are tracked as explicitly non-mathematical roadmap metadata."

  - id: stability-families-e4
    gh_issue: 193
    gh_sub_issues: [380, 387, 389]
    kind: epic
    parent: stability-families-m2
    title: "Upgrade to uniform quadratic support (§§18, 21)"
    status: done
    summary: "Closed after children #380, #387, and #389 established genuine quadratic support, the integral-to-real quotient comparison, and one common charge and form across varying abstract fiber categories. Geometric realization is tracked in SF4."
    priority: must
    size: L
    tags: [support, quadratic-form, uniformity]

  - id: stability-families-e5
    gh_issue: 194
    gh_sub_issues: [391, 393, 395, 397, 399, 401, 403, 405, 407, 414, 417, 425, 429]
    kind: epic
    parent: stability-families-m2
    title: "Define nonvacuous stability-in-families interfaces (§§20–23)"
    status: done
    summary: "Closed after children #391--#429 built the nonvacuous ordinary-family interfaces, categorical and scheme base change, residue and derived fibers, coherent exact derived pullback, and the theorem that flat scheme pullback is exact. General nonexact left-derived pullback, bounded coherent or perfect subcategories, geometric preimage/HN witnesses, geometric openness, relative HN structures, and bounded moduli remain explicit SF4/SF3 follow-ons."
    priority: should
    size: XL
    tags: [families, relative-hn]

  - id: stability-families-m3
    gh_milestone: 16
    kind: milestone
    title: "SF3 — Geometry and moduli prerequisites"
    status: done
    lane: now
    acceptance:
      - "Bounded coherent and perfect derived fibers are genuine categories with residue fibers, a proved Perf-to-bounded relationship, and nonvacuous pullback restriction contracts."
      - "A geometric pre-stability base-change witness is constructed for the finite-type family class actually used downstream."
      - "Finite-type semistable-locus openness and relative-HN existence are theorems from geometric hypotheses, not caller-supplied conclusions."
      - "Quot, stacks, boundedness, and good-moduli ownership stays explicit before Theorems 22.2, 24.1, or Part VI are claimed."

  - id: stability-families-e6
    gh_issue: 195
    gh_sub_issues: [460, 461, 462]
    kind: epic
    parent: stability-families-m3
    title: "Coordinate geometric and moduli prerequisites across ownership boundaries"
    status: done
    summary: "Closed after #460 delivered bounded coherent/perfect fibers, #461 delivered the geometric pre-stability base-change bridge, and #462 derived finite-type Zariski/generic openness and objectwise relative-HN existence. ADR-0013 chooses monorepo ownership and transfers Quot, stack, boundedness, semistable-reduction, and good-moduli follow-ons to SF8/SF9."
    priority: must
    size: XL
    tags: [families, geometry, moduli]
    links:
      repositories: [derived-alg-geo-lean, mathlib4]
      decisions: [.claude/decisions/ADR-0013-moduli-infrastructure-ownership.md]
      follow_on_epics: [517, 516]

  - id: stability-families-m4
    gh_milestone: 26
    kind: milestone
    title: "SF4 — Geometric slicing and relative HN (§§18–23)"
    status: done
    lane: now
    acceptance:
      - "Preimage-slicing witnesses are coherent under identity and composition and usable by category-valued families."
      - "Actual scheme-derived pullback consumes an honest geometric inducing package with every unproved presentable/Ind hypothesis explicit."
      - "Ordinary and generic semistability loci are genuine Zariski loci, and openness is proved only from named geometric hypotheses."
      - "Relative HN structures over curve/Dedekind bases carry object-level filtration data and have a nonvacuous constant-family model."
      - "Bounded coherent/perfect fibers, Quot, stacks, moduli, and Theorem 22.2 remain separate until their hypotheses are proved."

  - id: stability-families-e11
    gh_issue: 441
    gh_sub_issues: [443, 444, 445]
    kind: epic
    parent: stability-families-m4
    title: "Construct geometric slicing and openness witnesses"
    status: done
    summary: "Children #443–#445 supply compositional preimage witnesses, attach explicit left-adjoint inducing premises to concrete exact/flat scheme-derived pullback without inhabiting the presentable/Ind theorem, and realize universal and generic openness on actual residue-fiber semistable loci in the Zariski topology under the named base-change witness."
    priority: must
    size: XL
    tags: [families, slicing, openness, geometry]

  - id: stability-families-e12
    gh_issue: 442
    gh_sub_issues: [446]
    kind: epic
    parent: stability-families-m4
    title: "Construct relative HN structures over curves"
    status: done
    summary: "Child #446 supplies object-level relative HN data over actual scheme base changes, specializes the existence interface to regular Noetherian curve bases, records ordinary HN restrictions to every residue fiber with length, phase, and factor compatibility, supports pullback through the explicit geometric slicing witness, and gives a nonempty constant-family model. Unconditional geometric existence remains separate."
    priority: must
    size: XL
    tags: [families, relative-hn, curves, geometry]

  - id: stability-families-m5
    gh_milestone: 29
    kind: milestone
    title: "SF5 — T-structure calculus"
    status: done
    lane: now
    acceptance:
      - "Truncation functors commute naturally with integer shifts and the comparison respects the truncation triangle."
      - "Heart cohomology is homological and exact on distinguished triangles in every degree."
      - "Boundedness, nondegeneracy, and left/right/two-sided t-exactness have reusable identity, composition, adjunction, and heart APIs."
      - "The false unrestricted converse from nondegenerate to bounded is corrected rather than asserted."

  - id: stability-families-e10
    gh_issue: 216
    gh_sub_issues: [127]
    kind: epic
    parent: stability-families-m5
    title: "Cohomology functors, bounded t-structures, and t-exactness"
    status: done
    summary: "Closed by declaration-level audit of the migrated t-structure and all-degree cohomology modules. The repository proves bounded implies nondegenerate; the converse needs the finite-cohomology condition stated in Remark A.3 and is deliberately not claimed."
    priority: must
    size: L
    tags: [t-structure, cohomology, exactness]

  - id: stability-families-m6
    gh_milestone: 28
    kind: milestone
    title: "SF6 — Slicing order and adjoint transport"
    status: done
    lane: now
    acceptance:
      - "Strict and weak slicing orders have all Lemma-3.13 characterizations, transitivity, shift, and autoequivalence equivariance."
      - "Bayer and cofiltration packages preserve the paper's quantifier order at the abstract categorical level."
      - "Preimage/pullback/pushforward constructions expose Hom and HN obligations honestly and preserve phases, orders, shifts, and equivalences."
      - "The §3.4 registry coordinate is mapped with direct pinned-PDF evidence and remains a no-coverage claim."

  - id: stability-families-e8
    gh_issue: 211
    gh_sub_issues: []
    kind: epic
    parent: stability-families-m6
    title: "Slicing order relations and the Bayer/cofiltration properties"
    status: done
    summary: "Strict/weak order, all four phase characterizations, structural laws, abstract Bayer properties, and categorical cofiltration data are implemented and audited. Scheme-specific consequences remain outside this epic."
    priority: must
    size: L
    tags: [slicing, order, bayer, cofiltration]

  - id: stability-families-e9
    gh_issue: 215
    gh_sub_issues: []
    kind: epic
    parent: stability-families-m6
    title: "Abstract pullback and pushforward of slicings along an adjoint pair"
    status: done
    summary: "Raw phase preimages, genuine PreimageData, identity/composition/equivalence transport, phase reflection, order functoriality, and equivariance are implemented. The Appendix-A theorem remains an explicit, uninhabited boundary assigned to SF7."
    priority: must
    size: L
    tags: [slicing, transfer, adjunction]

  - id: stability-families-m7
    gh_milestone: 27
    kind: milestone
    title: "SF7 — Polishchuk/Appendix-A inducing theorem"
    status: done
    lane: now
    acceptance:
      - "Formula (A.8)'s phase-indexed source t-structures produce genuine preimage slicings by a proved finite truncation argument."
      - "Compact generation, Ind-extension, restriction to bounded coherent objects, and the A.17 recognition formulas are owned without axioms."
      - "The scheme-derived realization inhabits every hypothesis used by the supported geometric cases."
      - "HasLeftAdjointInducingTheorem and all downstream global theorem parameters are deleted."

  - id: stability-families-e13
    gh_issue: 475
    gh_sub_issues: [478, 477, 476, 479]
    kind: epic
    parent: stability-families-m7
    title: "Own the Polishchuk/Appendix-A inducing theorem"
    status: done
    summary: "Closed after #478 and #477 proved the finite-truncation and compact-generation/A.17 spine, #476 constructed the honest Dqc locus and supported coherent/perfect seam, and #479 removed the global theorem parameter. The broader general-scheme Dqc identifications were not claimed and continue as #528 under SF8."
    priority: must
    size: XL
    tags: [slicing, t-structure, ind, geometry]

  - id: stability-families-e13-1
    gh_issue: 478
    gh_sub_issues: []
    kind: deliverable
    parent: stability-families-e13
    title: "Derive preimage slicings from phase-indexed induced t-structures"
    status: done
    summary: "Charge-free head/tail decomposition plus truncation-triangle uniqueness gives a finite induction on target HN length. InducedTStructures.preimageData constructs the complete honest transfer witness."
    priority: must
    size: L
    tags: [slicing, t-structure, hn]

  - id: stability-families-e13-2
    gh_issue: 477
    gh_sub_issues: []
    kind: deliverable
    parent: stability-families-e13
    title: "Compact generation, Ind-extension, and Polishchuk A.17"
    status: done
    summary: "Complete through #507, #510, and #512. The repository owns compact objects, Coprod, the mapping telescope, Brown approximation tower, A.13 and (A.2), the A.14 restriction theorem, adjoint compactness, categorical A.17 Steps 1--4, bounded restriction, and A.3/A.4. Scheme-specific Dqc realization remains isolated in #476."
    priority: must
    size: XL
    tags: [compact-generation, ind, t-structure]

  - id: stability-families-e13-3
    gh_issue: 476
    gh_sub_issues: []
    kind: deliverable
    parent: stability-families-e13
    title: "Realize Appendix A on scheme-derived categories"
    status: done
    summary: "Closed at the supported scheme boundary by #514 and #515: the honest Dqc locus, coherent/bounded-coherent/perfect inclusions, and concrete identity inducing model are implemented without inhabiting unsupported cases. General triangulated/coproduct inheritance, bounded-coherent equivalence, compact-equals-perfect, generator closure, and S-linearity continue in #528."
    priority: must
    size: XL
    tags: [schemes, derived-category, locality]

  - id: stability-families-e13-4
    gh_issue: 479
    gh_sub_issues: []
    kind: deliverable
    parent: stability-families-e13
    title: "Remove the legacy global inducing-theorem parameter"
    status: done
    summary: "The global theorem switch and bounded left-adjoint shadow are deleted. Downstream geometry consumes phase-indexed InducedTStructures and constructs PreimageData through the owned finite truncation theorem."
    priority: must
    size: M
    tags: [cleanup, slicing, families]

  - id: stability-families-m8
    gh_milestone: 31
    kind: milestone
    title: "SF8 — Relative perfect complexes, Quot, and boundedness"
    status: in_progress
    lane: now
    acceptance:
      - "Relative-perfect and universally-gluable objects form honest scheme-derived loci with explicit Tor and negative-Ext witnesses."
      - "The relative-perfect moduli problem is groupoid-valued and functorial under only proved base changes."
      - "Boundedness is witnessed by a finite-type parameter scheme and a covering family, not by a caller-selected proposition."
      - "Quot and bounded-complex parameter spaces are reusable algebraic-geometry subjects and make no stack or good-moduli claim."

  - id: stability-families-e14
    gh_issue: 517
    gh_sub_issues: [528, 521, 518, 519, 520, 554]
    kind: epic
    parent: stability-families-m8
    title: "Own relative-perfect moduli and boundedness"
    status: in_progress
    summary: "#521, #518, #519, and #520 deliver the relative-perfect object, groupoid-valued moduli, finite-type boundedness, and supported Quot/parameter-space layers. #528 now has an honest affine quasi-coherent abelian/derived substrate, but the general unbounded Dqc equivalence, compact-equals-perfect, generator closure, and S-linearity remain before SF8 can close."
    priority: must
    size: XL
    tags: [relative-perfect, moduli, boundedness, quot]

  - id: stability-families-e14-0
    gh_issue: 528
    gh_sub_issues: []
    kind: deliverable
    parent: stability-families-e14
    title: "Complete the general-scheme Dqc realization"
    status: in_progress
    summary: "Constructed the genuine affine quasi-coherent sheaf category as abelian via Mathlib's tilde equivalence, its triangulated derived category, derived tilde/Gamma functors, and cohomology comparisons. The coproduct clause is now closed on both sides: #721 gave the honest Dqc(X) locus its small coproducts on an arbitrary scheme, and the affine realization functor is now proved to preserve them, through arbitrary-scheme coproducts on the quasi-coherent subcategory created by the inclusion into X.Modules, affine-only AB4 transported across the tilde equivalence, and a generic theorem that an exact coproduct-preserving functor's mapDerivedCategory preserves coproducts. General unbounded essential surjectivity into the honest Dqc locus, the bounded-coherent equivalence (#722, split as #1069--#1071), compact-equals-perfect, generator closure, and S-linearity (#723) remain explicitly open, and no compact-generation statement is claimed."
    priority: must
    size: XL
    tags: [dqc, compact-generation, schemes]

  - id: stability-families-e14-1
    gh_issue: 521
    gh_sub_issues: []
    kind: deliverable
    parent: stability-families-e14
    title: "Define relative-perfect and universally-gluable scheme objects"
    status: done
    summary: "Delivered honest Dqc total objects, the locally Noetherian pseudo-coherent criterion, pointwise open bounded stalkwise-flat Tor models, explicit geometric flat-model fibers, negative self-Ext vanishing, isomorphism invariance, and a concrete zero model."
    priority: must
    size: L
    tags: [relative-perfect, ext, fibers]

  - id: stability-families-e14-2
    gh_issue: 518
    gh_sub_issues: []
    kind: deliverable
    parent: stability-families-e14
    title: "Construct the groupoid-valued relative-perfect moduli problem"
    status: done
    summary: "Constructed a coherent Cat-valued pseudofunctor whose fibers are groupoids equivalent to the actual universally-gluable relative-perfect cores, with natural comparison to exact derived pullback on objects and morphisms. Added isomorphism classes, geometric-point fibers, and a concrete nonempty identity model without claiming arbitrary pullback preservation, descent, or algebraicity."
    priority: must
    size: L
    tags: [moduli, groupoid, base-change]

  - id: stability-families-e14-3
    gh_issue: 519
    gh_sub_issues: []
    kind: deliverable
    parent: stability-families-e14
    title: "Formalize finite-type boundedness witnesses for semistable complexes"
    status: done
    summary: "Replaced abstract boundedness by a finite-type parameter-scheme witness with an admitted universal relative-perfect family, explicit geometric fibers, and coverage up to isomorphism. Added isomorphism invariance, monotonicity, the Definition 21.15(5) adapter, base-change availability for the supported zero case, and a concrete identity/zero witness. General nonzero residue pullback and semistable coverage remain explicit SF8.4 work."
    priority: must
    size: XL
    tags: [boundedness, finite-type, semistable]

  - id: stability-families-e14-4
    gh_issue: 520
    gh_sub_issues: []
    kind: deliverable
    parent: stability-families-e14
    title: "Build Quot/parameter-space realizations for bounded complexes"
    status: done
    summary: "Built epimorphic module-sheaf quotient presentations with a category, target functor, and pullback/base-change functor. Added a representability interface requiring a Yoneda isomorphism and actual finite-type parameter morphism. Proved the zero-quotient functor is represented by the identity scheme, maps to the zero relative-perfect moduli model, and inhabits SF8.3 boundedness. General nonzero Quot representability remains unclaimed."
    priority: must
    size: XL
    tags: [quot, parameter-space, bounded-complexes]

  - id: stability-families-e14-5
    gh_issue: 554
    gh_sub_issues: [1495]
    kind: deliverable
    parent: stability-families-e14
    blocked_by: [1495]
    title: "Arbitrary derived pullback on the universally-gluable relative-perfect locus"
    status: in_progress
    summary: "Seven of thirteen deliverables are done: functorial flat/K-flat resolutions, the arbitrary left-derived pullback interface and caller-free inhabitant, exact-pullback comparison, universally-gluable locus interface, relative big-Zariski moduli boundary, and a nonidentity nonflat supported example. The free-Yoneda bicomplex now has a direct-sum totalization with flat terms; generic left-resolution rows have a natural objectwise quasi-isomorphic augmentation and degree-zero homology agreement. The induced integer-indexed bicomplex augmentation is rowwise quasi-isomorphic and totalizes to a map into a resolution-degree-zero bicomplex. A generic homology-preserving-colimit theorem, with an exact-colimit specialization, now preserves componentwise quasi-isomorphisms of complex diagrams and supplies one input for a compatible filtered resolution tower. Generic K-flat closure now needs only degreewise homology preservation and preservation of the particular diagram colimit by tensoring in either slot; an exact-colimit corollary supplies the homology premise under shape-wide tensor preservation. The proof uses naturality of preserved-colimit comparisons. This supplies a second input for a filtered resolution tower. Sheafified tensor on X.Modules preserves colimits of shapes in the site universe in either slot, with no flatness or invertibility premise. A single-zero bicomplex has a total without ambient coproducts and a natural comparison on either axis in any preadditive category with a zero object; its signed flipped comparison now assembles into a natural isomorphism under ambient total-existence. The arbitrary-shape, arbitrary-image generic single-extension comparison now normalizes the mapped bicomplex before totalization under only zero morphisms and a zero object. The adjacent-column filtered-total construction is now generic over the coefficient category with weak support and local total/cone premises, and the canonical total-quasi-isomorphism owner reuses that comparison. The free-Yoneda bicomplex augmentation now reaches that flipped single-zero target with rowwise quasi-isomorphisms, and its total target is naturally the input with a constructed total augmentation. The canonical integer-tail inclusion and nested maps now live with Mathlib's embedding owner, upstream of filtered totals. The normalized retained-component comparison and its bicomplex specialization now share that owner, preserving existing specialized names and component laws. A generic adjacent-column short exact core and a derived-triangle proof now give the actual total-map quasi-isomorphism for bicomplex maps supported on a finite horizontal strip in any abelian category, with finite-support totals constructed internally; this does not establish an unbounded total quasi-isomorphism. The canonical increasing good-truncation tower of an input cochain complex is functorial and its inclusion cocone is a colimit under local homology assumptions. Its literal image cocone remains colimiting after a termwise zero-morphism-preserving functor, without a general colimit-preservation premise. The named free-Yoneda resolved bicomplex cocone follows from this mapped-input theorem by specialization, and the literal direct-sum total cocone is colimiting: the generic total functor preserves diagram colimits under degree-fiber coproducts. This is a colimit statement, not a total-augmentation quasi-isomorphism. For the literal signed total, the incoming differential at each direct-sum coordinate is now calculated, and a columnwise exactness premise gives one epi-refined correction that clears the selected coordinate while preserving all lower cleared coordinates; finite epi-refinement iteration and a generic finite-support projection identity now prove degree-local exactness and acyclicity of the literal total when each diagonal has finite nonzero support and the supported column entries are exact, with a uniform upper-rectangle corollary. This does not yet prove the totalized augmentation is a quasi-isomorphism. A signed chain isomorphism now identifies the literal total of a vertical column cone with the mapping cone of the literal total map, with source and target branch laws and only pointwise and total-degree biproduct hypotheses; this supplies a structural comparison. With four independent, degreewise horizontal support bounds on source and target, a columnwise quasi-isomorphism now induces a quasi-isomorphism of literal totals via the signed cone comparison and finite-diagonal acyclicity. The four bounds now hold for the normalized free-Yoneda augmentation of every strictly bounded-above input, making its literal total map and signed augmentation quasi-isomorphisms; every canonical good truncation is a specialization. The stagewise quasi-isomorphisms now pass through the good-truncation colimit to prove the free-Yoneda total augmentation is a quasi-isomorphism for every input. The increasing lower stupid-truncation tower now recovers any cochain complex as a colimit without an upper bound or ambient colimit hypothesis; K-flat consumers may impose an upper bound to make its stages finite-support. A generic bifunctor total-map criterion now takes finite literal term support in one input and columnwise quasi-isomorphisms only on that support interval, with a signed-flip variant; it is a scheme-free input toward finite K-flatness, not geometric tensor preservation. Ordinary tensoring by an identity-flat module sheaf now preserves homology and termwise quasi-isomorphisms in both slots, but this alone does not make an arbitrary flat-term complex K-flat. A finite-term-support complex whose supported sheaves are identity-flat is now K-flat for the literal scheme total tensor in both slots; this does not cover unbounded flat-term resolutions. The generic literal two-slot homological-complex total now preserves diagram-shape colimits in each varying slot when the corresponding partial bifunctors at the fixed complex terms do and the target has total-degree-fiber coproducts; cochain corollaries use countable coproducts. The literal scheme-module totalTensor now preserves site-universe and small-shape diagram colimits in either fixed-input slot; sequential colimits are a specialization. Every strictly bounded-above complex of identity-flat module sheaves is K-flat for totalTensor, by finite lower stupid tails and exact sequential colimits. The free-Yoneda total of every input is now K-flat: each good truncation is resolved afresh into a bounded-above flat-term total, and the resolved truncation tower passes K-flatness through exact sequential colimits in both tensor slots. Its functorial total and already-proved quasi-isomorphic augmentation inhabit SchemeKFlatResolution with no caller-supplied resolution data. Arbitrary scheme pullback now preserves quasi-isomorphisms between K-flat complexes, so every SchemeKFlatResolution is pullback-acyclic without a flatness or exactness hypothesis. The canonical free-Yoneda resolution now supplies a named left-derived pullback inhabitant for every supported morphism, without caller-supplied acyclicity. The new caller-free inhabitant inherits canonical agreement with existing exact and flat pullback through LeftDerivedPullback.exactComparison and the existing flat specialization; exact-pullback comparison was already counted among the seven completed deliverables. Ordinary scheme-module pullback now preserves quasi-coherence for every morphism by transporting local presentations; this is an underived foundation and does not prove quasi-coherent cohomology preservation by derived pullback. Exact and flat derived pullback now preserve quasi-coherent cohomology without a caller-supplied preservation witness, by commuting with homology. For arbitrary nonexact morphisms, a termwise quasi-coherent K-flat complex now has quasi-coherent homology and its canonical derived pullback has quasi-coherent cohomology by comparison with ordinary pullback on that representative. The termwise-homology argument is now a scheme-free object-property theorem for any abelian category with kernel, cokernel and isomorphism closure; the geometric Dqc theorem specializes that generic result. No theorem yet supplies such a representative for every Dqc object, so general preservation remains open. On any supplied right-tensor-acyclic input, an objectwise canonical isomorphism now identifies arbitrary derived pullback with ordinary pullback, with ambient K-flatness as a specialization. For every affine ring map, including nonflat maps, a sheafified input whose right tensor inverts quasi-isomorphisms has an objectwise comparison with sheafified termwise scalar extension; ambient K-flatness is a specialization. This does not show that an arbitrary module complex sheafifies to an ambient K-flat complex or that every Dqc object has such a representative. A generic additive Hom comparison through degree-zero t-structure truncation now requires only two explicit outer-Hom vanishings; on affine schemes, bounded-below Dqc objects satisfy them by the existing negative affine vanishing theorem. This identifies structure-sheaf Hom with Hom into the degree-zero truncation. A generic pointwise isomorphism now identifies the pure degree-zero truncation with Mathlib H⁰ embedded by the single functor. For any scheme, Hom from the structure sheaf into that pure truncation identifies additively with top sections of H⁰; the bounded-below affine Dqc specialization composes this with the existing Hom-to-truncation comparison. A bounded-below ambient right-derived affine global-sections functor and its universal unit now exist using Mathlib injective resolutions. A generic natural degree-zero homology comparison now commutes a zero-preserving functor with a nonnegative cochain complex when degree-zero homology exists and the outgoing kernel is preserved. The bounded-below right-derived unit now induces an H⁰ isomorphism on strictly nonnegative complexes when the input outgoing kernel and the outgoing kernels of its nonnegative injective resolutions are preserved; on degree-zero singles only their own resolution kernels are needed to identify H⁰ of the right-derived functor pointwise with the underived functor, with a finite-limit-preserving specialization. The affine bounded-below RΓ+ unit now induces an H⁰ isomorphism on every strictly nonnegative complex of module sheaves; derived affine H⁰ for such a complex is ordinary affine sections of its input H⁰, with no quasicoherence assumption. On degree-zero singles the unit gives a pointwise H⁰ comparison with ordinary affine sections for every module sheaf. A canonical natural H⁰ comparison on generic degree-zero singles now follows from the natural derived unit under uniform input-specific resolution-kernel preservation, with finite-limit preservation as a corollary; component agreement with the earlier pointwise iso is proved, including its finite-limit specialization. The affine RΓ+ H⁰ comparison on degree-zero singles is now natural for every affine module sheaf and agrees componentwise with the earlier pointwise comparison. A generic K-injective target comparison now identifies degree-zero Hom-complex classes additively with derived-category morphisms; this does not itself identify H⁰ of affine RΓ+ on an arbitrary Dqc object. A new arbitrary-integer single-source Hom-complex comparison identifies it with degreewise additive coyoneda, retaining incoming boundaries; its represented specialization needs only preservation of zero morphisms and identifies homology in every integer degree. The bounded-below right-derived functor is now proved additive through the injective model; on affine schemes, the identity of the structure sheaf gives a specific additive, target-natural evaluation map into H⁰ of derived affine sections. Its injective-model agreement and bijectivity are not yet proved. Identification for arbitrary bounded-below Dqc inputs, unbounded affine Dqc identification, and removal of the lower bound remain open. Remaining deliverables include preservation of quasi-coherent cohomology, pseudo-coherence, local finite Tor amplitude and universal gluability, unit/compositor/triangle/pentagon/shift coherence, and full verification. The pseudo-coherence theorem must state which scope choice in #554 it takes. Blocks #522."
    priority: must
    size: XL
    tags: [derived-pullback, relative-perfect, k-flat, moduli]

  - id: stability-families-e14-5a
    gh_issue: 1495
    gh_sub_issues: []
    kind: deliverable
    parent: stability-families-e14-5
    title: "Establish flat generators for scheme module sheaves"
    status: done
    summary: "Proved that the canonical open free-Yoneda module sheaves have regular stalks on the open and zero stalks outside, hence are stalkwise flat over the identity, and constructed the natural M.val.Elements-indexed coproduct epimorphism at the generic small-ringed-site owner with its X.Modules specialization. This prerequisite does not supply an unbounded K-flat resolution or arbitrary pullback acyclicity."
    priority: must
    size: L
    tags: [module-sheaves, flat-generators, derived-pullback]

  - id: stability-families-m9
    gh_milestone: 30
    kind: milestone
    title: "SF9 — Moduli of σ-semistable objects (BLMNPS §§9, 21)"
    status: in_progress
    lane: now
    acceptance:
      - "M_pug(𝒟/S) is open in Lieblich's stack, with bounded extensions (§9; #1757, #1758)."
      - "M^st_σ(v, φ), M_σ(v, φ) and P_σ(I; v) are open substacks (Def 21.11, Lemma 21.12; #522)."
      - "Langton–Maruyama semistable reduction for hearts over a Dedekind scheme (Thm 16.1, Cor 16.5; #525)."
      - "Quasi-properness (Lemma 21.22; #1774)."
      - "In characteristic 0, a good moduli space proper over S (Thm 21.24(3); #524), closing gaps F3 and F4 for M_σ(v)."
      - "Algebraicity of M_pug, AHLH existence and Quot representability enter only as typed hypotheses over genuine definitions; no general stack, descent or good-moduli theory is built in this milestone."

  - id: stability-families-e15
    gh_issue: 516
    gh_sub_issues: [523, 522, 525, 524]
    kind: epic
    parent: stability-families-m9
    blocked_by: [1624, 1628, 1629, 522, 524, 525, 1757, 1758, 1774, 1759]
    title: "SF9: Moduli of σ-semistable objects over a base (BLMNPS §§9, 21)"
    status: in_progress
    summary: "The consumer epic for BLMNPS §§9 and 21: moduli of σ-semistable objects over a base, stated on genuine stacks from STK/GM/MOD and proved from that general theory, not rebuilt inside the stability lane. Children: M_pug(𝒟/S) for a semiorthogonal component with bounded extensions (#1757, #1758); the open substacks M^st_σ(v, φ), M_σ(v, φ), P_σ(I; v) (#522); semistable reduction (#525); quasi-properness (#1774); the route to Θ-reductivity and S-completeness (#1759); the proper good moduli space (#524)."
    priority: must
    size: XL
    tags: [moduli, semistable-reduction, good-moduli, blmnps]

  - id: stability-families-e15-1
    gh_issue: 523
    gh_sub_issues: []
    kind: deliverable
    parent: stability-families-e15
    title: "Formalize stacks in groupoids and scheme descent"
    status: done
    summary: "Built a reusable StackInGroupoids wrapper over Mathlib's effective descent, explicit covering families and Čech descent equivalences, morphism descent, stack transport across fiber/descent equivalences, pullback-compatible representability interfaces, and nonconstant representable big-Zariski stacks from Yoneda subcanonicity."
    priority: must
    size: XL
    tags: [stacks, groupoids, descent]

  - id: stability-families-e15-2
    gh_issue: 522
    gh_sub_issues: []
    kind: deliverable
    parent: stability-families-e15
    blocked_by: [1770, 1634, 1687, 1735, 1740, 1757, 1772, 1628]
    title: "SF9.2: M^st_σ(v, φ), M_σ(v, φ) and P_σ(I; v) are cut out of M_pug(𝒟/S) by open properties (BLMNPS Def 21.11, Lemma 21.12)"
    status: in_progress
    summary: "BLMNPS Def 21.11 and Lemma 21.12, re-scoped: the three moduli functors are sub-pseudofunctors of #1735's M_pug cut out by local, pullback-stable properties (#1634), open in #1687's sense, and algebraic under #1740's typed algebraicity hypothesis. The earlier big-Zariski presentation layer is superseded by the natural notions of STK1–STK3 (#1626 renames it Pointwise…)."
    priority: must
    size: M
    tags: [algebraic-stack, relative-perfect, open-substack]

  - id: stability-families-e15-3
    gh_issue: 525
    gh_sub_issues: []
    kind: deliverable
    parent: stability-families-e15
    blocked_by: [1760, 1762, 1763, 1765, 1063, 1066, 1651, 1653, 1656, 1697, 1695]
    title: "SF9.3: Langton–Maruyama semistable reduction for hearts over a Dedekind scheme (BLMNPS Thm 16.1, Cor 16.5)"
    status: planned
    summary: "BLMNPS Thm 16.1 with Def 16.3, Remark 16.4 and Cor 16.5: semistable reduction for a heart over a Dedekind scheme C, on SF13's HN structures over C. Quasi-properness (Lemma 21.22) moved to #1774."
    priority: must
    size: L
    tags: [semistable-reduction, valuative]

  - id: stability-families-e15-4
    gh_issue: 524
    gh_sub_issues: []
    kind: deliverable
    parent: stability-families-e15
    blocked_by: [522, 1773, 1774, 1759, 1737, 1738, 1706, 1713, 1714, 1700, 1740]
    title: "SF9.4: M_σ(v, φ) has a good moduli space proper over an excellent base of characteristic 0 (BLMNPS Thm 21.24(3))"
    status: planned
    summary: "BLMNPS Thm 21.24(3) stated with #1713's HasProperGoodModuliSpace over #1706's IsGoodModuliSpace, from #1714's typed AHLH statement, the affine diagonal of M_σ(v, φ) (#1737, #1738) and the restricted Θ-reductivity and S-completeness (#1700, route #1759)."
    priority: must
    size: L
    tags: [good-moduli, properness]

  # ============ SF10 — the Fourier--Mukai adjoint substrate ============
  #
  # This milestone carries NO `gh_milestone`, deliberately. The lane's work is
  # already merged and its two open issues are unparented on the tracker, so
  # minting a GitHub milestone would create a roadmap-owned milestone whose
  # RM-06 obligation someone then has to maintain -- for a lane whose remaining
  # items are two issues. `numerical-k-theory.yaml` uses the same shape for the
  # same reason. If the lane grows sub-issues, mint the milestone then.
  #
  # WHY THIS FILE. The lane is split across `CategoryTheory/Triangulated/
  # FourierMukai/` and `AlgebraicGeometry/StabilityCondition/Families/`, and the
  # second is this roadmap's own territory: the geometric ledgers below sit
  # beside `KernelCorrespondence`, `KernelConvolution`, and `KernelUnit`, which
  # this track already owns implicitly. The abstract half has no track and did
  # not warrant minting one for two issues.
  - id: stability-families-m10
    kind: milestone
    title: "SF10 — Fourier--Mukai adjoints, the dual kernel, and the equivalence"
    status: in_progress
    lane: next
    acceptance:
      - "Both adjoints of a kernel transform are kernel-presented, abstractly and geometrically."
      - "A dual kernel is derived rather than supplied, from named contracts alone."
      - "The equivalence stops being supplied data."
      - "At least one contract in the adjoint ledger has an instance."

  - id: stability-families-e16
    gh_issue: 795
    gh_sub_issues: []
    kind: epic
    parent: stability-families-m10
    title: "Layer 3: derive the equivalence from the adjoint kernels, then find a criterion for it"
    status: planned
    summary: "Closing #559 left layer 3 undelivered: `KernelAutoequivalence` still takes its equivalence and comparison isomorphism as supplied data, and nothing proves any transform is an equivalence. The issue splits it in two, and the split is the useful part. Half one is nearly free and was verified against the pin rather than asserted -- a `RightAdjointKernelData` whose adjunction has pointwise-invertible unit and counit yields an equivalence through `Adjunction.toEquivalence` whose functor is the transform DEFINITIONALLY, so `iso` is `Iso.refl`; its consumer already exists, since `geometricKernelAutoequivalence` could then stop taking an equivalence at all. Half two is the actual theorem: nothing supplies that invertibility, and classically it is Bondal--Orlov plus Serre duality, neither of which this repository has. The ordering is load-bearing -- a class asserting invertibility filed before half one would be supplied data nothing consumes, which is the defect #559 existed to prevent."
    priority: must
    size: L
    tags: [lean, fourier-mukai, adjunction, equivalence]

  - id: stability-families-e17
    gh_issue: 796
    gh_sub_issues: []
    kind: epic
    parent: stability-families-m10
    title: "Discharge the Fourier--Mukai adjoint ledger: ten uninhabited classes, no witness"
    status: planned
    summary: "The #559 arc (#768, #770, #777, #779, #780) landed `geometricDualKernel` with dual kernel `Rsigma_*(K^v (x) omega_q)` -- the classical `P^v (x) p^* omega_X [dim X]` -- assembled entirely from named contracts. Every one of those contracts is uninhabited: six new (`HasDerivedPullbackAdjunction`, `HasKernelDual`, `HasTwistedInversePullback`, `HasDualizingTwist`, `HasPullbackSwap`, `HasPushforwardSwap`) plus four pre-existing (`HasCoherentPullback`, `HasDerivedPushforward`, `HasCoherentDerivedTensor`, `HasProjectionFormulaRight`). So the result is a conditional statement with no witness, and the gain is separability rather than existence: ten distinct classical theorems, dischargeable one at a time with the consumer already waiting. The issue orders them easiest-first -- the two swap classes are pure functoriality, Grothendieck duality and the relative dualizing complex are last -- and records the two contracts easy to miss, `HasDerivedPushforward` at the PULLBACK's morphism and `HasCoherentPullback` at the PUSHFORWARD's, which together are what make the reversed correspondence expressible."
    priority: should
    size: XL
    tags: [lean, geometry, fourier-mukai, ledger, inhabitance]

  - id: stability-families-m11
    gh_milestone: 53
    kind: milestone
    title: "SF11 — Base change of t-structures (§§3–6)"
    status: in_progress
    lane: next
    acceptance:
      - "Base change categories D_T exist as a construction, with their two functors as theorems about it."
      - "S-local t-structures are inside the formalized layer rather than named as excluded."
      - "Theorem 5.3 is proved on top of the Ind-extension the repository already owns, not a second copy of it."
      - "Theorem 5.7(2), noetherianity is inherited under base change, is proved: everything downstream rests on it."
      - "Proposition 5.9(3) holds for a not necessarily finitely generated extension."

  - id: stability-families-e18
    gh_issue: 1060
    gh_sub_issues: []
    kind: epic
    parent: stability-families-m11
    title: "Base change categories D_T for a morphism T to S (§3)"
    status: done
    summary: "The repository now constructs the K-flat external product, perfect envelope, quasicoherent coproduct closure, intrinsic bounded-coherent intersection, and the corresponding perfect/quasicoherent/bounded semiorthogonal sequences. Pullback, pushforward, adjunction, essential-surjectivity reductions, tensor linearity, and the projection formula are available componentwise and at sequence level. Proposition315CoreData assembles perfect/quasicoherent semiorthogonality and fullness from explicit geometric witnesses. Its projection pipeline constructs quasicoherent right projectors from compact-generator approximation t-structures, transfers cocontinuity from zero truncation, and restricts compact-preserving projectors to perfect components through an explicit compact-intersection equality. Finite cohomological amplitude now formally yields preservation of the t-bounded locus, restriction of the chosen projections to it, and strongness of the restricted sequence, and the bounded-coherent locus is split so that only its coherence half remains an input. The task 2.1--2.3 formal boundary is complete: Tensor duality, fullness/generation, projection/restriction, bounded coherence, finite amplitude, and component detection remain explicit geometry-owner inputs, while the conditional closure, restriction, comparison, and detection APIs are proved without hiding those obligations."
    priority: must
    size: XL
    tags: [lean, derived, base-change, prerequisite]

  - id: stability-families-e19
    gh_issue: 1061
    gh_sub_issues: []
    kind: epic
    parent: stability-families-m11
    title: "S-local t-structures and S-local slicings (§4)"
    status: done
    summary: "The S-local t-structure and slicing APIs are now formalized over the existing base-change family and phase ownership. The quantifier over quasi-compact opens, aisle-based uniqueness, theorem-backed noetherian and filtration-lifting adapters, affine witness data, and restriction/phase compatibility are available. The underlying base-change, generation, locality, and lifting geometry remains explicit owner input rather than a fabricated theorem."
    priority: must
    size: L
    tags: [lean, derived, t-structure, locality]

  - id: stability-families-e20
    gh_issue: 1062
    gh_sub_issues: []
    kind: epic
    parent: stability-families-m11
    title: "Theorem 5.3: the base-changed t-structure on (Dqc)_T"
    status: done
    summary: "The formal Ind-extension bridge for Theorem 5.3 is now present without a second carrier: IndExtensionData proves the inclusion t-exactness and exposes its coproduct-aisle comparison, while a PreservesColimit boundary records the filtered-colimit owner input. Families/Theorem53.lean adds source-shaped bounded-range descent and tensor/t-exactness owner data for (5.1), (5.2), (3), (4a)-(4d), and (5.3)/(5.4), with formal composition theorems. The scheme-level flat/fpqc, tensor, and base-change geometry remains explicit and is not claimed proved by inhabiting the boundary structures."
    priority: must
    size: XL
    tags: [lean, derived, base-change, ind-extension]

  - id: stability-families-e21
    gh_issue: 1063
    gh_sub_issues: [1496]
    kind: epic
    parent: stability-families-m11
    title: "Theorems 5.6, 5.7 and Corollary 5.8: the bounded coherent case"
    status: in_progress
    summary: "Theorem 5.6 induces a t-structure on D_T for a finite perfect morphism, as the analogue of Polishchuk 2.3.5 -- a sibling of the Polishchuk 2.1.2 result the repository owns through SF7 and #1054-#1057, and the relationship should be recorded rather than assumed. Theorem 5.7 is the main base change result for essentially perfect morphisms, and its clause (2), that noetherianity is inherited, is the load-bearing one: Step 2 of Theorem 12.17 gets finiteness of an a priori infinite filtration from it. Child #1496 addresses only the finite-cover Noetherian-heart detection step; bounded-coherent restriction and geometric local Noetherianity remain open. Corollary 5.8 specialises to a point of the base."
    priority: must
    size: XL
    tags: [lean, derived, base-change, noetherian]

  - id: stability-families-e21-1
    gh_issue: 1496
    gh_sub_issues: []
    kind: deliverable
    parent: stability-families-e21
    title: "Finite-cover Noetherian-heart detection"
    status: done
    summary: "Proved that triangulated t-exact restrictions preserve heart monomorphisms and derived the finite-family local-to-global Noetherian implication from the existing detector. Geometric conservativity, local Noetherianity on bounded-coherent D_T, and Theorem 5.7(2) remain under parent #1063."
    priority: must
    size: M
    tags: [lean, derived, t-structure, noetherian]

  - id: stability-families-e22
    gh_issue: 1064
    gh_sub_issues: []
    kind: epic
    parent: stability-families-m11
    title: "Proposition 5.9: base change of hearts along a field extension, with descent to a finitely generated subfield"
    status: planned
    summary: "For a not necessarily finitely generated extension k in l, the t-structure on (Dqc)_l induces a bounded one on D_l, base change is t-exact, and -- clause (3), the point of the item -- every object of the base-changed heart descends to a subfield finitely generated over k. That descent is what makes Theorem 12.17 work for arbitrary extensions, and its absence is exactly why Slicing.IndExtensions.largePreimage (#1057) can only carry HN existence as a hypothesis."
    priority: must
    size: L
    tags: [lean, derived, base-change, descent]

  - id: stability-families-e23
    gh_issue: 1065
    gh_sub_issues: []
    kind: epic
    parent: stability-families-m11
    title: "Tilted-noetherian t-structures, and their comparison with the existing noetherian hearts"
    status: planned
    summary: "The standing hypothesis of Proposition 5.9 and the conclusion of Theorem 5.7(1). The repository has NoetherianTorsionSubcategory and SubobjectChain.Terminates in Weak/Heart/Noetherian.lean, and a Happel-Reiten-Smalo tilt in progress under SF1 (#192), but nothing named tilted-noetherian. The comparison with the existing heart vocabulary must be a proved theorem and not a second noetherianity hierarchy, per the abstraction tree."
    priority: must
    size: M
    tags: [lean, stability, noetherian, tilting]

  - id: stability-families-e24
    gh_issue: 1066
    gh_sub_issues: []
    kind: epic
    parent: stability-families-m11
    title: "Flat, torsion and torsion-free objects over a base (§6)"
    status: planned
    summary: "Lemma 6.12, torsion-free implies flat, and Remark 6.16, both consumed by Step 2 of Theorem 12.17: by Lemma 12.12 and Theorem 5.7(2) the heart is noetherian, by Remark 6.16 the object may be assumed torsion free, and thus flat by Lemma 6.12. Forty-two files in the tree mention torsion-freeness, but they are numerical and sheaf-theoretic rather than the section 6 torsion theory relative to a base; any overlap must be a proved comparison."
    priority: must
    size: L
    tags: [lean, geometry, torsion, flatness]

  - id: stability-families-m12
    kind: milestone
    title: "SF12 — Stability conditions and base change (§12)"
    status: planned
    lane: later
    acceptance:
      - "Theorem 12.17(1)-(4) holds for an arbitrary, not necessarily finitely generated, field extension."
      - "The construction (12.3) from base-changed hearts is proved equal to Slicing.IndExtensions.largePreimagePhase."
      - "Slicing.IndExtensions.largePreimage loses its HN hypothesis, or the ledger records why it keeps it."

  - id: stability-families-e25
    gh_issue: 1067
    gh_sub_issues: []
    kind: epic
    parent: stability-families-m12
    title: "Theorem 12.17: stability conditions under an arbitrary field extension"
    status: planned
    summary: "This is Theorem 2.8 of arXiv:2607.28411v1, which that paper cites rather than proves, and the result the projective-families lane (#1033) has been building toward. The construction is via base-changed HEARTS: with A^phi the interval heart, P_l(phi) is the intersection of the A_l^phi' over the phases that cut it (12.3), from Proposition 5.9. The Ind-extension phrasing of Theorem 2.8(1) is a restatement, not the definition, which is why the lane's Ind-extension route reaches the phase-indexed t-structures and Hom-vanishing but not HN existence -- largePreimage never mentions a heart, and the finiteness lives there. Four steps: finite l by conservativity of the pushforward, which the lane already owns as semistable_iff_preimage_of_coproduct (#1055); l = k(x) with Z over Q[i] by noetherianity and discreteness; arbitrary l by the descent of Proposition 5.9(3); general Z by continuity and the GL2+(R) action. XL and expected to split after scoping, as #528 did."
    priority: must
    size: XL
    tags: [lean, stability, base-change, field-extension]

Exported from .claude/roadmap/stability-families.yaml at the commit that removed it from the tree.

Home

Architecture

CI and runners

Decision records

Roadmaps

Clone this wiki locally