docs: the summary is maintenance rules, then facts, then week-by-week

Three changes, all asked for by the maintainer:

The "where it stands" prose is gone.  It read as an assessment of the
project — feature scope covered, what is "left", waiting on decisions
rather than features — which is a judgement this file has no standing to
make.  The factual table stays: test tallies, documented members, visual
regression coverage, benchmark pairs, style parity, bundle size, CI.

The maintenance rules move to the top, where the next person editing the
file sees them before editing rather than after.  Two rules are new: the
weeks carry headline changes and *what they buy*, and no judgement calls
about project status — both of which this pass exists to enforce.

The timeline table becomes five week sections, each a list of what
changed and what it bought at a high level.  What was previously scattered
across "what v4 is", "performance" and a one-line-per-week table is now
in the week that produced it, with its benefit attached: the columnar
model buys the performance case, goldens plus parity diffs buy the answer
to "is it right", the GPU executors buy 642x on Markov clustering, the
concurrent benchmark runner buys 8.4 minutes instead of 55.

The rendered page now holds zero prose blocks — three tables and 62
points.

Claude-Session: https://claude.ai/code/session_01CZuBgFBGSDy6QtAYFL5sE3

Author: maxkfranz

EXECUTIVE_SUMMARY.md

@@ -3,26 +3,40 @@
 The v4 rewrite: a columnar model and a WebGPU renderer, per
 [#3486](https://github.com/cytoscape/cytoscape.js/issues/3486).
 
-- **Status**: not released, and not close. `cytoscape@3` remains the shipping
-  library.
+- **Status**: not released. `cytoscape@3` remains the shipping library.
 - **Scope of this record**: the v4 prototype, from **2026-07-22**.
 - **Last updated**: 2026-08-12, as round 68 closed.
-- **Derived document.** [`PLAN.md`](PLAN.md) is the development record and the
-  source of truth; this is the five-minute version — see *Maintaining this
-  file*.
+
+## How to maintain this file
+
+- **[`PLAN.md`](PLAN.md) is the source; this file is derived from it.** Nothing
+  is recorded here that is not recorded there first.
+- **Rewrite when a round closes** — by re-reading the record and restating it,
+  never by appending.
+- **Point form only.** Short points, high level, readable in five minutes.
+- **Week by week, and headline changes only.** Each week is a list of what
+  changed and **what it buys** — the benefit, not the implementation. A change
+  earns a line only if it changed what the library *is* or what it can do; most
+  do not.
+- **No round narrative.** Rounds, file names, sub-round numbering and
+  implementation detail belong in `PLAN.md`. Round numbers appear here only
+  where one is the sole handle on an open question.
+- **Restate, don't append.** Later rounds routinely change what an earlier
+  decision meant, so earlier weeks need correcting too.
+- **Facts, not judgement.** The numbers table below stays current; assessments
+  of how close a release is, what is "left", or whether the work is on track do
+  not belong here — they need information this file does not carry.
+- **Quote a number only if it is current.** Test tallies, member counts and
+  benchmark figures go stale first; re-run the relevant command (see
+  `AGENTS.md`) rather than copying one forward.
+- **Keep the decisions table honest**: an item leaves it when the decision is
+  made, not when the work is scheduled.
+- Update *Last updated* and the round it covers.
 
 ---
 
 ## Where it stands
 
-- Feature scope covered; the work is **hardening, measurement and release
-  preparation**.
-- CI green as of 2026-08-06, having been red on every push for weeks.
-- `npm test` passes from a clean checkout.
-- What is left waits on **a decision, another platform, or release
-  credentials** — not on unbuilt features.
-- An inventory, not a schedule: several rounds v4 needs are not scoped yet.
-
 | | |
 |---|---|
 | Automated tests | 2,205 unit · 427 module · 24 soak · 376 browser (251 run; 125 skip for want of a WebGPU adapter) |
@@ -31,31 +45,80 @@ The v4 rewrite: a columnar model and a WebGPU renderer, per
 | Benchmarks | 25 suites, 4 published profiles · **all 366 v3-comparative pairs read v4-faster** (geometric mean 13.7×, minimum 1.03×) · GPU algorithm executors 13× geo-mean over their CPU reference |
 | Style parity | v4 accepts 157 of v3's 291 style property names; the rest dropped by decision |
 | Bundle | 691 KiB minified / 185 KiB gzipped — ~1.5× v3 (410 / 126 KiB); the WGSL shaders, which v3 has no equivalent of, are minified at build time |
-
-**Headline case** — a 19,607-node / 464,657-edge network:
-
-- initialises in **~0.95 s against v3's ~19 s**;
-- holds **33 ms frames where v3 takes 4,460 ms**;
-- lays out live on the GPU force layout in **1.3 s**.
+| CI | Green as of 2026-08-06; `npm test` passes from a clean checkout |
 
 ---
 
-## What v4 is
+## Week 1 — 22–24 July
 
 - **A columnar, CPU-canonical model** — typed-array columns, stable slots,
-  coalesced dirty spans, a CSR adjacency index.
-- Reads stay synchronous, so the public API did not become async.
-- **A WebGPU renderer** — SDF node shapes, compute culling, indirect draws, GPU
-  picking, GPU text.
-- **A co-signed model↔renderer contract** in one file: changing the column and
-  flag layout is a deliberate act.
-- **Structured queries instead of a selector language** — objects and plain
-  functions.
-- **A serializable mapper DSL** for style, evaluated on the GPU for paint
-  channels.
-- Element state (`:selected`, `:active`, `:grabbed`, …) is ordinary style
-  condition, not a shader constant — so an application can restyle or disable
-  every affordance.
+  coalesced dirty spans, a CSR adjacency index. Buys the whole performance
+  case: the graph is already in the shape the GPU wants.
+- **Reads stay synchronous.** Buys a public API that did not have to become
+  async to get the model.
+- **A WebGPU renderer** — SDF shapes, compute culling, indirect draws, GPU
+  picking, GPU text. Buys interaction on graphs that stall v3: **33 ms frames
+  where v3 takes 4,460**.
+- **A co-signed model↔renderer contract in one file.** Buys a layout change
+  that cannot happen by accident on one side only.
+- **Structured queries instead of a selector language.** Buys queries that are
+  data — composable, inspectable, no parser.
+- **A serializable mapper DSL for style.** Buys paint channels evaluated on the
+  GPU, and style that can cross a wire.
+
+## Week 2 — 27 July – 2 August
+
+- **Visible-behaviour parity with v3** — curves, compounds, labels, images,
+  animation. Buys an existing app the same picture.
+- **Goldens *plus* live v3-vs-v4 parity diffs.** Buys the answer to *is it
+  right*, which a golden alone cannot give: goldens compare v4 against its own
+  past.
+- **Benchmarks with an HTML report.** Buys claims that are numbers.
+- **Design sittings, decisions recorded when taken.** Buys a migration guide
+  that could later be compiled rather than reconstructed.
+
+## Week 3 — 3–8 August
+
+- **Error and documentation contracts gated at zero tolerance.** Buys a guard
+  no test has ever fired failing the build, and a public member without docs
+  failing it too.
+- **The repository split: v4 is the package, v3 lives whole in `v3/`.** Buys a
+  v4 that ships alone, and a real v3 to keep measuring parity against.
+- **Packaging gate, migration guide, soak tier, status site.** Buys release
+  mechanics that are testable before there is a release.
+- **CI made honest.** Buys a green run that means something: every defect found
+  was in something never executed on a fresh checkout, on a runner, or in a
+  browser nobody could launch locally.
+
+## Week 4 — 9–11 August
+
+- **Both long-standing API decisions taken, each by declining the surface its
+  own homework had priced.** Buys a smaller library and a recorded reason.
+- **The force layout rebuilt on published methods.** Buys a layout that stays
+  numerically stable past node degree ~20, and lays out the 465k-edge network
+  live in **1.3 s**.
+- **The nine expensive whole-graph algorithms went async with GPU executors.**
+  Buys Markov clustering at up to **642×** its CPU reference, k-medoids 146×,
+  Floyd–Warshall 28×; the traversal tier stays synchronous.
+- **Per-element bypasses returned as a first-class stylesheet section.** Buys
+  v3's ergonomics without v3's cost — id-keyed, surviving re-add, carried by
+  `cy.json()` (v3 loses them), twice as fast.
+- **Cross-commit benchmark comparison pages.** Buys regressions caught by the
+  archive rather than by a user: one four-day-old regression on first use.
+- **The load path measured, then taken apart.** Buys **1.9×** on the 465k-edge
+  load with a byte-identical rendered frame, and initialisation of that network
+  in **~0.95 s against v3's ~19 s**.
+
+## Week 5 — 12 August
+
+- **The benchmark runner runs jobs concurrently.** Buys a publishing run of the
+  `all` profile in **8.4 minutes instead of ~55**, so re-measuring stops being
+  a reason not to measure.
+- **Concurrency folded into the harness fingerprint, and the published archive
+  kept serial.** Buys the guarantee that a faster *instrument* can never be
+  read as a faster *library*.
+
+---
 
 ## What changed for users of v3
 
@@ -67,45 +130,13 @@ The v4 rewrite: a columnar model and a WebGPU renderer, per
   sequencing.
 - **display/visibility split** into a structural tier and a paint-only tier.
 - **`cy.layout({ impl })` is the whole extension story.**
-- **The nine expensive whole-graph algorithms return promises** and take
-  `executor: 'cpu' | 'gpu' | 'auto'`; the traversal tier stays synchronous.
-- **Per-element bypasses are back** — a first-class `bypasses` stylesheet
-  section, id-keyed, surviving re-add, carried by `cy.json()` (v3 loses them),
-  twice as fast as v3's.
+- **The expensive whole-graph algorithms return promises** and take
+  `executor: 'cpu' | 'gpu' | 'auto'`.
 - **`cy.collection()` throws if passed an argument**; **`cy.$()` and
   `cy.byId()`** restored as aliases.
-
-## Performance
-
-- Every v3-comparative benchmark pair reads v4-faster.
-- The published archive and its cross-commit comparison pages keep it that way
-  — they caught a four-day-old regression on first use.
-- **GPU algorithm executors** on a real adapter: Markov clustering up to 642×
-  its CPU reference, k-medoids 146×, Floyd–Warshall 28×, betweenness 18×; only
-  PageRank and hierarchical clustering sit in low single digits.
-- **The load path**, rebuilt on measurement: **1.9×** on the 465k-edge load,
-  rendered frame byte-identical.
-- **First frame**: deferring the feature pipelines cut the software-renderer
-  first frame from 4.6 s to 2.7 s.
-- **The benchmark runner** runs jobs concurrently: an `all`-profile publishing
-  run went from ~55 min to **8.4**.
-- Concurrency is treated as a change of instrument, not a faster clock: it is
-  folded into the harness fingerprint, and **the published archive stays
-  serial**.
-
----
-
-## Timeline
-
-| Week | |
-|---|---|
-| **1** — 22–24 Jul | Architecture proved end to end. |
-| **2** — 27 Jul – 2 Aug | Visible-behaviour parity with v3, and the measurement infrastructure to keep it. |
-| **3** — 3–8 Aug | Contracts gated; the repository split so v4 is the package; packaging, migration guide, soak tier, status site; CI made honest. |
-| **4** — 9–11 Aug | Both long-standing decisions taken; the force layout rebuilt; the expensive algorithms async onto the GPU; bypasses back; the load path taken apart. |
-| **5** — 12 Aug | The benchmark runner stopped using one core of eight. |
-
----
+- **Element state** (`:selected`, `:active`, `:grabbed`, …) is ordinary style
+  condition rather than a shader constant, so an application can restyle or
+  disable every affordance.
 
 ## Open decisions
 
@@ -115,7 +146,7 @@ The v4 rewrite: a columnar model and a WebGPU renderer, per
 | Edge overlay band width | v3 draws the halo `2 × padding` wide (invisible at small paddings), v4 `width + 2 × padding` (always visible). Either resolution changes rendered output |
 | Hollow *mid* arrows | Still show the line through them: they sit mid-edge, where a trim cannot reach. May end up unsupported rather than fixed |
 
-## What remains before 4.0
+## Not yet built
 
 | | |
 |---|---|
@@ -128,8 +159,6 @@ The v4 rewrite: a columnar model and a WebGPU renderer, per
   through the declarative mapper system; splitting the largest implementation
   files.
 
----
-
 ## How this project works
 
 1. **A control for every claim** — each test run once with the behaviour it
@@ -141,24 +170,3 @@ The v4 rewrite: a columnar model and a WebGPU renderer, per
    browser, a person opening the page.
 5. **A change that should be invisible is the best test of the tools that watch
    for changes.**
-
----
-
-## Maintaining this file
-
-- **`PLAN.md` is the source; this file is derived from it.**
-- **Rewrite when a round closes** — by re-reading the record and restating it,
-  never by appending.
-- **Point form only**: short points, high level, readable in five minutes.
-- **No round narrative.** Rounds, file names and implementation detail belong
-  in `PLAN.md`; a round earns a line here only when it changed what the library
-  *is*, and most earn none.
-- Round numbers appear only where one is the sole handle on an open question.
-- **Restate, don't append.** Later rounds routinely change what an earlier
-  decision meant, so earlier sections need correcting too.
-- **Quote numbers only if they are current.** Test tallies, member counts and
-  benchmark figures go stale first; re-run the relevant command (see
-  `AGENTS.md`) rather than copying a figure forward.
-- **Keep the decisions table honest**: an item leaves it when the decision is
-  made, not when the work is scheduled.
-- Update *Last updated* and the round it covers.