diff --git a/eval/harness/agent-cli.mjs b/eval/harness/agent-cli.mjs
index 67f205b3..8668b0d9 100644
--- a/eval/harness/agent-cli.mjs
+++ b/eval/harness/agent-cli.mjs
@@ -1,5 +1,6 @@
// What the agent runs in its directory: `./check` (the arm's compiler feedback) and `./test` (the visible tests).
-// Every run is appended to `.runlog` (one JSON line each), which the score reads for iterations and when green came.
+// Every run is appended to `.runlog` (one JSON line each: `at` when it ended, `ms` how long it took), which the score
+// reads for iterations, when green came, and how much of the agent's time went to these tools.
// Usage: node harness/agent-cli.mjs
import { appendFileSync } from "node:fs";
import path from "node:path";
@@ -8,7 +9,8 @@ import { runFile } from "./test.mjs";
const [cmd, dirArg] = process.argv.slice(2);
const dir = path.resolve(dirArg);
-const log = entry => appendFileSync(path.join(dir, ".runlog"), JSON.stringify({ at: new Date().toISOString(), cmd, ...entry }) + "\n");
+const started = Date.now();
+const log = entry => appendFileSync(path.join(dir, ".runlog"), JSON.stringify({ at: new Date().toISOString(), ms: Date.now() - started, cmd, ...entry }) + "\n");
if (cmd === "check") {
const r = check(dir);
diff --git a/eval/harness/score.mjs b/eval/harness/score.mjs
index cd975691..424b5394 100644
--- a/eval/harness/score.mjs
+++ b/eval/harness/score.mjs
@@ -1,5 +1,8 @@
// Scores finished runs: for each run directory (named --), the arm's check, the visible and the hidden
// tests, and the agent's own log of `./check` and `./test` runs. Writes results.json and prints a summary table.
+// Time: `toolMs` is the agent's time inside `./check` and `./test` (each `.runlog` entry's `ms`), `greenAfterMs` the time
+// from its first run to its first passing `./test`; `agentMs`, the whole run's wall time, comes from `/.agent.json`
+// (`{ "ms": … }`, written by whoever ran the agent) when there is one. The model's own time is agentMs - toolMs.
// With --overlap, each task's hidden-overlap.mjs (round 3) runs too and counts toward the defect.
// Tests named in `alongside` are run and reported but do not count toward the defect (round 4 pre-registration).
// Usage: node harness/score.mjs [out.json] [--overlap]
@@ -30,6 +33,10 @@ for (const name of readdirSync(runs).sort()) {
const alongside = round5.has(task) ? round5Alongside : round4Alongside;
const log = existsSync(path.join(dir, ".runlog")) ? readFileSync(path.join(dir, ".runlog"), "utf8").trim().split("\n").filter(Boolean).map(l => JSON.parse(l)) : [];
const firstGreen = log.findIndex(e => e.cmd === "test" && e.ok);
+ const startOf = e => Date.parse(e.at) - (e.ms ?? 0);
+ const toolMs = log.every(e => e.ms != null) && log.length ? log.reduce((t, e) => t + e.ms, 0) : null;
+ const greenAfterMs = firstGreen < 0 || !log.length ? null : Date.parse(log[firstGreen].at) - startOf(log[0]);
+ const agent = existsSync(path.join(dir, ".agent.json")) ? JSON.parse(readFileSync(path.join(dir, ".agent.json"), "utf8")) : null;
const checked = check(dir);
const visible = await runFile(dir, path.join(tasksDir, task, "visible.mjs"));
const hidden = await runFile(dir, path.join(tasksDir, task, "hidden.mjs"));
@@ -41,6 +48,7 @@ for (const name of readdirSync(runs).sort()) {
run: name, task, arm, n: Number(n),
checks: log.filter(e => e.cmd === "check").length, tests: log.filter(e => e.cmd === "test").length,
cyclesToGreen: firstGreen < 0 ? null : firstGreen + 1,
+ agentMs: agent?.ms ?? null, toolMs, greenAfterMs,
checkOk: checked.ok, built: hidden.build.ok,
visible: `${visible.results.filter(r => r.ok).length}/${visible.results.length}`,
hidden: `${hidden.results.length - failedHidden.length}/${hidden.results.length}`,
@@ -51,6 +59,6 @@ for (const name of readdirSync(runs).sort()) {
failedVisible: visible.results.filter(r => !r.ok).map(r => ({ name: r.name, error: r.error })),
});
const r = rows.at(-1);
- console.log(`${r.run.padEnd(18)} check ${r.checkOk ? "ok " : "ERR"} visible ${r.visible} hidden ${r.hidden.padEnd(5)}${r.overlap ? ` overlap ${r.overlap}` : ""} cycles ${r.checks + r.tests} (green after ${r.cyclesToGreen ?? "never"})${r.shippedDefect ? " DEFECT" : ""}${r.failedAlongside.length ? " (alongside failed)" : ""}`);
+ console.log(`${r.run.padEnd(18)} check ${r.checkOk ? "ok " : "ERR"} visible ${r.visible} hidden ${r.hidden.padEnd(5)}${r.overlap ? ` overlap ${r.overlap}` : ""} cycles ${r.checks + r.tests} (green after ${r.cyclesToGreen ?? "never"})${r.agentMs != null ? ` agent ${(r.agentMs / 60000).toFixed(1)}m` : ""}${r.toolMs != null ? ` tools ${(r.toolMs / 60000).toFixed(1)}m` : ""}${r.shippedDefect ? " DEFECT" : ""}${r.failedAlongside.length ? " (alongside failed)" : ""}`);
}
writeFileSync(outArg ? path.resolve(outArg) : path.join(runs, "results.json"), JSON.stringify(rows, null, 2) + "\n");
diff --git a/eval/results/desk13-exact-2026-11-13/README.md b/eval/results/desk13-exact-2026-11-13/README.md
new file mode 100644
index 00000000..85b9b7b8
--- /dev/null
+++ b/eval/results/desk13-exact-2026-11-13/README.md
@@ -0,0 +1,27 @@
+# desk9 in Exact, Opus, 3 apps, our harness (2026-11-13)
+
+Not preregistered. The `contract` arm (ported from expo/exact's bench patch, #59) run the way we ran the .rx probe:
+Opus as general-purpose subagents, `--brief plain --inline-docs` (Exact's `docs/start-here.md` inlined, the rest of
+its docs readable), `./check`/`./test` as the only build feedback (15 runs). Exact at main 8c742e6a, from a
+`git archive` with `bench/` removed, so agents could not read Exact's own desk9 reference.
+
+| app | hidden | overlap | fuzz holes | defects | lines (non-blank, non-comment) | subagent tokens | minutes |
+| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
+| contract-1 | 77 | 37 | 0 | 1 | 605 | 155k | 12.4 |
+| contract-2 | 76 | 37 | 0 | 2 | 568 | 145k | 11.4 |
+| contract-3 | 77 | 38 | 0 | **0** | 605 | 168k | 10.5 |
+
+Mean **1.0**. For comparison, the last three .rx apps (rx-7..9, same harness, guide with the lifetime rule, Solid
+next f00ce6f52) score 2, 0, 0 = **0.67**, at 99-107k tokens, about 5 minutes and 411-452 lines each. Three apps per
+arm: neither difference means anything statistically.
+
+## The defects
+
+- **Settings save, contract-1 and contract-2** (overlap "a later save failing while an earlier one is out shows its
+ error and then the earlier result"). Every issue write is a `parallel` mutation, but both apps left
+ `mutation saved` on Exact's default mode, under which a later save replaces the earlier one, so the earlier
+ result never shows. contract-3 made its save `parallel` too (`mutation saved … parallel refreshes settings`) and passes.
+- **contract-2, "rows: row errors and detail errors leave each other alone"**: row actions and the detail share an
+ error path (agent).
+
+No fuzz holes in any Exact app.
diff --git a/eval/results/desk13-exact-2026-11-13/contract-1.grade.txt b/eval/results/desk13-exact-2026-11-13/contract-1.grade.txt
new file mode 100644
index 00000000..f89f7bdb
--- /dev/null
+++ b/eval/results/desk13-exact-2026-11-13/contract-1.grade.txt
@@ -0,0 +1,7 @@
+== hidden
+77/77 passed
+== overlap
+✗ base: settings: a later save failing while an earlier one is out shows its error and then the earlier result
+37/38 passed
+== fuzz
+desk13-contract-1: 0/20 seeds found a hole
diff --git a/eval/results/desk13-exact-2026-11-13/contract-2.grade.txt b/eval/results/desk13-exact-2026-11-13/contract-2.grade.txt
new file mode 100644
index 00000000..aefbe2c5
--- /dev/null
+++ b/eval/results/desk13-exact-2026-11-13/contract-2.grade.txt
@@ -0,0 +1,8 @@
+== hidden
+✗ rows: row errors and detail errors leave each other alone
+76/77 passed
+== overlap
+✗ base: settings: a later save failing while an earlier one is out shows its error and then the earlier result
+37/38 passed
+== fuzz
+desk13-contract-2: 0/20 seeds found a hole
diff --git a/eval/results/desk13-exact-2026-11-13/contract-3.grade.txt b/eval/results/desk13-exact-2026-11-13/contract-3.grade.txt
new file mode 100644
index 00000000..66efee5d
--- /dev/null
+++ b/eval/results/desk13-exact-2026-11-13/contract-3.grade.txt
@@ -0,0 +1,6 @@
+== hidden
+77/77 passed
+== overlap
+38/38 passed
+== fuzz
+desk13-contract-3: 0/20 seeds found a hole
diff --git a/eval/results/desk13-probe-2026-11-13/README.md b/eval/results/desk13-probe-2026-11-13/README.md
index 617685b5..217729ed 100644
--- a/eval/results/desk13-probe-2026-11-13/README.md
+++ b/eval/results/desk13-probe-2026-11-13/README.md
@@ -85,3 +85,26 @@ and no error boundary. Setup:
When the other action lands (its re-ask goes out), and then the detail lands, the `` renders empty while
`detail.body` beside it shows. Without the other action the test passes.
+
+## On next f00ce6f52 + solidjs/solid#4835 (the #4834 fix, still an open PR)
+
+Built at c1949ba3a. Both repros (`repro/blank-rows.spec.tsx`, `repro/blank-heading.spec.tsx`) pass. Grades in `graded-4835/`.
+
+| app | hidden | overlap | fuzz holes | defects | before |
+| --- | --- | --- | --- | --- | --- |
+| rx-1 | 76 | 37 | 1 | 3 | 4 |
+| rx-2 | 76 | 37 | 0 | 2 | 2 |
+| rx-3 | 76 | 37 | 2 | 4 | 5 |
+| rx-4 | 77 | 37 | 0 | 1 | 2 |
+| rx-5 | 77 | 37 | 1 | 2 | 2 |
+| rx-6 | 77 | 36 | 0 | 2 | 3 |
+| rx-7 | 75 | 38 | 0 | 2 | 2 |
+| rx-8 | 77 | 38 | 0 | 0 | 0 |
+| rx-9 | 77 | 38 | 0 | 0 | 0 |
+| reference | 77 | 38 | 0 | 0 | 0 |
+
+"Before" is the last grade of each app: rx-1, rx-3 and rx-8 on f00ce6f52, the others on the Solid of their first grade.
+The blank heading (seed 15 in rx-1 and rx-3) is gone. rx-4 s13 and rx-6 s16 (count outside the list's region) no
+longer fire; rx-5 s17 still does. Everything left is agent or guide-caused, except two untraced holes: rx-1 s11
+(the #3 detail shows #4's guessed title "Bravo"; the trail replayed alone ends correctly, so it may depend on timing)
+and rx-3 s11/s19 (page hangs). Apps on the current guide (rx-7..9): 2, 0, 0.
diff --git a/eval/results/desk13-probe-2026-11-13/graded-4835/reference.grade b/eval/results/desk13-probe-2026-11-13/graded-4835/reference.grade
new file mode 100644
index 00000000..980f6dca
--- /dev/null
+++ b/eval/results/desk13-probe-2026-11-13/graded-4835/reference.grade
@@ -0,0 +1,6 @@
+== hidden
+77/77 passed
+== overlap
+38/38 passed
+== fuzz
+ref-desk9: 0/20 seeds found a hole
diff --git a/eval/results/desk13-probe-2026-11-13/graded-4835/rx-1.grade b/eval/results/desk13-probe-2026-11-13/graded-4835/rx-1.grade
new file mode 100644
index 00000000..17b05ceb
--- /dev/null
+++ b/eval/results/desk13-probe-2026-11-13/graded-4835/rx-1.grade
@@ -0,0 +1,10 @@
+== hidden
+✗ base: tabs: the settings page loads again when reopened, with a fresh draft
+76/77 passed
+== overlap
+✗ base: tabs: a save that answers after the page was left shows the reloaded settings
+37/38 passed
+== fuzz
+desk13-rx-1: 1/20 seeds found a hole
+ seed 11 step 43: detail title: "#3 Bravofooter says 2019finding related helpretry assign to ", server "Delta"
+ tab help ; tab issues ; tab help ; land loadSettings: ; tab issues ; filter closed ; tab help ; land fetchIssues:open ; land fetchIssues:closed ; tab issues ; tab help ; tab issues ; focus ; land fetchIssues:open ; select #4 ; rename Alpha ; rename Bravo ; post me too ; post fixed? ; land fetchIssue:4 ; row-assign 0 ; row-close 0 ; select #3 ; focus ; close ; land fetchComments:4 ; row-close 0
diff --git a/eval/results/desk13-probe-2026-11-13/graded-4835/rx-2.grade b/eval/results/desk13-probe-2026-11-13/graded-4835/rx-2.grade
new file mode 100644
index 00000000..dc7efdb7
--- /dev/null
+++ b/eval/results/desk13-probe-2026-11-13/graded-4835/rx-2.grade
@@ -0,0 +1,8 @@
+== hidden
+✗ base: tabs: the settings page loads again when reopened, with a fresh draft
+76/77 passed
+== overlap
+✗ base: tabs: a save that answers after the page was left shows the reloaded settings
+37/38 passed
+== fuzz
+desk13-rx-2: 0/20 seeds found a hole
diff --git a/eval/results/desk13-probe-2026-11-13/graded-4835/rx-3.grade b/eval/results/desk13-probe-2026-11-13/graded-4835/rx-3.grade
new file mode 100644
index 00000000..ed341b86
--- /dev/null
+++ b/eval/results/desk13-probe-2026-11-13/graded-4835/rx-3.grade
@@ -0,0 +1,12 @@
+== hidden
+✗ base: tabs: the settings page loads again when reopened, with a fresh draft
+76/77 passed
+== overlap
+✗ base: tabs: a save that answers after the page was left shows the reloaded settings
+37/38 passed
+== fuzz
+desk13-rx-3: 2/20 seeds found a hole
+ seed 11 step -1: page crash: session timed out (120 s): the page hangs
+
+ seed 19 step -1: page crash: session timed out (120 s): the page hangs
+
diff --git a/eval/results/desk13-probe-2026-11-13/graded-4835/rx-4.grade b/eval/results/desk13-probe-2026-11-13/graded-4835/rx-4.grade
new file mode 100644
index 00000000..58e12498
--- /dev/null
+++ b/eval/results/desk13-probe-2026-11-13/graded-4835/rx-4.grade
@@ -0,0 +1,7 @@
+== hidden
+77/77 passed
+== overlap
+✗ base: tabs: a save that answers after the page was left shows the reloaded settings
+37/38 passed
+== fuzz
+desk13-rx-4: 0/20 seeds found a hole
diff --git a/eval/results/desk13-probe-2026-11-13/graded-4835/rx-5.grade b/eval/results/desk13-probe-2026-11-13/graded-4835/rx-5.grade
new file mode 100644
index 00000000..5cc165a4
--- /dev/null
+++ b/eval/results/desk13-probe-2026-11-13/graded-4835/rx-5.grade
@@ -0,0 +1,9 @@
+== hidden
+77/77 passed
+== overlap
+✗ base: tabs: a save that answers after the page was left shows the reloaded settings
+37/38 passed
+== fuzz
+desk13-rx-5: 1/20 seeds found a hole
+ seed 17 step 8: count: tab Issues (2), 1 open rows shown
+ user bob ; land fetchIssues:open ; row-assign 1 ; land assign:3,bob ; land fetchIssues:open ; row-close 2 ; land closeIssue:4 ; row-close 1
diff --git a/eval/results/desk13-probe-2026-11-13/graded-4835/rx-6.grade b/eval/results/desk13-probe-2026-11-13/graded-4835/rx-6.grade
new file mode 100644
index 00000000..5a85de3c
--- /dev/null
+++ b/eval/results/desk13-probe-2026-11-13/graded-4835/rx-6.grade
@@ -0,0 +1,8 @@
+== hidden
+77/77 passed
+== overlap
+✗ base: tabs: a save that answers after the page was left shows the reloaded settings
+✗ rows: a row action that answers after the issues page was left changes nothing on its return
+36/38 passed
+== fuzz
+desk13-rx-6: 0/20 seeds found a hole
diff --git a/eval/results/desk13-probe-2026-11-13/graded-4835/rx-7.grade b/eval/results/desk13-probe-2026-11-13/graded-4835/rx-7.grade
new file mode 100644
index 00000000..e7ef0e81
--- /dev/null
+++ b/eval/results/desk13-probe-2026-11-13/graded-4835/rx-7.grade
@@ -0,0 +1,8 @@
+== hidden
+✗ base: help: selecting after a failure shows loading…, not the old failure
+✗ base: help: selecting after not found shows loading…, not the old message
+75/77 passed
+== overlap
+38/38 passed
+== fuzz
+desk13-rx-7: 0/20 seeds found a hole
diff --git a/eval/results/desk13-probe-2026-11-13/graded-4835/rx-8.grade b/eval/results/desk13-probe-2026-11-13/graded-4835/rx-8.grade
new file mode 100644
index 00000000..eabf887a
--- /dev/null
+++ b/eval/results/desk13-probe-2026-11-13/graded-4835/rx-8.grade
@@ -0,0 +1,6 @@
+== hidden
+77/77 passed
+== overlap
+38/38 passed
+== fuzz
+desk13-rx-8: 0/20 seeds found a hole
diff --git a/eval/results/desk13-probe-2026-11-13/graded-4835/rx-9.grade b/eval/results/desk13-probe-2026-11-13/graded-4835/rx-9.grade
new file mode 100644
index 00000000..40d2a377
--- /dev/null
+++ b/eval/results/desk13-probe-2026-11-13/graded-4835/rx-9.grade
@@ -0,0 +1,6 @@
+== hidden
+77/77 passed
+== overlap
+38/38 passed
+== fuzz
+desk13-rx-9: 0/20 seeds found a hole
diff --git a/eval/results/desk14-2026-11-13/README.md b/eval/results/desk14-2026-11-13/README.md
new file mode 100644
index 00000000..be3aa30f
--- /dev/null
+++ b/eval/results/desk14-2026-11-13/README.md
@@ -0,0 +1,52 @@
+# Round 14 (exploratory): frontier .rx vs React vs Exact, 2026-11-13
+
+Not preregistered. 18 Opus apps on the desk9 brief (`--brief plain --inline-docs`), 6 per arm, run in three waves of
+six (two per arm per wave). Graded with the usual suite: hidden 77, overlap 38, 20 fuzz seeds (`--r9`).
+
+- Solid next dba2aadbb (includes #4831 and #4835, the #4834 fix). rxc built from main + this branch (54853d0).
+- React arm: React 19 + TypeScript, @tanstack/react-query installed and offered.
+- Exact arm: Exact `contract` (app.contract + app.ts) from the checkout described in `desk13-exact-2026-11-13`.
+- The .rx guide is the one-lifetime-rule guide from the desk13 probe.
+- The hand-written .rx reference grades 0 (77/77, 38/38, 0/20).
+
+## Results
+
+| arm | defects per app | mean | apps at 0 | mean minutes | mean tokens |
+| --- | --- | --- | --- | --- | --- |
+| .rx | 1, 0, 0, 0, 0, 1 | **0.33** | 4/6 | 7.1 | 112k |
+| React | 2, 3, 1, 3, 2, 2 | 2.17 | 0/6 | 6.7 | 89k |
+| Exact | 2, 2, 2, 1, 0, 3 | 1.67 | 1/6 | 9.9 | 150k |
+
+Minutes are each agent's wall time (from its task notification, in `/.agent.json`). Three waves of six ran on
+one machine, so a wave's builds and browsers shared the CPU.
+
+## What the defects are
+
+- **.rx (2): both are page hangs, and both are a Solid bug.** rx-1 seed 2 and rx-6 seed 19 time out (120 s): selecting
+ another issue while a guess is out (a rename, or an assign) never returns. Reduced to
+ `repro/guess-select-loop.spec.tsx` (plain Solid, about 35 lines; filed solidjs/solid#4840): a detail rendered per selection through ``
+ under a component that passes its children by getter, a heading memo reading a `createOptimistic` guess, and a
+ selection change while that guess's action is out. Solid's dev build reports "Potential Infinite Loop Detected. Kept
+ alive by scheduled work; last staged node: "; the production build spins. rx-6 s19 is the rename case exactly;
+ rx-1 s2 has assign guesses out instead (not reduced separately). The same loop is behind desk13 rx-3 s11/s19.
+- **React (13): mostly one class.** 6/6 lose a guessed comment or label when the list it sits in reloads while the
+ post or add is out (hidden "labels that load while an add is out keep the add" 6/6, "comments ... keep the post"
+ 5/6). react-2 and react-4 also show a posted comment twice at rest (fuzz s16).
+- **Exact (10): mostly the settings save.** 4/6 get overlapping saves wrong ("a later save failing while an earlier
+ one is out"; contract-3 also "an earlier save failing after a later one succeeded"), the default mutation mode
+ replacing an earlier write with a later one, as in the desk13 Exact probe. 2/6 let a row error and a detail error
+ clear each other; contract-1 doesn't reload settings when the page reopens (2 tests); contract-6 asks the server
+ for an invalid name.
+
+## Also traced: desk13 rx-1 seed 11 (Solid, filed solidjs/solid#4841)
+
+The detail of issue #3 showed issue #4's data. Instrumenting the app showed the detail memo computing #3's answer
+while the screen kept #4's. `repro/wrong-detail-after-error.spec.tsx` (plain Solid): a heading memo reading a rename
+guess; select #3 while closes and renames are out; #3's detail fails, a rename's refresh asks again, and the answer
+lands while a close is still out. The screen shows "#3 Alpha (detail of #4)" instead of "#3 Delta (detail of #3)".
+
+## Deviations
+
+`deviations.txt`. Most agents in every arm added their own tests to `tests/visible.mjs` and restored the file;
+contract-6 kept its extra tests in `.exact/extra.mjs`. rx-6 removed an unused import after its last test run (it
+type-checks; grading builds it fine).
diff --git a/eval/results/desk14-2026-11-13/deviations.txt b/eval/results/desk14-2026-11-13/deviations.txt
new file mode 100644
index 00000000..07f06b23
--- /dev/null
+++ b/eval/results/desk14-2026-11-13/deviations.txt
@@ -0,0 +1,11 @@
+rx-1: added 9 own tests to tests/visible.mjs, then restored it (agent says exactly)
+rx-2: added 14 own tests to tests/visible.mjs, then restored it
+contract-1: added 17 own tests to tests/visible.mjs, then restored it
+contract-2: added 12 own tests to tests/visible.mjs, then restored it
+contract-4: added 11 own tests to tests/visible.mjs, then removed
+react-3: added 12 own tests to tests/visible.mjs, then restored
+contract-3: added 13 own tests, then removed
+rx-6: last edit (removed unused import) checked but not re-tested
+react-6: added 9 own tests, then restored. contract-6: extra tests kept in .exact/extra.mjs (not graded)
+contract-5: 15 extra tests, then restored
+rx-5: 19 extra tests, then restored
diff --git a/eval/results/desk14-2026-11-13/graded/contract-1.grade b/eval/results/desk14-2026-11-13/graded/contract-1.grade
new file mode 100644
index 00000000..3376e31d
--- /dev/null
+++ b/eval/results/desk14-2026-11-13/graded/contract-1.grade
@@ -0,0 +1,8 @@
+== hidden
+✗ base: tabs: the settings page loads again when reopened, with a fresh draft
+76/77 passed
+== overlap
+✗ base: tabs: a save that answers after the page was left shows the reloaded settings
+37/38 passed
+== fuzz
+desk14-contract-1: 0/20 seeds found a hole
diff --git a/eval/results/desk14-2026-11-13/graded/contract-2.grade b/eval/results/desk14-2026-11-13/graded/contract-2.grade
new file mode 100644
index 00000000..43daf8cc
--- /dev/null
+++ b/eval/results/desk14-2026-11-13/graded/contract-2.grade
@@ -0,0 +1,8 @@
+== hidden
+✗ rows: row errors and detail errors leave each other alone
+76/77 passed
+== overlap
+✗ base: settings: a later save failing while an earlier one is out shows its error and then the earlier result
+37/38 passed
+== fuzz
+desk14-contract-2: 0/20 seeds found a hole
diff --git a/eval/results/desk14-2026-11-13/graded/contract-3.grade b/eval/results/desk14-2026-11-13/graded/contract-3.grade
new file mode 100644
index 00000000..22e32a4e
--- /dev/null
+++ b/eval/results/desk14-2026-11-13/graded/contract-3.grade
@@ -0,0 +1,8 @@
+== hidden
+77/77 passed
+== overlap
+✗ base: settings: an earlier save failing after a later one succeeded leaves the later one
+✗ base: settings: a later save failing while an earlier one is out shows its error and then the earlier result
+36/38 passed
+== fuzz
+desk14-contract-3: 0/20 seeds found a hole
diff --git a/eval/results/desk14-2026-11-13/graded/contract-4.grade b/eval/results/desk14-2026-11-13/graded/contract-4.grade
new file mode 100644
index 00000000..9d4568aa
--- /dev/null
+++ b/eval/results/desk14-2026-11-13/graded/contract-4.grade
@@ -0,0 +1,7 @@
+== hidden
+77/77 passed
+== overlap
+✗ base: settings: a later save failing while an earlier one is out shows its error and then the earlier result
+37/38 passed
+== fuzz
+desk14-contract-4: 0/20 seeds found a hole
diff --git a/eval/results/desk14-2026-11-13/graded/contract-5.grade b/eval/results/desk14-2026-11-13/graded/contract-5.grade
new file mode 100644
index 00000000..9d8c23bf
--- /dev/null
+++ b/eval/results/desk14-2026-11-13/graded/contract-5.grade
@@ -0,0 +1,6 @@
+== hidden
+77/77 passed
+== overlap
+38/38 passed
+== fuzz
+desk14-contract-5: 0/20 seeds found a hole
diff --git a/eval/results/desk14-2026-11-13/graded/contract-6.grade b/eval/results/desk14-2026-11-13/graded/contract-6.grade
new file mode 100644
index 00000000..43a83659
--- /dev/null
+++ b/eval/results/desk14-2026-11-13/graded/contract-6.grade
@@ -0,0 +1,9 @@
+== hidden
+✗ base: settings: saving an invalid name asks nothing
+✗ rows: row errors and detail errors leave each other alone
+75/77 passed
+== overlap
+✗ base: settings: a later save failing while an earlier one is out shows its error and then the earlier result
+37/38 passed
+== fuzz
+desk14-contract-6: 0/20 seeds found a hole
diff --git a/eval/results/desk14-2026-11-13/graded/react-1.grade b/eval/results/desk14-2026-11-13/graded/react-1.grade
new file mode 100644
index 00000000..9b4a70f5
--- /dev/null
+++ b/eval/results/desk14-2026-11-13/graded/react-1.grade
@@ -0,0 +1,8 @@
+== hidden
+77/77 passed
+== overlap
+✗ base: comments that load while a post is out keep the post
+✗ labels that load while an add is out keep the add
+36/38 passed
+== fuzz
+desk14-react-1: 0/20 seeds found a hole
diff --git a/eval/results/desk14-2026-11-13/graded/react-2.grade b/eval/results/desk14-2026-11-13/graded/react-2.grade
new file mode 100644
index 00000000..94fa18f8
--- /dev/null
+++ b/eval/results/desk14-2026-11-13/graded/react-2.grade
@@ -0,0 +1,10 @@
+== hidden
+77/77 passed
+== overlap
+✗ base: comments that load while a post is out keep the post
+✗ labels that load while an add is out keep the add
+36/38 passed
+== fuzz
+desk14-react-2: 1/20 seeds found a hole
+ seed 16 step 37: at rest: comments: #2 shows [guest: ping | guest: ping], server [guest: ping]
+ user guest ; fail fetchIssues:open ; focus ; tab help ; tab issues ; filter closed ; land fetchIssues:open ; focus ; land fetchIssues:closed ; select #2 ; land fetchLabels:2 ; close ; post me too ; fail addComment:2,guest,me too ; row-assign 0 ; row-assign 0 ; close ; post ping ; assign ; land addComment:2,guest,ping ; land fetchIssue:2 ; fail assign:2,guest ; select #2 ; rename Bravo ; assign
diff --git a/eval/results/desk14-2026-11-13/graded/react-3.grade b/eval/results/desk14-2026-11-13/graded/react-3.grade
new file mode 100644
index 00000000..3211d1c9
--- /dev/null
+++ b/eval/results/desk14-2026-11-13/graded/react-3.grade
@@ -0,0 +1,7 @@
+== hidden
+77/77 passed
+== overlap
+✗ labels that load while an add is out keep the add
+37/38 passed
+== fuzz
+desk14-react-3: 0/20 seeds found a hole
diff --git a/eval/results/desk14-2026-11-13/graded/react-4.grade b/eval/results/desk14-2026-11-13/graded/react-4.grade
new file mode 100644
index 00000000..15010665
--- /dev/null
+++ b/eval/results/desk14-2026-11-13/graded/react-4.grade
@@ -0,0 +1,10 @@
+== hidden
+77/77 passed
+== overlap
+✗ base: comments that load while a post is out keep the post
+✗ labels that load while an add is out keep the add
+36/38 passed
+== fuzz
+desk14-react-4: 1/20 seeds found a hole
+ seed 16 step 36: at rest: comments: #2 shows [guest: ping | guest: ping], server [guest: ping]
+ user guest ; fail fetchIssues:open ; focus ; tab help ; tab issues ; filter closed ; land fetchIssues:open ; focus ; land fetchIssues:closed ; select #2 ; land fetchLabels:2 ; close ; post me too ; fail addComment:2,guest,me too ; row-assign 0 ; row-assign 0 ; close ; post ping ; assign ; land addComment:2,guest,ping ; land fetchIssue:2 ; fail assign:2,guest ; select #2 ; rename Bravo ; assign
diff --git a/eval/results/desk14-2026-11-13/graded/react-5.grade b/eval/results/desk14-2026-11-13/graded/react-5.grade
new file mode 100644
index 00000000..57e7c987
--- /dev/null
+++ b/eval/results/desk14-2026-11-13/graded/react-5.grade
@@ -0,0 +1,8 @@
+== hidden
+77/77 passed
+== overlap
+✗ base: comments that load while a post is out keep the post
+✗ labels that load while an add is out keep the add
+36/38 passed
+== fuzz
+desk14-react-5: 0/20 seeds found a hole
diff --git a/eval/results/desk14-2026-11-13/graded/react-6.grade b/eval/results/desk14-2026-11-13/graded/react-6.grade
new file mode 100644
index 00000000..6c781c6e
--- /dev/null
+++ b/eval/results/desk14-2026-11-13/graded/react-6.grade
@@ -0,0 +1,8 @@
+== hidden
+77/77 passed
+== overlap
+✗ base: comments that load while a post is out keep the post
+✗ labels that load while an add is out keep the add
+36/38 passed
+== fuzz
+desk14-react-6: 0/20 seeds found a hole
diff --git a/eval/results/desk14-2026-11-13/graded/reference-rx.grade b/eval/results/desk14-2026-11-13/graded/reference-rx.grade
new file mode 100644
index 00000000..407420dd
--- /dev/null
+++ b/eval/results/desk14-2026-11-13/graded/reference-rx.grade
@@ -0,0 +1,6 @@
+== hidden
+77/77 passed
+== overlap
+38/38 passed
+== fuzz
+ref-rx: 0/20 seeds found a hole
diff --git a/eval/results/desk14-2026-11-13/graded/rx-1.grade b/eval/results/desk14-2026-11-13/graded/rx-1.grade
new file mode 100644
index 00000000..6189ae47
--- /dev/null
+++ b/eval/results/desk14-2026-11-13/graded/rx-1.grade
@@ -0,0 +1,8 @@
+== hidden
+77/77 passed
+== overlap
+38/38 passed
+== fuzz
+desk14-rx-1: 1/20 seeds found a hole
+ seed 2 step -1: page crash: session timed out (120 s): the page hangs
+
diff --git a/eval/results/desk14-2026-11-13/graded/rx-2.grade b/eval/results/desk14-2026-11-13/graded/rx-2.grade
new file mode 100644
index 00000000..10901bff
--- /dev/null
+++ b/eval/results/desk14-2026-11-13/graded/rx-2.grade
@@ -0,0 +1,6 @@
+== hidden
+77/77 passed
+== overlap
+38/38 passed
+== fuzz
+desk14-rx-2: 0/20 seeds found a hole
diff --git a/eval/results/desk14-2026-11-13/graded/rx-3.grade b/eval/results/desk14-2026-11-13/graded/rx-3.grade
new file mode 100644
index 00000000..1c86884c
--- /dev/null
+++ b/eval/results/desk14-2026-11-13/graded/rx-3.grade
@@ -0,0 +1,6 @@
+== hidden
+77/77 passed
+== overlap
+38/38 passed
+== fuzz
+desk14-rx-3: 0/20 seeds found a hole
diff --git a/eval/results/desk14-2026-11-13/graded/rx-4.grade b/eval/results/desk14-2026-11-13/graded/rx-4.grade
new file mode 100644
index 00000000..2098eeb7
--- /dev/null
+++ b/eval/results/desk14-2026-11-13/graded/rx-4.grade
@@ -0,0 +1,6 @@
+== hidden
+77/77 passed
+== overlap
+38/38 passed
+== fuzz
+desk14-rx-4: 0/20 seeds found a hole
diff --git a/eval/results/desk14-2026-11-13/graded/rx-5.grade b/eval/results/desk14-2026-11-13/graded/rx-5.grade
new file mode 100644
index 00000000..f3ec928d
--- /dev/null
+++ b/eval/results/desk14-2026-11-13/graded/rx-5.grade
@@ -0,0 +1,6 @@
+== hidden
+77/77 passed
+== overlap
+38/38 passed
+== fuzz
+desk14-rx-5: 0/20 seeds found a hole
diff --git a/eval/results/desk14-2026-11-13/graded/rx-6.grade b/eval/results/desk14-2026-11-13/graded/rx-6.grade
new file mode 100644
index 00000000..443aeaf3
--- /dev/null
+++ b/eval/results/desk14-2026-11-13/graded/rx-6.grade
@@ -0,0 +1,8 @@
+== hidden
+77/77 passed
+== overlap
+38/38 passed
+== fuzz
+desk14-rx-6: 1/20 seeds found a hole
+ seed 19 step -1: page crash: session timed out (120 s): the page hangs
+
diff --git a/eval/results/desk14-2026-11-13/repro/guess-select-loop.spec.tsx b/eval/results/desk14-2026-11-13/repro/guess-select-loop.spec.tsx
new file mode 100644
index 00000000..9f0865d3
--- /dev/null
+++ b/eval/results/desk14-2026-11-13/repro/guess-select-loop.spec.tsx
@@ -0,0 +1,37 @@
+/**
+ * @jsxImportSource @solidjs/web
+ * @vitest-environment jsdom
+ */
+import { test, expect } from "vitest";
+import { For, Loading, action, createOptimistic, createMemo, createSignal, flush } from "solid-js";
+import { render } from "../src/index.js";
+const tick = async () => { for (let i = 0; i < 4; i++) { await new Promise(r => setTimeout(r, 0)); flush(); } };
+test("selecting another issue while a rename guess is out does not loop", async () => {
+ const asks: { key: string; res: (v: any) => void }[] = [];
+ const ask = (key: string) => new Promise(res => asks.push({ key, res }));
+ const pick = (key: string) => { const i = asks.findIndex(a => a.key === key); if (i < 0) throw new Error("no ask " + key); return asks.splice(i, 1)[0]; };
+ const c = document.body.appendChild(document.createElement("div"));
+
+ const [selected, setSelected] = createSignal(null);
+ const [renaming, guessRename] = createOptimistic<{ id: number; title: string }[]>(() => []);
+ const detail = createMemo(() => selected() != null ? ask(`detail:${selected()}`) : undefined);
+ const heading = createMemo(() => { const d: any = detail(); return d == null ? "" : renaming().reduce((t, r) => r.id === d.id ? r.title : t, d.title); });
+ const selList = createMemo(() => selected() == null ? [] : [selected()]);
+ const rename = action(function* (id: number, title: string) { guessRename(rs => [...rs, { id, title }]); yield ask(`rename:${id}`); });
+
+ // A page component: its children are read through a getter, as a compiled `if` passes them.
+ const Page = (p: { children: any }) => <>{p.children}>;
+ const Detail = (props: { id: number }) =>
+
loading #{props.id}}>
+ #{props.id} {heading()}
{(detail() as any)?.body}
+
+
;
+ render(() => {(s: any) => }, c);
+
+ setSelected(4); await tick();
+ rename(4, "Bravo"); await tick(); // the guess stays out: rename never answers
+ pick("detail:4").res({ id: 4, title: "Typo in footer", body: "four" }); await tick();
+ expect(c.textContent).toContain("#4 Bravo");
+ setSelected(3); await tick(); // here: "Potential Infinite Loop Detected" (a hang in production)
+ expect(c.textContent).toContain("loading #3");
+});
diff --git a/eval/results/desk14-2026-11-13/repro/wrong-detail-after-error.spec.tsx b/eval/results/desk14-2026-11-13/repro/wrong-detail-after-error.spec.tsx
new file mode 100644
index 00000000..f1f1361c
--- /dev/null
+++ b/eval/results/desk14-2026-11-13/repro/wrong-detail-after-error.spec.tsx
@@ -0,0 +1,59 @@
+/**
+ * @jsxImportSource @solidjs/web
+ * @vitest-environment jsdom
+ */
+import { test, expect } from "vitest";
+import { Loading, Errored, Show, action, createOptimistic, createMemo, createSignal, refresh, flush } from "solid-js";
+import { render } from "../src/index.js";
+const tick = async () => { for (let i = 0; i < 4; i++) { await new Promise(r => setTimeout(r, 0)); flush(); } };
+test("after a failed detail, its re-ask's answer shows, not the detail of the issue selected before", async () => {
+ const asks: { key: string; p: Promise; res: (v: any) => void; rej: (e: any) => void }[] = [];
+ // one flight per key while it is out, like the app's backend plumbing
+ const ask = (key: string) => { const o = asks.find(a => a.key === key); if (o) return o.p; let res: any, rej: any; const p = new Promise((a, b) => { res = a; rej = b; }); asks.push({ key, p, res, rej }); return p; };
+ const pick = (key: string) => { const i = asks.findIndex(a => a.key === key); if (i < 0) throw new Error("no ask " + key + " in " + asks.map(a => a.key)); return asks.splice(i, 1)[0]; };
+ const c = document.body.appendChild(document.createElement("div"));
+ const frames: string[] = []; const snap = (l: string) => frames.push(`${l.padEnd(18)} ${c.innerHTML} | in the air: ${asks.map(a => a.key)}`);
+
+ const [selected, setSelected] = createSignal(null);
+ const [renaming, guessRename] = createOptimistic<{ id: number; title: string }[]>(() => []);
+ const [closing, guessClose] = createOptimistic(() => []);
+
+ const list = createMemo(() => ask("list"));
+ const answer = createMemo(() => selected() != null ? ask(`detail:${selected()}`) : undefined);
+ const detail = createMemo(() => selected() == null ? null : answer());
+ const heading = createMemo(() => { const d: any = detail(); return d == null ? "" : renaming().reduce((t, r) => r.id === d.id ? r.title : t, d.title); });
+
+ const select = action(function* (id: number) { setSelected(id); });
+ const rename = action(function* (id: number, title: string) { guessRename(rs => [...rs, { id, title }]); yield ask(`rename:${id}`); refresh(list); refresh(answer); refresh(detail); });
+ const close = action(function* (id: number) { guessClose(cs => [...cs, id]); yield ask(`close:${id}`); refresh(list); });
+
+ render(() =>
+ select
}>
+ loading #{selected()}}>
+ detail error: {e()?.message}
}>
+ #{selected()} {heading()} (detail of #{(detail() as any)?.id})
+
+
+
+ , c);
+
+ const issues = [{ id: 1, title: "Login fails" }, { id: 3, title: "Crash on save" }, { id: 4, title: "Typo in footer" }];
+ const steps: [string, () => void][] = [
+ ["list lands", () => pick("list").res(issues)],
+ ["select #4", () => select(4)],
+ ["rename #4", () => rename(4, "Alpha")],
+ ["detail #4 lands", () => pick("detail:4").res({ id: 4, title: "Typo in footer" })],
+ ["select #3", () => select(3)],
+ ["close #3", () => close(3)],
+ ["close #4", () => close(4)],
+ ["rename #3", () => rename(3, "Delta")],
+ ["rename #4 lands", () => pick("rename:4").res({})],
+ ["detail #3 fails", () => pick("detail:3").rej(new Error("declined"))],
+ ["close #4 lands", () => pick("close:4").res({})],
+ ["rename #3 lands", () => pick("rename:3").res({})],
+ ["detail #3 lands", () => pick("detail:3").res({ id: 3, title: "Delta" })],
+ ];
+ for (const [n, f] of steps) { f(); await tick(); snap(n); }
+ console.log(frames.join("\n"));
+ expect(c.querySelector("h2")?.textContent).toBe("#3 Delta (detail of #3)");
+});