fix(run): fidelity readout never rounds a near-unity F to a bare 1

[kateebonner]

What

formatFidelity (problem.ts) rendered with toFixed(4), so any $F \geq 0.99995$ displayed as a flat "F = 1" in the run chip, the ring-2 Run hero, and the Run modal — Kate hit this today on a transmon X-gate solve that finished at F ≈ 1 but almost certainly not exactly 1.

How

Adaptive precision in the one shared formatter: ordinary values keep the 4-decimal form (existing renders unchanged — F=0.9998 · 60 iter, solving… F=0.968), but when rounding would collapse the value to exactly 1 or 0, precision extends one digit at a time (capped at 10) until the gap survives. Only a true 1/0 renders bare. The gap from perfect is the result; the display must never erase it.

Applies automatically to every consumer of runChipText/runVerdict — chip, hero, modal.

Verification

• bun test src/amicode/problem.test.ts: 34 pass, including new near-unity cases (0.99997 → F=0.99997, 0.99999997 → F=0.99999997, solving f=3e-5 → F=0.99997, true 1 → F=1)
• npm run lint: 0 errors; tsgo -b in packages/app: clean

🤖 Generated with Claude Code

[kateebonner]

What

formatFidelity (problem.ts) rendered with toFixed(4), so any $F \geq 0.99995$ displayed as a flat "F = 1" in the run chip, the ring-2 Run hero, and the Run modal — Kate hit this today on a transmon X-gate solve that finished at F ≈ 1 but almost certainly not exactly 1.

How

Adaptive precision in the one shared formatter: ordinary values keep the 4-decimal form (existing renders unchanged — F=0.9998 · 60 iter, solving… F=0.968), but when rounding would collapse the value to exactly 1 or 0, precision extends one digit at a time (capped at 10) until the gap survives. Only a true 1/0 renders bare. The gap from perfect is the result; the display must never erase it.

Applies automatically to every consumer of runChipText/runVerdict — chip, hero, modal.

Verification

• bun test src/amicode/problem.test.ts: 34 pass, including new near-unity cases (0.99997 → F=0.99997, 0.99999997 → F=0.99999997, solving f=3e-5 → F=0.99997, true 1 → F=1)
• npm run lint: 0 errors; tsgo -b in packages/app: clean

🤖 Generated with Claude Code

[kateebonner]

What

formatFidelity (problem.ts) rendered with toFixed(4), so any $F \geq 0.99995$ displayed as a flat "F = 1" in the run chip, the ring-2 Run hero, and the Run modal — Kate hit this today on a transmon X-gate solve that finished at F ≈ 1 but almost certainly not exactly 1.

How

Adaptive precision in the one shared formatter: ordinary values keep the 4-decimal form (existing renders unchanged — F=0.9998 · 60 iter, solving… F=0.968), but when rounding would collapse the value to exactly 1 or 0, precision extends one digit at a time (capped at 10) until the gap survives. Only a true 1/0 renders bare. The gap from perfect is the result; the display must never erase it.

Applies automatically to every consumer of runChipText/runVerdict — chip, hero, modal.

Verification

• bun test src/amicode/problem.test.ts: 34 pass, including new near-unity cases (0.99997 → F=0.99997, 0.99999997 → F=0.99999997, solving f=3e-5 → F=0.99997, true 1 → F=1)
• npm run lint: 0 errors; tsgo -b in packages/app: clean

🤖 Generated with Claude Code

[kateebonner]

What

formatFidelity (problem.ts) rendered with toFixed(4), so any $F \geq 0.99995$ displayed as a flat "F = 1" in the run chip, the ring-2 Run hero, and the Run modal — Kate hit this today on a transmon X-gate solve that finished at F ≈ 1 but almost certainly not exactly 1.

How

Adaptive precision in the one shared formatter: ordinary values keep the 4-decimal form (existing renders unchanged — F=0.9998 · 60 iter, solving… F=0.968), but when rounding would collapse the value to exactly 1 or 0, precision extends one digit at a time (capped at 10) until the gap survives. Only a true 1/0 renders bare. The gap from perfect is the result; the display must never erase it.

Applies automatically to every consumer of runChipText/runVerdict — chip, hero, modal.

Verification

• bun test src/amicode/problem.test.ts: 34 pass, including new near-unity cases (0.99997 → F=0.99997, 0.99999997 → F=0.99999997, solving f=3e-5 → F=0.99997, true 1 → F=1)
• npm run lint: 0 errors; tsgo -b in packages/app: clean

🤖 Generated with Claude Code

[kateebonner]

What

formatFidelity (problem.ts) rendered with toFixed(4), so any $F \geq 0.99995$ displayed as a flat "F = 1" in the run chip, the ring-2 Run hero, and the Run modal — Kate hit this today on a transmon X-gate solve that finished at F ≈ 1 but almost certainly not exactly 1.

How

Adaptive precision in the one shared formatter: ordinary values keep the 4-decimal form (existing renders unchanged — F=0.9998 · 60 iter, solving… F=0.968), but when rounding would collapse the value to exactly 1 or 0, precision extends one digit at a time (capped at 10) until the gap survives. Only a true 1/0 renders bare. The gap from perfect is the result; the display must never erase it.

Applies automatically to every consumer of runChipText/runVerdict — chip, hero, modal.

Verification

• bun test src/amicode/problem.test.ts: 34 pass, including new near-unity cases (0.99997 → F=0.99997, 0.99999997 → F=0.99999997, solving f=3e-5 → F=0.99997, true 1 → F=1)
• npm run lint: 0 errors; tsgo -b in packages/app: clean

🤖 Generated with Claude Code

[kateebonner]

What

formatFidelity (problem.ts) rendered with toFixed(4), so any $F \geq 0.99995$ displayed as a flat "F = 1" in the run chip, the ring-2 Run hero, and the Run modal — Kate hit this today on a transmon X-gate solve that finished at F ≈ 1 but almost certainly not exactly 1.

How

Adaptive precision in the one shared formatter: ordinary values keep the 4-decimal form (existing renders unchanged — F=0.9998 · 60 iter, solving… F=0.968), but when rounding would collapse the value to exactly 1 or 0, precision extends one digit at a time (capped at 10) until the gap survives. Only a true 1/0 renders bare. The gap from perfect is the result; the display must never erase it.

Applies automatically to every consumer of runChipText/runVerdict — chip, hero, modal.

Verification

• bun test src/amicode/problem.test.ts: 34 pass, including new near-unity cases (0.99997 → F=0.99997, 0.99999997 → F=0.99999997, solving f=3e-5 → F=0.99997, true 1 → F=1)
• npm run lint: 0 errors; tsgo -b in packages/app: clean

🤖 Generated with Claude Code

[kateebonner]

What

formatFidelity (problem.ts) rendered with toFixed(4), so any $F \geq 0.99995$ displayed as a flat "F = 1" in the run chip, the ring-2 Run hero, and the Run modal — Kate hit this today on a transmon X-gate solve that finished at F ≈ 1 but almost certainly not exactly 1.

How

Adaptive precision in the one shared formatter: ordinary values keep the 4-decimal form (existing renders unchanged — F=0.9998 · 60 iter, solving… F=0.968), but when rounding would collapse the value to exactly 1 or 0, precision extends one digit at a time (capped at 10) until the gap survives. Only a true 1/0 renders bare. The gap from perfect is the result; the display must never erase it.

Applies automatically to every consumer of runChipText/runVerdict — chip, hero, modal.

Verification

• bun test src/amicode/problem.test.ts: 34 pass, including new near-unity cases (0.99997 → F=0.99997, 0.99999997 → F=0.99999997, solving f=3e-5 → F=0.99997, true 1 → F=1)
• npm run lint: 0 errors; tsgo -b in packages/app: clean

🤖 Generated with Claude Code

[kateebonner]

What

formatFidelity (problem.ts) rendered with toFixed(4), so any $F \geq 0.99995$ displayed as a flat "F = 1" in the run chip, the ring-2 Run hero, and the Run modal — Kate hit this today on a transmon X-gate solve that finished at F ≈ 1 but almost certainly not exactly 1.

How

Adaptive precision in the one shared formatter: ordinary values keep the 4-decimal form (existing renders unchanged — F=0.9998 · 60 iter, solving… F=0.968), but when rounding would collapse the value to exactly 1 or 0, precision extends one digit at a time (capped at 10) until the gap survives. Only a true 1/0 renders bare. The gap from perfect is the result; the display must never erase it.

Applies automatically to every consumer of runChipText/runVerdict — chip, hero, modal.

Verification

• bun test src/amicode/problem.test.ts: 34 pass, including new near-unity cases (0.99997 → F=0.99997, 0.99999997 → F=0.99999997, solving f=3e-5 → F=0.99997, true 1 → F=1)
• npm run lint: 0 errors; tsgo -b in packages/app: clean

🤖 Generated with Claude Code