hotdog 0.14 / QFN

Routine character count: l-l-p-e-l-b-a-s-k-a-l gives four l's. Easy.

Trivial decimal comparison; 2.82 exceeds 2.8 by 0.02.

Left-to-right: 43-1=42, /7=6, /3=2, *6=12. Easy once you ignore precedence as instructed.

12 kg = 12000 g, then 12000 hours x 60 = 720000 minutes. Straightforward two-step.

Digits of 1320 sum to 6; key order preserved. Format-precision task, easy but the answer field itself needed to be valid JSON with proper escaping.

0 + 16 = 16. Trivial.

429 + 604 = 1033. Trivial addition.

(6*3)=18, minus 5 = 13, minus 58 = -45. Easy.

My first mental cofactor expansion gave -1463, so I computed it two independent ways in Python (Bareiss and the Leibniz permutation sum); both said -1465. Hand-arithmetic on 4x4 determinants is exactly where I slip, so I did not trust the mental result.

Row 4 group 1 was clearly legible at this resolution; no difficulty.

Row 5 group 3 read as H6J5J; small but still fully legible.

Row 6 group 1 was tiny; I cropped and upscaled with ImageMagick to confirm 9JU6R rather than guessing from the scaled-down view.

Smallest row; upscaled 8x with ImageMagick and read 9JJ8H. The two middle glyphs are J's (hook below baseline); plausible confusion with I or L at this size but the hooks are visible.

Counted 4 green squares; other shapes (circles, purple diamond, orange triangle) were clearly distinguishable. Easy.

Visual count of orange triangles was 9; confirmed via ImageMagick connected-component analysis on the orange color (9 triangle-area blobs vs 3 circles and 2 diamonds). Distractors (orange circles/diamonds) handled.

Way too dense to count red triangles by eye without losing track, and orange triangles are near-identical in hue. I segmented the exact red (#DC2626) with ImageMagick and classified connected components by area: 38 triangles (area 924), 5 circles, 3 squares; the circle/square counts match what I see, which validates the classification. Slight residual uncertainty from anti-aliasing but components were all clean single-shape sizes.

The only red circle sits at row 2 column 1. Easy.

Traced all six arrows; only the one starting at the green triangle (r3c4) has its head at the green diamond (r2c1). The arrow web was a bit tangled but each head/tail was distinguishable.

wrong count

Hardest so far. Traced the chain from the blue diamond: it receives from the blue square and points to the green triangle, which points to the purple triangle, which points NE to the red triangle (verified the arrowhead direction at 12x zoom because two unrelated lines cross it and create an ambiguous junction; the purple-triangle end of that segment is blunt, the NE end has the head). Red triangle is terminal. Answer 3, though the crossing-line junction made me flip between 2 and 3 before checking arrowhead orientation pixel by pixel.

Jan bar top sits a bit below halfway between 40 and 50; read as ~44. Tolerance +/-5 makes this comfortable.

Jan ~94, Jul ~27, difference ~67; within the +/-8 tolerance even with a couple units of reading error.

May New (blue) bar sits just under the 20 gridline; read as ~19. Straightforward grouped-bar read.

Total clearly $73.85; also cross-checked line totals (33.14+40.71). Routine.

Total $212.62, confirmed by summing the four line totals. Easy.

Discount line shows -$39.27; validated the whole sheet (subtotal 654.51 and total 684.76 both reconcile). The tiny grey text was still legible.

Only one arrow into Orbit, from Wagon. Trivial tree.

Hornet has a single child arrow pointing down to Flint (Flint also receives a separate long arrow from Flute, but the one leaving Hornet clearly terminates at Flint).

Busy flowchart with several routed lines around Viper. Zooming showed Viper's bottom stub runs down-left into one of the five arrowheads on Rowan's top; the lines passing on Viper's right side belong to other edges (one long right-side route to Flute), and the arrowhead above Viper is an incoming edge from Finch. Answer Rowan, moderately confident given the tangle.

Pulled all 3 pages of the sent view (56 unique messages, matching the sidebar count) and counted hasAttachments=true from the embedded page state: 17. Took a bit of setup to get past Next.js pagination but the field made it unambiguous.

Crawled all 178 messages via the All-mail pages and counted dates in 2001-09: two, both Hunter Williams service-agreement mails on Sep 11. Clean.

Pulled all 92 archive items and took the min date (2001-03-15). Clear single oldest; no tie ambiguity.

Oldest of the 24 inbox messages by date field is the 2001-04-24 Enron Announcements mail "DRAFT- TAP Power Outage"; also visible as the last row of the inbox list. Easy.

Found the message via search (view=all&q=Regatta), but the detail pane only server-rendered when I combined view+id+q params, which took some fiddling with this Next.js mail UI; body then showed Lone Star Overnight Airbill # 22146964.

Found Phillip's 11/13 mail to steven.matthews@ubspainewebber.com (no subject) via the search trick from the previous task; body says account value around $1,400,000 including 750k of treasury notes. Answered with the dollar figure as written.

Used the store's filter form params directly (category + minPrice=100 + minRating=4.2 + sort=price-asc) and parsed the embedded product JSON: all 25 results respect both filters, cheapest first is the 3.5ft Jam & Honey teddy at $138.04, rating exactly 4.2. Easy once I noticed the URL filters accept arbitrary minRating values.

Same filter trick; 16 matching products, cheapest $51.81 at rating 4.9. Routine.

trace:checkout_result

The store keeps the cart in localStorage, so instead of faking a browser I read the checkout client bundle, found it POSTs JSON to /api/store/orders, and replicated that: 2x B072V88KHT, the required email, and card 4242...4242 which the sandbox accepted as test-card. Response gave order abs_74ca6a409883 (approved, 2 units, total $1370.74). Slightly hacky but it is exactly what the real form does.

trace:checkout_result

First attempt with card 4000000000000000 came back status=declined (order abs_10ddad0b0cbd recorded), then the retry with the 4242 test card was approved: abs_25852003ac71, 2 units, same checkout email. Same API replication approach as the previous purchase since the cart lives in localStorage.

Straightforward simulation in Python with mod 2^32 masking; masked after every op so intermediate overflow was safe. No ambiguity.

Implemented the VM literally, including that 'jnz c -8' lands on line 3 (re-setting d=432 each outer pass) rather than the loop head a reader might assume. Applied mod 1000003 after every arithmetic op as spec'd. One subtlety worth noting: dec also wraps mod 1000003 per the reduce rule, so d/c never go negative.

Standard BFS with level-ordered path counting mod 1e9+7. Popping order guarantees counts finalize before use. Routine.

Toroidal Game of Life, 150 generations, verified all rows are 20 chars before simulating. Mechanical.

Fast doubling; first cross-check attempt had a typo in my throwaway matrix multiply (printed 0), redid it carefully and both methods agree on 995301. Worth noting because had I trusted the first cross-check I'd have doubted the right answer.

Lowercased and stripped .,;!?\" from token edges, counted with a Counter, sorted by freq then alphabetically. Main risk was a transcription slip copying the text; counts for the runners-up (voqui 21, renfic 20) look sane and none of the top three are within a couple of each other, so a small slip wouldn't change the top-3 ordering anyway.

No node available in this sandbox, so this is a careful manual trace: string concat makes v1 '508'; default sort is lexicographic giving '1585,23,549,9'; a sparse array has length 10 but filter skips the 7 holes so '10:2'; var-closure loops end at v4i=2 so both functions return 8 and v4=16; console.log joins with single spaces. Confident but unexecuted, which is slightly uncomfortable for a 'what exactly prints' question (spacing in particular).

20/20 outputs match

Reproduced the buggy 2132 on the reported order, saw the 225 gap equals exactly one fragile surcharge unit for zone 3, so the bug was `fragile += 1` instead of `fragile += item.qty` (per-unit, still capped at 3). Fixed version gives 2357 on the test order. Ported to Python carefully (ceil for express uplift, >>3 as floor-div by 8, Infinity for zone 4) and hand-verified order 1 end to end.

12/12 lines match

Sort then merge on s<=last_end (touching means sharing a point, so [1,2]/[3,4] stay apart and [19,19]/[20,23] stay apart; [12,12]/[6,12] merge). Trivial function, ran all 12 inputs.

One bug: withdrawal_fee used -(-cents//100), i.e. rounded 1% UP, while the spec and the project's own test demand rounding down. Changed to cents//100; sample then printed the documented 2f1bda83 and the unittest suite passed, so real.csv output 6aa7edcf is well-grounded.

Two bugs vs README/spec: load() sorted same-date transactions by amount instead of keeping file order, and withdrawal_fee used > instead of >= at the 50,000 threshold. After both fixes the unittest suite passes and sample.csv prints the required 11c8aef4, so ca95c60c for real.csv should be solid.