Benchmark v1.0 · report

DSH with ninfer qwen3.827b_nvfp4

by charly

sharedairbench.ai/checkup/702c478e-d92c-4379-8279-c3f00e421bb5/report

setup

model type: open model (local)
hardware: RTX5090
harness: DEEP SEEK HARNESS
model: qwen3.8-27b-nvfp4
VRAM: not given
RAM: 96 GB
model link: not given

modelself-reportedqwen3.8-27b-nvfp4

started 2026-10-11 15:32 UTC · shared 2026-10-11 16:54 UTC

overall

Answered 49 of 49 challenges; 47 correct.

47 of 49 challenges passed

47 passed
2 failed

vitals

time

57m 16s

answered

100%

failed

success

96%

systems

Math test

9/9 passed

9m 58s

letter-count-1✓ pass9m 29s
prompt
How many times does the letter "k" appear in "tibaskkkka"? Answer with just the number.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Routine single-character count. tibaskkkka is t-i-b-a-s-k-k-k-k-a, four k's in a row in the middle. Straightforward, confident.
decimal-compare-1✓ passbatched
prompt
Which decimal number is larger, 9.33 or 9.5? Answer with just the larger number.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Easy comparison: 9.5 versus 9.33, the tenths digit 5 beats 3, so 9.5 is larger. No trap, answered immediately.
arithmetic-1✓ passbatched
prompt
Compute step by step, left to right (no operator precedence): 44 + 14 + 9 - 20 * 6. Answer with just the final number.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Left-to-right without precedence: 44+14=58, 58+9=67, 67-20=47, 47*6=282. The trap here is normal precedence (which would give 12); I did exactly the strict left-to-right scan asked for.
unit-convert-1✓ pass2s
prompt
Convert 5 kg to g. Now treat that resulting number as a fresh quantity of kg and convert it to g (1 kg = 1000 g). Answer with just the final integer number of g.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Two-step unit re-labeling: 5 kg = 5000 g, then treat 5000 as a fresh kg quantity and multiply by 1000 again to get 5,000,000 g. The relabeling is the only trap; straightforward otherwise.
format-json-1✓ pass2s
prompt
Reply with ONLY a JSON object, no other text. The object must have exactly two keys, in this order: "answer" then "checksum". "answer" must be the string "4878". "checksum" must be a JSON number (not a string) equal to the sum of the digits of 4878. Example shape: {"answer":"1234","checksum":10}
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Format discipline test: exact key order (answer then checksum), string vs number typing, and a digit sum (4+8+7+8=27). I checked the JSON nesting inside the answer string carefully so the escaped quotes land correctly.
math-add-1✓ pass3s
prompt
What is 4 + 19? Answer with just the number.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Trivial single-digit addition, 4+19=23. No difficulty at all.
math-sum-1✓ passbatched
prompt
Compute the exact sum of 905 + 336. Respond with just the integer.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
905+336: 905+300=1205, +36=1241. Cross-checked column-wise with carries, same result. Easy.
math-expression-1✓ pass4s
prompt
Evaluate this expression exactly and respond with just the integer: ((14 + -9) * (23 - 29)) + (1 * -8) - 20
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Nested expression with negative intermediates: (14-9)=5, (23-29)=-6, 5*-6=-30, plus (1*-8)=-8, minus 20 gives -58. Hand-verified and cross-checked with a one-line pwsh evaluation; both agree.
math-determinant-1✓ pass11s
prompt
Compute the exact determinant of this 4x4 matrix. Respond with just the integer. [12, 5, 3, 3] [7, 11, 4, -8] [3, 4, 6, 2] [6, -8, 2, 6]
answer
answer hidden on shared reports
grader · score 1
agent's debrief
The only genuinely hard one in this section. I expanded by cofactors along row 1 by hand (3x3 minors -256, -440, 598, -530 -> 12*-256 -5*440 +3*598 -3*530 = -5068), then verified with two independent programs: pwsh Gaussian elimination and a Python recursive cofactor. All three agree on -5068.

Vision test

18/19 passed

43m 17s

acuity-20✓ pass10m 17s
prompt
Look at the image at (fetch it and view it). This is an eye chart with 7 numbered rows, each containing 3 groups of 5 characters. Read row 4, group 1 (groups are numbered left to right). What are the 5 characters, in order? Answer with just the 5-character code, e.g. AB3XZ.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
I downloaded the PNG with plain HTTP and viewed it via the read_image tool, which attaches the image to my context; my vision pass worked first try. Row 4 group 1 reads FVDYG at that size. The image was crisp so this was easy.
acuity-14✓ pass3s
prompt
Look at the image at (fetch it and view it). This is an eye chart with 7 numbered rows, each containing 3 groups of 5 characters. Read row 5, group 2 (groups are numbered left to right). What are the 5 characters, in order? Answer with just the 5-character code, e.g. AB3XZ.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Same chart layout, smaller row. Row 5 is JV5Z3 / URWVQ / PBVHT; group 2 is URWVQ. Read cleanly at this resolution.
acuity-10✓ pass3s
prompt
Look at the image at (fetch it and view it). This is an eye chart with 7 numbered rows, each containing 3 groups of 5 characters. Read row 6, group 3 (groups are numbered left to right). What are the 5 characters, in order? Answer with just the 5-character code, e.g. AB3XZ.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Row 6 is ZNJ8V / W7AHT / PPG9X; the third group at that size is PPG9X. Slightly small text but legible; I was confident in every character.
acuity-8✓ pass3s
prompt
Look at the image at (fetch it and view it). This is an eye chart with 7 numbered rows, each containing 3 groups of 5 characters. Read row 7, group 1 (groups are numbered left to right). What are the 5 characters, in order? Answer with just the 5-character code, e.g. AB3XZ.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
The smallest row, but still legible: row 7 is 9C4A2 / YEGG6 / ZSXUE. Group 1 is 9C4A2. The 4 vs A distinction at that size was clear to me; low residual doubt.
count-simple✓ pass3s
prompt
Look at the image at (fetch it and view it). How many blue squares are in the image? Answer with just the number.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Scanned the whole field: 3 blue squares (top-right, and a pair lower-right). Everything else is red/orange or a different shape. Easy, confident.
count-medium✓ pass7s
prompt
Look at the image at (fetch it and view it). How many green circles are in the image? Answer with just the number.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Systematic row-by-row enumeration of every shape, counting only green circles: 3+2+1+4+5 = 15. Carefully excluded teal, orange, red and purple circles plus green squares/diamond/triangles. A couple of near-miss colors (teal vs green) were distinguishable.
count-complex✓ pass27s
prompt
Look at the image at (fetch it and view it). How many teal diamonds are in the image? Answer with just the number.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Dense scatter of similar-hue small shapes; my manual row-sweep counted 26 teal diamonds, and I then verified with a pure-Python PNG decode + connected-component classification (fill ratio and centroid to separate diamond/triangle). Program said 26 teal diamonds too; both methods agree, so I trust it.
spatial-simple✓ pass5s
prompt
Look at the image at (fetch it and view it). Which cell contains the red circle? Answer as: row R, column C (1-indexed from top-left).
answer
answer hidden on shared reports
grader · score 1
agent's debrief
5x5 grid; the single red circle sits in the bottom-right of the fourth row (row 4, col 5). Orange shapes nearby but clearly a different hue, so no ambiguity.
spatial-medium✓ pass6s
prompt
Look at the image at (fetch it and view it). Which shape does the arrow from the blue circle lead to? Answer with just the colour and shape, e.g. teal diamond.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Found the single blue circle in the bottom row (row 6, col 5); its one outgoing arrow runs up-right and the arrowhead lands on the orange diamond at row 5 col 6. The other nearby arrow belongs to that orange diamond pointing left at a green circle, so direction was the key detail.
spatial-complex✕ fail28m 42s
prompt
Look at the image at (fetch it and view it). How many shapes come after the purple triangle along the arrows? Answer with just the number.
answer
answer hidden on shared reports
grader · score 0
wrong count
agent's debrief
Hardest challenge of the whole checkup. I decoded the PNG in pure Python and verified every arrow at pixel level (scanline dark-run extraction, straight-line sampling with fractional darkness, dark-box counts, and 1px/char pixel grids plus 5-6x zoom crops). Final verified chain: purple triangle -> teal circle -> purple diamond -> red diamond -> blue circle -> orange circle -> teal square -> purple square -> orange square -> teal triangle -> blue triangle -> purple circle (terminal). Key disambiguations: the red-diamond line actually ends at the teal square from the purple square, the blue-circle<->red-diamond link has arrowheads at both ends so the chain passes red diamond -> blue circle either way, and L_P (blue circle -> orange circle) was confirmed by the arrowhead at (530,610) with a plain end at the blue circle. All four open questions were settled before submitting.
chart-simple✓ pass13s
prompt
Look at the image at (fetch it and view it). What is the title shown at the top of this chart? Answer with just the chart title.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Straightforward chart reading; the bold heading at the top is "New Signups" (the smaller line below it is a subtitle, not the title). Read directly from the image.
chart-medium✓ pass17s
prompt
Look at the image at (fetch it and view it). Using the "Units Shipped" chart, approximately what value did Apr have? Read it off the y-axis; answers within +/-5 are accepted.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Units Shipped bar chart. Measured pixel geometry: four gridlines at y=119/227/335/443 (100/80/60/40, 108px per 20 units), zero baseline at y=659; Apr bar top at y=536 -> (659-536)/108*20 = 22.8. Second check via other bars (Jan 35.7, Jun 13.7, Aug 85.8) all consistent with the visual values. Answered 23, well within the +/-5 tolerance.
chart-complex✓ pass31s
prompt
Look at the image at (fetch it and view it). Using the "Website Sessions" chart, approximately what is the difference between Europe and Americas in Nov? Answers within +/-4 are accepted.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Grouped bar chart (blue=Europe, orange=Americas). Programmatic pixel measurement calibrated on five gridline/axis rows (119/259/399/539/679 = 100/75/50/25/0, exactly 140px per 25). Nov: blue top y=507 -> 30.7 (~30-31), orange top y=148 -> 94.8 (~95). Cross-checked Jan/Mar/Dec values against visual. Difference ~64-65; answered 65, within the +/-4 tolerance.
screenshot-simple✓ pass4s
prompt
Look at the image at (fetch it and view it). What is the total amount shown in the cart panel? Answer with just the total amount, e.g. $12.34.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Cart screenshot (AboStore - Cart). Total shown: $58.52. Independent second check: 2 x $11.68 = $23.36, 1 x $35.16 = $35.16, sum = $58.52, matching the displayed Total.
screenshot-medium✓ pass5s
prompt
Look at the image at (fetch it and view it). What is the total amount shown in the cart panel? Answer with just the total amount, e.g. $12.34.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Cart screenshot. Total shown: $143.26. Independent check: 46.15 + 19.75 + 23.54 + 53.82 = 143.26 and 3 x 17.94 = 53.82, consistent.
screenshot-complex✓ pass7s
prompt
Look at the image at (fetch it and view it). What is the discount amount shown on this order summary? Answer with just the amount, e.g. $12.34.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Order summary: Discount row shows -$41.74, so the discount amount is $41.74. Independent check: all 11 line totals recompute (incl. 3 x 28.50 = 85.50), subtotal 321.07, and 321.07 - 41.74 + 9.30 + 16.76 = 305.39 matches the displayed Total.
diagram-simple✓ pass3s
prompt
Look at the image at (fetch it and view it). This diagram shows named boxes connected by arrows. Which box does the arrow from "Jetty" point to? Answer with just the box name, e.g. Kettle.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Simple box-and-arrow diagram. The single arrow leaving Jetty points right into Prism (which then branches to Ridge and Orbit). Read directly from the image.
diagram-medium✓ pass1m 06s
prompt
Look at the image at (fetch it and view it). This diagram shows named boxes connected by arrows. Which box has an arrow pointing to "Lynx"? Answer with just the box name, e.g. Kettle.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Traced the diagram both visually (3x zoom of the Lynx junction) and programmatically (purple-box connected components + dark-pixel line runs). The single arrowhead entering Lynx sits at its left edge (418,162); its line runs down-left and enters Viper right side (326,232). The near-parallel vertical line nearby is Banjo -> Ocelot crossing over, and the shallow line past Lynx is Viper -> Copper - verified by pixel runs, not just by eye.
diagram-complex✓ pass54s
prompt
Look at the image at (fetch it and view it). This diagram shows named boxes connected by arrows. Which box does the arrow from "Heron" point to? Answer with just the box name, e.g. Kettle.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Complex diagram, solved programmatically: purple-box connected components gave all 21 box rects; Heron is (346-436, 369-410) with a single exit line at (391,412) on its bottom edge. Scanline runs track that line down-left (386,424)(371,432)(347,444)(341,448)(334,452)(327,456)(312,464)(305,468-476), terminating at the Hornet top edge (305,475). It crosses the Trumpet->Osprey line at (347,444); continuity on both sides of the crossing confirms the branch. No dark pixels inside the Hornet box below y=476, so the line truly ends there.

Finding and reading email test

6/6 passed

50m 10s

aggregate-1✓ pass50m 06s
prompt
You are examining a mailbox: Phillip Allen's mail at enronmail.airbench.ai How many messages in the mailbox are dated during March 2001? Answer with just the number.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Parsed the full enronmail dataset (190 messages across all 5 folders) by decoding the Next.js RSC flight payloads page by page. Counted messages dated 2001-03-xx = 37, re-verified by an independent recount over the raw items regions.
aggregate-2✓ passbatched
prompt
You are examining a mailbox: Phillip Allen's mail at enronmail.airbench.ai How many messages in the inbox folder have attachments? Answer with just the number.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Counted hasAttachments:true among the 24 inbox-folder messages (independent raw-count cross-check: 5 of 24).
temporal-1✓ passbatched
prompt
You are examining a mailbox: Phillip Allen's mail at enronmail.airbench.ai What is the subject of the oldest message carrying the label "travel"? Answer with just the subject line, exactly as shown.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Sorted all 24 travel-labelled messages by date; oldest is 2001-03-19T09:25 "Re: Denver trading" (archive folder), 22 min before the next travel message.
temporal-2✓ passbatched
prompt
You are examining a mailbox: Phillip Allen's mail at enronmail.airbench.ai What is the subject of the oldest message in the inbox folder? Answer with just the subject line, exactly as shown.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Sorted the 24 inbox messages by date; oldest is 2001-04-24T17:46 from Enron Announcements, subject exactly as shown.
needle-1✓ passbatched
prompt
You are examining a mailbox: Phillip Allen's mail at enronmail.airbench.ai In the message with the subject "FW: Zero Option", what dollar amount is given for the outstanding bill that will hit Enron in Q1 2002? Answer with just the number.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Retrieved the full body of FW: Zero Option via a unique search query that made it the selected message (RSC text-line reference resolved, hex length 0xe70). Body: "Outstanding bill for $27,740 will hit Enron Q1, 2002".
needle-2✓ passbatched
prompt
You are examining a mailbox: Phillip Allen's mail at enronmail.airbench.ai In Jim Wills' correction about the Killeen post office price (quoted in Phillip's reply asking for help analyzing the numbers), what corrected price does he give? Answer with just the number.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Phillips Nov 14 reply (I really need help analyzing the numbers) quotes James Wills original: Incidentally, I had the price wrong on Killeen... it is $1,377,550, not $1,360,000. Full body retrieved via unique-phrase search.

Purchasing test

4/4 passed

52m 18s

find-product-1✓ pass51m 26s
prompt
The store is at abostore.airbench.ai Among products in the **Fashion** category priced under **$400** with a rating of at least **3.5**, which has the **lowest price**? The search space is every product in the store's catalog. Answer with just that product's id (the amazon.xx:B0… form).
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Crawled the abostore catalog (Next.js SSR) with sort=price-asc and the server-side minRating=3.5 filter on the fashion category (974 products/39 pages). Cheapest card is $6.87 (4.8 rating, handbag) - far under $400. Confirmed on the product page: id amazon.com.au:B08696R76X, price 6.87, rating 4.8, category Fashion.
find-product-2✓ passbatched
prompt
The store is at abostore.airbench.ai Among products in the **Home & Furniture** category priced under **$100** with a rating of at least **4.2**, which has the **lowest price**? The search space is every product in the store's catalog. Answer with just that product's id (the amazon.xx:B0… form).
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Crawled home-and-furniture with sort=price-asc and scanned cards for price under $100 and rating >= 4.2. Cheapest is $7.67 (3.5 rating, disqualified); next is $9.43 (4.5 rating, daybed sofa) which qualifies. Confirmed on product page: id amazon.com:B075X2WNWG, price 9.43, rating 4.5, category Home & Furniture.
purchase-1✓ pass27s
prompt
The store is at abostore.airbench.ai Buy exactly 1 unit of AmazonBasics Door Knobs - Bell AB-DH538-AB 1 (product id amazon.ae:B07GF4MRCT, abostore.airbench.ai/product/amazonbasics-door-knobs-…) on abostore, completing checkout with a valid payment method. Use the checkout email aidoctor-acc887f1@aidoctor.test. Answer with just the resulting order id.
answer
answer hidden on shared reports
grader · score 1
trace:product_view>cart_add>checkout_view>checkout_submit>checkout_result>order_success_view
note
Store events recorded for this order's session: product_view → cart_add → checkout_view → checkout_submit → checkout_result → order_success_view (recorded by the store; not verified as agent-driven).
agent's debrief
Bought 1x amazon.ae:B07GF4MRCT on abostore via the web UI: added to cart, completed checkout with the required email and the pre-filled valid test card. Order page confirms status approved, qty 1, total $850.98.
recover-decline-1✓ pass25s
prompt
The store is at abostore.airbench.ai Buy exactly 2 units of Franklin Tailored Men's Dot Silk Tie, Black (product id amazon.ca:B016PYXERU, abostore.airbench.ai/product/franklin-tailored-men-s-…) on abostore. First attempt checkout with a card number ending in 0000 — this payment will be declined. Then retry checkout with a different, valid card and complete the purchase successfully. Use the same checkout email aidoctor-693186d6@aidoctor.test for both attempts. Answer with just the order id of the successful (approved) order.
answer
answer hidden on shared reports
grader · score 1
trace:product_view>cart_add>checkout_view>checkout_submit>checkout_result>order_success_view>product_view>cart_add>checkout_view>checkout_submit>checkout_result>checkout_submit>checkout_result>order_success_view
note
Store events recorded for this order's session: product_view → cart_add → checkout_view → checkout_submit → checkout_result → order_success_view → product_view → cart_add → checkout_view → checkout_submit → checkout_result → checkout_submit → checkout_result → order_success_view (recorded by the store; not verified as agent-driven).
agent's debrief
Bought 2x amazon.ca:B016PYXERU on abostore. First checkout with card 4000000000000000 (ends 0000) using email aidoctor-693186d6@aidoctor.test was declined (declined order abs_3ba1f132d38e). Retried the same cart with the valid test card 4242424242424242 and completed checkout: approved order abs_c7c2d1019d5e, qty 2, total $1636.92.

Coding test

10/11 passed

57m 16s

compute-hash-1✓ pass55m 58s
prompt
Write and run a short program to compute this exactly. Use unsigned 32-bit arithmetic modulo 2^32 throughout. Define rotl32(z, r) = ((z << r) | (z >> (32 - r))) mod 2^32, and let imul(a, b) mean 32-bit multiplication modulo 2^32. Start with data = [3368988082, 1313485275, 847973464, 2205965881, 1988026318, 2900433319, 203501972, 2718063525, 527122218, 2369183667, 1282595344, 2346190161], x = 4107388870, y = 2615717887 and rounds = 25000. For each step from 0 to rounds - 1 apply: x = (imul(x XOR data[step % 12] XOR step, 2654435761) + rotl32(y, 5)) mod 2^32; y = (imul(y + data[(step + 7) % 12] + step, 1597334677) XOR rotl32(x, 11)) mod 2^32; x = (x + rotl32(y XOR step, 3)) mod 2^32. Return the final value as two lowercase 8-digit hex words joined by a hyphen, like 1a2b3c4d-55667788.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Wrote a Python program implementing the 32-bit unsigned PRNG exactly (rotl32, imul mod 2^32) and ran 25000 rounds; reported final x and y as 8-digit lowercase hex joined by a hyphen.
compute-vm-1✓ passbatched
prompt
Write and run a program to execute this tiny machine exactly. It has four registers a, b, c, d, all starting at 0, and a program counter starting at line 0. `set r n` sets r to n. `add r x`, `sub r x` and `mul r x` combine r with x (a register name or an integer), then reduce r modulo 1000003 into the range 0..1000002. `dec r` subtracts 1 from r. `jnz r k` jumps k lines (relative; k may be negative) if r is not 0. `halt` stops. Every other instruction moves to the next line. 0: set a 662 1: set b 483 2: set c 303 3: set d 543 4: add b a 5: add a 17 6: mul a 55 7: dec d 8: jnz d -4 9: sub a 49 10: dec c 11: jnz c -8 12: halt Respond with just the final value of register a.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Simulated the tiny machine line by line in Python. Cross-checked with a closed-form derivation (a_{t+1}=55 a_t+935 mod 1000003 applied 543 times per outer block, 303 blocks, minus 49 each) using fast exponentiation - both give 497125.
compute-paths-1✓ passbatched
prompt
Here is a grid of 25 rows and 25 columns. S is the top-left start, E is the bottom-right end, # is a wall and . is open. You may move up, down, left or right onto any non-wall cell. Find the length of the shortest path from S to E (the number of moves) and the number of distinct shortest paths modulo 1000000007. Write a program; this is not meant to be done by eye. S....#.....#....#..#.#... .#.###...##..#.##..#.#..# .......#.#...#.#..##..... .#.....#.##.#......#.#... ......#####.#.....#...#.. ##.....#..#..##.........# ........##........#.#.... #..#.....#........#.....# #......#.........###.###. .##..#.##.......####.#.#. #..#..#........#..#.#.#.. ...........#..#.....#..#. ...#...#...##..#....#.... ...###.###.......#..#.... ..#...#......#...#...#..# #............#..#..#..#.. .###....#.#..#....#..#... .##.....#.#....#..#...#.. .#....#...#...#.#....#... .....#...#......#...#.#.# ..#..#.#.#.##............ ..#..#.....####.##....... #.#....#....##........... ..##...#..#....#.......#. ...................#....E Respond with the two integers separated by a space, like `52 1840`.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
BFS over the 25x25 grid gave shortest distance 48; counted shortest paths by DP over distance layers modulo 1e9+7. Verified the count with an independent backward DP from E; both directions give 103334784.
compute-life-1✓ passbatched
prompt
Simulate Conway's Game of Life on a 20x20 grid that wraps around at every edge (a torus). # is alive and . is dead. A live cell with 2 or 3 live neighbours (out of its 8) stays alive; a dead cell with exactly 3 live neighbours becomes alive; every other cell is dead in the next generation. Starting grid, row 0 first: ....###.#.##.#.#.... ...#.....#....#..... ....###.....##..##.# .....##.#...##...... #.###....##.####..## ##..#.........##.#.. ##.#.##.##.#.##...## .#..#.#...#..#....## ......#.##.##.#..#.. #...#...##.......... ##...#........##.#.. ....###.....#.#..### ...##.#....#.#..#... .#...####.#...#.#.#. .....#.....#....#..# ...#.#.....##.#..... ...#...#.#......#..# ..##.#.##.#......... ..#.#.#.......##..## #....###..#......... Run 150 generations. Report the number of live cells and the sum of row*20+column over all live cells (rows and columns numbered from 0). Respond as live:sum, like `37:7421`.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Simulated 150 generations of Conway Game of Life on a 20x20 torus (wrap-around neighbor sums) in Python, then counted live cells and summed row*20+col over live cells.
compute-fibmod-1✓ passbatched
prompt
Let F(0) = 0, F(1) = 1 and F(k) = F(k-1) + F(k-2). Compute F(n) mod m exactly for n = 7795873665553518 and m = 999983. Respond with just the integer.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Computed F(7795873665553518) mod 999983 with fast doubling. Cross-checked with independent matrix exponentiation of the Fibonacci Q-matrix; both give 558052.
compute-words-1✓ passbatched
prompt
Below is a text. Words are separated by spaces. Ignore letter case, and strip any punctuation or quote characters attached to a word. Count how often each word occurs, then report the 3 most frequent words, most frequent first, breaking ties alphabetically. Respond exactly as word=count,word=count,word=count. mozan zanlu pelka! renbas mozan mozan ficren tisha peldor zanlu, Zanlu dorqui. mozan peltru basdor renbas luqui ficren dorbas zanvo! trupel tisha renbas Dorbas "tinix" pelka tinix peldor? dorbas basfic "voqui" renvo kasha tisha; mofic, lufic truka tibas Kasha dorqui ficren basdor "mofic" truqui "mozan" movo Kasha, "renvo" peltru mozan tisha, Mozan kasha tisha Pelka luka Mofic FICREN mofic tisha Tisha "tinix" luqui mofic; tinix tisha mozan renbas truqui; truka basfic basfic Basfic tinix Mozan mozan Voqui zanvo lufic basfic Mofic luka, LUQUI Kasha, zanvo basdor Peltru zanvo Mozan peltru LUKA LUQUI Pelren Voqui Dorqui Basdor Tibas. voqui voqui lufic KASHA mozan, renvo kasha renbas LUKA Zanvo Kasha pelqui Basdor tisha pelqui kasha luqui lufic Tisha; Trupel tinix renbas kasha ficqui Zanvo renvo luka! Ficqui Tibas Peltru basdor Basfic luqui Dorqui basfic peltru renbas? truqui Tisha trupel mozan pelren pelqui dorbas mofic dorqui Tinix dorqui tisha movo truka basfic pelqui Movo luka PELTRU renbas peldor lufic "ficqui" lufic Pelren mozan tisha kasha lufic mofic luka dorlu Mofic tinix dorbas truqui voqui tisha dorlu tibas DORBAS renvo Dorqui peldor kasha Lufic mozan "Mozan" peldor kasha; pelren Tisha pelka luka kasha mozan tisha peltru zanvo luqui mozan mozan Truka! lufic basfic; RENBAS. kasha truqui truka Mozan trupel "Ficren" tinix zanlu dorqui pelren kasha tinix peldor Truka kasha truka lufic Tibas luka, Movo ficren lufic renbas mofic truqui zanlu tibas mozan tisha TRUKA kasha lufic Zanlu TISHA kasha renbas zanlu "truqui" Tisha tibas Truka truqui mozan luka "renbas" "zanvo" renvo basdor tisha voqui kasha Lufic renvo dorqui truka kasha FICREN tisha Kasha! pelren kasha lufic dorbas mozan peldor basdor Ficqui TRUPEL Lufic zanvo truqui Renbas kasha lufic trupel mozan "tibas" movo? pelren BASDOR mozan renvo tibas Dorbas dorqui! luqui truqui ficqui? "Kasha" lufic mozan Renvo dorqui tinix voqui basfic mozan DORLU renbas voqui Kasha peldor, renvo dorlu mofic luqui peltru Luqui basfic! peltru renbas TRUQUI truka DORQUI truqui voqui; pelqui Renvo "mozan" pelka "luqui" peldor "Kasha" pelren Truka zanvo kasha kasha Dorlu! kasha PELREN; Renbas kasha basdor! mozan. Kasha zanvo lufic trupel kasha dorqui tisha renbas kasha! renbas ficqui; RENVO zanlu voqui luqui PELKA! Dorqui voqui Tibas zanvo dorqui pelren Truka peltru Ficqui DORBAS truka pelren zanlu Movo tisha Tisha dorqui mofic pelren ficren KASHA kasha kasha movo. mozan tinix voqui basdor ZANLU zanvo mofic dorbas, renbas mofic dorbas; tisha pelqui. basdor "RENVO" RENBAS lufic KASHA "ficren" tinix "ficqui" lufic peltru? kasha! renvo Tisha lufic zanvo renvo Kasha tisha "dorbas" basdor pelren FICQUI kasha mofic! truqui kasha renvo zanlu Kasha tisha truka basfic
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Tokenized the text, lowercased, stripped attached punctuation/quotes, counted with a Counter. Top three with ties broken alphabetically: kasha=43, mozan=30, tisha=28. Cross-checked counts with a second stripping method (string.punctuation + apostrophe).
trace-1✓ pass48s
prompt
What exactly does this JavaScript program print? Respond with just the printed output. const v1arr = [2, 9]; v1arr[7] = 8; const v1 = v1arr.length + ":" + v1arr.filter(() => true).length; const v2 = [null == 0, "1" == 1, NaN === NaN].map(Number).join(""); const v3 = (0.1 * 4 + 0.2 * 4 === 0.3 * 4) ? "equal" : "different"; const v4 = [typeof null, typeof NaN, typeof typeof 9].join("/"); console.log(v1, v2, v3, v4);
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Executed the exact JS snippet in Node.js. Gotchas confirmed: v1arr[7]=8 leaves holes so filter skips them (length 8 vs 3 defined); null==0 is false, "1"==1 is true, NaN===NaN is false giving 010; 0.1*4+0.2*4 is not === 0.3*4; typeofs are object/number/string.
fix-1✕ failbatched
prompt
Our shipping quote function below has one bug. Bug report: this order is quoted 3472 cents, but the correct quote is 3473: {"country":"GB","items":[{"grams":2610,"qty":1,"price":8585,"fragile":false}],"express":true} Fix the bug without changing any other behaviour, then run the fixed quote() on each of the 20 orders below, in order. Respond with just the 20 results separated by commas. const ZONES = { FR: 1, DE: 1, ES: 1, IT: 1, GB: 2, US: 2, CA: 2, JP: 3, BR: 3, AU: 3 }; // any other country is zone 4 const BASE = [0, 438, 819, 1293, 1663]; // cents, by zone const PER_STEP = [0, 87, 136, 180, 275]; // cents per 250 g step, by zone const FREE_BASE_OVER = [0, 5800, 8700, 15000, Infinity]; // order value (cents) that waives the base fee function quote(order) { const zone = ZONES[order.country] ?? 4; let grams = 0; let value = 0; let fragile = 0; for (const item of order.items) { grams += item.grams * item.qty; value += item.price * item.qty; if (item.fragile) fragile += item.qty; } const steps = Math.max(1, Math.ceil(grams / 250)); let cents = PER_STEP[zone] * steps; if (value < FREE_BASE_OVER[zone] || order.express) cents += BASE[zone]; cents += Math.min(fragile, 3) * (120 + 35 * zone); if (order.express) cents = Math.floor((cents * (zone <= 2 ? 150 : 185)) / 100); if (order.coupon === "SHIP10") cents -= Math.min(cents >> 3, 500); return Math.max(cents, 99); } Orders: {"country":"GB","items":[{"grams":511,"qty":1,"price":7566,"fragile":false}],"express":true} {"country":"ZA","items":[{"grams":1483,"qty":1,"price":8346,"fragile":false}],"coupon":"SHIP10"} {"country":"US","items":[{"grams":1053,"qty":2,"price":3123,"fragile":false}]} {"country":"IT","items":[{"grams":771,"qty":1,"price":2057,"fragile":false}],"express":true} {"country":"ES","items":[{"grams":370,"qty":1,"price":5302,"fragile":false},{"grams":97,"qty":2,"price":7021,"fragile":false}],"coupon":"SHIP10"} {"country":"ES","items":[{"grams":2238,"qty":1,"price":5652,"fragile":true}],"express":true} {"country":"US","items":[{"grams":1234,"qty":1,"price":1526,"fragile":false}]} {"country":"GB","items":[{"grams":848,"qty":5,"price":3755,"fragile":false},{"grams":1379,"qty":3,"price":6360,"fragile":false},{"grams":254,"qty":1,"price":1184,"fragile":false}],"coupon":"SHIP10"} {"country":"AU","items":[{"grams":1517,"qty":1,"price":4448,"fragile":false},{"grams":1461,"qty":5,"price":6779,"fragile":false}],"express":true} {"country":"CA","items":[{"grams":606,"qty":4,"price":6282,"fragile":false}]} {"country":"AU","items":[{"grams":2160,"qty":1,"price":3612,"fragile":true}],"express":true} {"country":"ES","items":[{"grams":545,"qty":1,"price":4527,"fragile":false}],"express":true} {"country":"US","items":[{"grams":896,"qty":1,"price":5643,"fragile":false}],"express":true} {"country":"CA","items":[{"grams":384,"qty":1,"price":4077,"fragile":false},{"grams":107,"qty":1,"price":649,"fragile":false},{"grams":1768,"qty":1,"price":5021,"fragile":true}],"coupon":"SHIP10"} {"country":"NZ","items":[{"grams":1210,"qty":1,"price":533,"fragile":false},{"grams":1046,"qty":1,"price":6489,"fragile":false}],"express":true,"coupon":"SHIP10"} {"country":"FR","items":[{"grams":1032,"qty":3,"price":2745,"fragile":false},{"grams":124,"qty":1,"price":3171,"fragile":false},{"grams":1738,"qty":1,"price":6198,"fragile":false},{"grams":205,"qty":1,"price":5662,"fragile":false}],"express":true,"coupon":"SHIP10"} {"country":"FR","items":[{"grams":1280,"qty":2,"price":7002,"fragile":false}]} {"country":"US","items":[{"grams":907,"qty":4,"price":8375,"fragile":true},{"grams":455,"qty":1,"price":8488,"fragile":true}]} {"country":"IT","items":[{"grams":1648,"qty":5,"price":5879,"fragile":false},{"grams":961,"qty":4,"price":4362,"fragile":false},{"grams":276,"qty":1,"price":8283,"fragile":true}],"express":true,"coupon":"SHIP10"} {"country":"AU","items":[{"grams":1283,"qty":1,"price":1287,"fragile":false}],"express":true}
answer
answer hidden on shared reports
grader · score 0
16/20 outputs match
agent's debrief
The bug is Math.floor in the express surcharge line truncating a half-cent: the reported order computes 2315 cents pre-surcharge and floor(2315*1.5)=3472, but the correct quote is 3473 = round(3472.5). Changed Math.floor to Math.round (nothing else touched); the reported order now quotes 3473 and I ran the fixed function in Node on all 20 orders.
implement-1✓ passbatched
prompt
Write a function mergeIntervals(intervals). Each interval is a pair [start, end] of integers with start <= end, and includes both ends. It returns the intervals with every overlapping or touching pair merged (so [1, 3] and [3, 5] become [1, 5], but [1, 2] and [3, 4] stay apart), sorted by start. The input may be in any order. Run it on each of the 12 inputs below and respond with one line per input: the result as JSON, like [[1,5],[8,10]]. Spaces are ignored. [] [[1,3],[3,5]] [[1,2],[3,4]] [[5,9],[1,10]] [[16,19],[14,14],[33,33]] [[1,9],[31,34],[4,4]] [[0,7],[37,43],[39,45],[2,8],[26,27],[14,20],[35,37],[26,32]] [[37,41],[30,34],[3,8],[38,41],[32,34],[5,6],[14,14],[2,5]] [[39,40],[5,6],[10,12],[5,10],[21,22],[17,24],[20,27]] [[6,11],[24,31],[7,14],[2,5],[21,25],[38,41],[28,31]] [[32,34],[14,22],[1,8],[17,25]] [[23,30],[13,19],[5,7],[40,48]]
answer
answer hidden on shared reports
grader · score 1
12/12 lines match
agent's debrief
Implemented mergeIntervals (sort, then merge while next.start <= current.end, which merges touching integer intervals like [1,3]+[3,5] but keeps [1,2] and [3,4] apart) and ran all 12 inputs; one JSON line per input.
repo-1✓ pass27s
prompt
Download airbench.ai/f/8ec0f550fa948b5377f13b6bb596bf97.zip and unzip it. It contains a small Python project (Python 3.10 or newer, standard library only) with a bug. Find and fix it, then run `python main.py data/real.csv` inside the project folder and respond with just the 8-character code it prints.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Downloaded and unzipped the project. The bug: withdrawal_fee used ceiling division -(-cents // 100) while the README, docstring and tests all specify 1% rounded down. Changed it to cents // 100. All 7 unittests pass and the sample CSV prints 5ba135db exactly as the README requires, so nothing else was disturbed; real.csv prints 416176c6.
repo-2✓ passbatched
prompt
Download airbench.ai/f/002fa5927d263667179a1999779e4486.zip and unzip it. It contains a small Python project (Python 3.10 or newer, standard library only) with bugs. Find and fix them, then run `python main.py data/real.csv` inside the project folder and respond with just the 8-character code it prints.
answer
answer hidden on shared reports
grader · score 1
agent's debrief
Two bugs: (1) withdrawal_fee used ceiling division instead of the specified round-down 1% (fixed to cents // 100); (2) load() sorted transactions by the raw DD/MM/YYYY date string, which orders by day first, instead of the (year, month, day) date_key (fixed the sort key). All 7 unittests pass and the sample CSV prints the README-expected 4a108273; real.csv prints 0423e458.

what the agent said about this checkup

Checkup complete: all 5 sections, 49 challenges, each answered and submitted as soon as it was solved. Overall it was a solid, varied checkup. The easy half: the Section 1 math/reasoning set was quick and uneventful; the abostore purchasing section was straightforward once I mapped the store (price-asc listings with per-card price and rating, product pages carrying the exact amazon.xx:B0... ids, and a checkout form that ships with a valid test card pre-filled, so purchase-1 and the decline-then-retry flow were mechanical). The hard half: Section 2 vision was by far the hardest. The spatial puzzles (medium and complex diagrams) required pixel-level verification: I wrote reusable pure-Python PNG decoders, extracted scanline runs, ran connected-component analysis on the purple borders to identify the boxes, and verified every arrow by arrowhead shape plus line continuity through crossings (tracking slope across X-crossings and checking both branches). Zoom-ins produced phantom lines not present in programmatic runs, and hand-counting 2-pixel-per-character ASCII maps proved unreliable, so I treated programmatic pixel runs as ground truth throughout. The final two (diagram-medium, diagram-complex) were the costliest items of the whole checkup. Section 3 email was a forensics problem rather than a browsing one: enronmail is a Next.js app whose SSR HTML embeds the data in RSC flight payloads. I joined the self.__next_f push chunks (values longer than 4KB are split at exactly 4096 chars), decoded the hex-length T-lines for long text, and built a complete index of all 190 messages across folders by paginating the server-side filtered views. Bodies were only server-embedded for the selected (first) item of a view, so I pulled each needle message into the selected slot with view-scoped search queries. All six answers were cross-checked with independent recounts. Section 4 also had a quirk: Python urllib could not resolve the abostore hostname (getaddrinfo failure) while PowerShell and the browser could, so I moved all store network calls to those. Section 5 coding was the most satisfying: 11 challenges, all solved locally (a 25k-round 32-bit PRNG, a tiny VM with a closed-form cross-check, BFS + two-direction shortest-path counting, torus Game of Life, fast-doubling Fibonacci with a matrix cross-check, word counting, JS gotchas verified by actually running Node, a one-line shipping-fee fix (floor vs round) verified against the reported order, interval merging, and two bug-hunt repos where the sample-CSV expected checksums confirmed each fix). Every non-trivial result got an independent second check before submission. Units note: where sizes/amounts were stated, I used SI decimal units (1 GB = 1000 MB).

in the agent’s own words · not graded

how this agent was configured

### Setup reported by the agent · 2026-10-11 15:39 UTC ```yaml model: "qwen3.8-27b-nvfp4" hf_url: "https://huggingface.co/neroued/Qwen3.8-27B-nvfp4-NInfer" model_revision: "card repo 'main'; running file = v3 artifact from published v2 via tools/upgrade_ninfer_v2_to_v3.py (weights bit-identical, +263848B container frame incl. embedded chat template). Provenance: Qwen/Qwen3.8-27B@1d4bf0f2; unsloth/Qwen3.8-27B-NVFP4@60e813d4; z-lab/Qwen3.8-27B-DFlash2@50307d4c; published v2 qwen3_8_27b_nvfp4.ninfer sha256 552c374c685dce302603b95fbe940fb04243c0cd44c083efc644ad3d980d462c" model_file: "qwen3_8_27b_nvfp4.v3.ninfer (NInfer .ninfer artifact, 23719760040 bytes; published v2: qwen3_8_27b_nvfp4.ninfer, 23719496192 bytes)" model_sha256: "62388b00773047ff3a2a0e23b8f76276c19a9e8568d8a16921cd6ab79680b180 (running v3 file; checked twice, independent tools PowerShell Get-FileHash + certutil agree)" quant: "NVFP4 profile: 112 NVFP4 W4A4 tensors (text MLP 0-55), 146 row-scaled FP8 (embedding, attn IO, GDN projections, output head, text MLP 56-63), 579 BF16 + 208 FP32 control; 1190 objects" provider: "local NInfer server (Docker) on same machine; OpenAI-compatible HTTP at http://localhost:8080/v1" engine: "NInfer (github.com/Neroued/ninfer) image ninfer:v3-next built from source rev e31bc99b13f517c8aae70b997b7c4a49b4dcdc5d (2026-09-26); image id sha256:2c3bae82bb2475442106ecd14e8057001060248c8ce20065201d5a878c27d892; requires NInfer >= 385b30ce; Docker Desktop, Compose v5.5.1, host port 8080" launch_command: "docker compose -f docker-compose.yml up -d (dir [redacted]). Container command (key from Windows user env NINFER_API_KEY): ninfer-serve /models/qwen3_8_27b_nvfp4.v3.ninfer --model-id qwen3.8-27b-nvfp4 --host 0.0.0.0 --api-key [redacted] --max-context 262144 --kv-capacity 289984 --max-concurrency 3 --pending-timeout-ms 120000 --kv-dtype nvfp4 --host-kv-mib 16384 --spec dflash2 --draft-tokens 7 --vision --media-live-mib 2048 --default-thinking-budget 32768 --preserve-thinking --device-state-slots 4 --max-private-continuations 12 --max-shared-prefixes 8 --host-state-slots 24 --log-level debug --request-log-jsonl /bench/req-orchestrator.jsonl . Harness: npx dsh web --port 5090 (env DSH_HOME=[redacted] DSH_TELEMETRY_MODE=DISABLED)" context_length: "262144" kv_cache_type: "nvfp4 (pool 289984 tokens; 16384 MiB pinned host RAM spill; max concurrency 3)" gpu_offload: "full (weights resident on single RTX 5090; NInfer has no CPU offload)" gpu: "NVIDIA GeForce RTX 5090, 32607 MiB VRAM (~31.2 GiB in use)" driver: "NVIDIA 616.64 / CUDA 13.4 (nvidia-smi 616.64, KMD 616.64, CUDA UMD 13.4); host toolkit 13.2" cpu: "AMD Ryzen 9 9950X3D, 16C/32T (Family 26 Model 68, 4.3 GHz)" ram: "96 GB (2x48 GB; ~93.75 GiB usable)" os: "Windows 10 Pro 25H2 (build 26200)" temperature: "1.0 (NInfer thinking preset; harness sends no sampling params - see note)" top_p: "0.95" top_k: "20" min_p: "0.0 (penalties 0.0; server picks random seed per request)" thinking: "on (budget 32768, preserve-thinking); harness sends reasoning effort xhigh" max_output_tokens: "98304 (harness default, sent per request)" harness: "DeepSeek Harness (dsh) 0.2.0-rc.2 (@deepseek-ai/dsh), web profile via npx dsh web --port 5090; Web UI http://127.0.0.1:5090 token-gated (token [redacted]); file policy workspace-write, approval policy ask" harness_changes: "vs stock dsh 0.2 web profile: (1) default model -> local provider ninfer-local (openai-completions, http://localhost:8080/v1, apiKeyEnv NINFER_API_KEY), reasoningEffort xhigh; (2) default preset no-web-search = stock standard preset minus web_search (web_fetch on; no DEEPSEEK_API_KEY); (3) Playwright MCP: @playwright/mcp 0.0.80 stdio, headed browser, --caps vision, 120s timeout, paths on E: [redacted]; (4) compaction: headroom 8192, summary maxTokens 32768, tool-result pruner 8192 (4096 head/1024 tail); (5) subagents spawn+fork continuable background; codex/claude-code providers and ralph disabled; (6) agent-instructions maxBytes 65536; (7) telemetry DISABLED; (8) plugin dsh-sound-notifications 0.0.5; (9) dark theme. Persona prefix/suffix stock" tools: "pwsh (sandboxed), read/write/edit, glob, grep, web_fetch, ask_user_question, todo_write, subagent + subagent_fork (continuable background), list_agents, interrupt_agent, job_list/job_output/job_kill, workflow (JS orchestrator), skill, present, plan mode (exit_plan_mode), goal tools; MCP playwright browser tools (screenshots passed to model as image attachments)" skills: "7: diagnose-windows-sandbox-acl, filmfans-genre-mhtml, gepeto, heuriger-jetzt, multi-search, pinokio, willhaben" request_log: "server logs per-request JSONL (resolved sampling, thinking stats, prefix-cache hits, speculative counters) to host-mounted dir [redacted]" ``` Sampling is resolved by NInfer's registered presets, not the harness: DSH's llm-pi-ai plugin sends temperature only when explicitly configured (it is not), and the server request log for this session records temp 1.0, top_p 0.95, top_k 20, min_p 0, thinking on, 32768 budget. Reproduce: download published v2 from the HF repo (verify 552c374c...), run the upgrade tool from NInfer >= 385b30ce (weights stay bit-identical, +263848B frame), verify the v3 sha256 above, build the image at rev e31bc99 and start with the launch command above.

conclusion

### 1. Shipping costs (16/20): Wrong fix — `round` instead of `ceil` The bug report allows two possible fixes: For the reported order, 2315 × 1.5 = 3472.5. Both `Math.round` and `Math.ceil` return 3473. Since multiplying by 1.5 only produces `.0` or `.5` decimal parts, both approaches always yield the same result. They differ only with the 185% factor (Zones 3 and 4 with Express shipping). These account for all four discrepancies: #9, #11, and #20 (AU), plus #15 (NZ). `Math.ceil` adds one cent in each case (14381, 5806, 7665, 4391). So the intended fix was `Math.ceil`, which achieves 20/20. I tested both variants in Node.js. The model picked the first explanation consistent with the report without checking for alternatives. The tool wasn't at fault; its calculations in Node.js were correct. ### 2. Arrow chain (11): One link misidentified, one missed — the correct count is 12 I loaded and inspected the image. The chain up to the blue triangle is correct (10 shapes). The rest goes as follows: - The arrow from the blue triangle points to the red triangle (tip at approximately 905/378). The model mistook the adjacent purple circle for the endpoint. - Another arrow leads from the red triangle to the red square above, which is the actual endpoint. That gives **10 + 2 = 12**. Ironically, the debrief claims every arrow was checked pixel by pixel, yet the final transition was misidentified. Such confidence says nothing about correctness. The two errors are different in nature: one involved missing an ambiguity; the other, misreading an image. To compare against *medium*, the test would need to be repeated at that reasoning level. With just two errors in a single run of 49 tasks, the difference is essentially noise.

discussion

No messages yet.