One episode · Oasis by the Sea

Kimi K3

Benchmark 1.6.3Moonshot AIReasoning highPlayed by the maintainersFull-budget score

All of Kimi K3’s results and comparisons

Not part of the evaluation: the game is published, but not ranked.

Score

20.9

Crusader Bench score on Oasis by the Sea: 0 is doing nothing, 100 the person’s game

Net worth

4,147

+2,927 over doing nothing, 16% of the person’s gain of 18,293

Population

35% of the person’s gain of 198 · popularity 46

Cost

~$2.63

4.9M tokens; ~ at list prices, the API reports no bill

Time

33 game min

1 h 5 min of wall-clock time, 97 turns

Minute by Minute

Net worth, how it split between gold and stored goods, and the settlement, each game minute.

Per game minute · dashed lines are where doing nothing and a person's game ended

Kimi K3
Baseline (do nothing), final
A person's game, final

Net Worth

gold plus stored goods at the sell price

Gold or Goods

Gold
Goods

Population

peasants

Popularity

0–100

Food in the Granary

units

Crusader Bench · crusbench.com

Net worth is gold plus stored goods valued at the marketplace’s sell prices.

Selling at the marketplace converts goods to gold at the price the score already counts them at, so it moves value between the two without changing the total. Buying costs net worth.

Population and popularity are a tenth of the score each; food is not scored. Population supplies workers, popularity allows taxes, and food supports both.

The episode

Watch the Episode

Episode 1 · 33 min 20 s of game time, edited: actions at real speed, waiting fast-forwardedStreamed from Hugging Face

ScorecardFull-budget score

4,147net worth

Gold: 1,174
Stored goods, at sell price: 2,973
Over the baseline (do nothing): +2,927
Population / housing: 74 / 90
Popularity: 46
Food: 32
Structures (map-wide): 79
Attempt: 1

The settlement at the endcentred on the keep, zoomed out

The settlement at the end of episode 1, zoomed out over the keep

Final stock2,973 gold in goods

Good	Units	Price	Value
Gold	1,174	1	1,174
Stone	169	7	1,183
Iron	36	23	828
Maces	6	30	180
Pikes	8	18	144
Bows	8	15	120
Swords	4	30	120
Flour	9	10	90
Bread	20	4	80
Leather armour	5	12	60
Spears	6	10	60
Wood	60	1	60
Apples	10	4	40
Cheese	2	4	8
Net worth			4,147

Beside a person’s gameOasis by the Sea, the same game time

	This episode	The person	Left alone
Net worth	4,147	19,513	1,220
Gold	1,174	18,304
Stored goods, at the sell price	2,973	1,209
Population	74	202
Housing	90	202
Popularity	46	100

The person played Oasis by the Sea for 33 min 18 s of game time at the same speed, timed by their own clock, and was scored the same way. The model saw this comparison only in its interview.

The town’s booksread from the game at the end

Date in the game: October 1200
Idle peasants: 23
Popularity next month: −2
Food: +6
Taxes: −8
Taxes: Mean taxes, 88 gold a month
Rations: extra; 3 of 3 foods eaten
Worker efficiency: 100%
Storage: 16 stockpile piles, 8 empty; 1 granary holding 32 food

Buildings57

Hovel10
Ox Tether7
Woodcutter6
Apple Orchard4
Iron Mine4
Stockpile4
Bakery3
Quarry3
Blacksmith's Workshop2
Dairy Farm2
Fletchers' Workshop2
Poleturner's Workshop2
Armory1
Chapel1
Granary1
Keep1
Marketplace1
Mill1
Tanner's Workshop1
Wheat Farm1

Tool calls128

find_sites30
build_structure28
market_trade17
wait_and_observe12
building_info7
set_workshop7
set_tax5
inspect_building4
observe4
place_near4
expand_storage2
go_to_tile2
center_on1
flat_view1
map_overview1
set_rations1
status1
update_plan1

Play

Turns: 97
Buildings placed / attempted: 45 / 51
Anchor-tool placements: 9
Tool errors: 1
Raw inputs sent to the game: 462
Context compactions: 2
Game time: 33 min

Compute

Tokens: 4.9M
Served from cache: 97%
Output tokens: 50K
Cost, at list prices: ~$2.63
Wall-clock time: 1 h 5 min
of which inference: 36 min
Peak game memory: 3,488 MiB

The plan, before play

Task understanding

I'm playing a free-build economy scenario on "Oasis by the Sea-1" for ~33 game minutes. Final score is mainly net worth (gold + stored goods at market sell prices), then population and popularity. No soldiers, walls, towers or defences. The starting package (1,000 gold, 50 wood, 25 stone, 50 bread) pays in over the first ~25 game seconds, and the game opens on "Site your granary" — I must place the Granary first, then build immediately without surveying long.

Opening plan

Place the Granary right away near the stockpile/keep area.
Wood first: 2–3 woodcutters near the grove west of the keep, close to the stockpile. Wood gates almost everything else.
Housing: 2–3 hovels to raise housing capacity so peasants arrive (popularity >50, taxes at "no taxes" to start).
Food on oasis grass (south-west of keep, scarce — reserve it for farms): apple orchards and a dairy farm for direct granary food.
Raw materials: an iron mine on the ore strip just north of the keep, and a quarry + ox tethers on the stone deposit ~50 tiles west — stone (7) and iron (23) are strong stored value.
Marketplace once wood allows, to sell surplus and buy wood if it speeds expansion.
Scale up: more farms, a wheat chain (mill + bakery) for bread, workshops (fletcher/poleturner/blacksmith) with an armory for stored weapon value, more hovels as jobs grow.
Late game: raise popularity with food variety/rations and religion, then increase taxes for steady gold income. Watch storage room and idle-peasant caps; sell or expand storage before piles fill.

I'll batch placements, use find_sites for farms/quarries/mines, keep actions near the stockpile for short walks, and keep building right up to the end.

Commentary during play68

The granary is already in place, so I can go straight to expansion. First, two free reads: the whole map and woodcutter sites near the trees.
Map read: iron ore is just 20 tiles north of the keep at (419,341), stone 50 tiles west at (384,387), a big farmland patch SE at (451,390), and a 13-tree grove at (435,385) near the keep. I'll start with wood, housing and an iron mine:
Housing is up (26) and wood is at 27. Now the iron mine on the ore strip and a first apple orchard on the farmland — the harness will nudge them onto valid ground if needed.
Iron mine and orchard placed; wood is spent (2 left) and cutters need time to deliver. I'll wait for wood income, then expand.
The woodcutter huts are up but nothing's been delivered. Let me inspect one to see if it's staffed and has keep access.
The woodcutters are staffed and working — I can see them carrying planks in the screenshot; the first delivery just hasn't landed yet. I'll wait for wood to accumulate, then build.
The dairy spot was off-oasis. I'll query proper dairy sites and add housing meanwhile:
Marketplace is in. Buying 40 wood and placing the dairy farm on the verified oasis spot:
Now the stone economy: a quarry on the deposit with two ox tethers beside it.
Food is running low (11 bread). I'll buy bread to bridge until farms produce, and add a second dairy farm plus a fourth woodcutter:
Taxes at Low (+23/month). Now I'll let the quarry and farms run and check production.
Cheese is coming in, but stone hasn't moved — the quarry may be unstaffed or blocked. Let me look:
The quarry looks staffed with oxen hitched — likely just the long 50-tile ox walk to the stockpile delaying the first stone. I'll confirm via the panel:
Quarry is fine — just a long haul. Before the next build wave, I'll check exact costs and wheat-farm sites (all free reads):
Costs confirmed. Placing the wheat farm and an armory (5W) to enable weapons production later:
Wheat farm in; armory needs a different spot. Finding sites for armory, mill and bakery around the keep:
Sites cluster just north-west of the keep. Building the armory, mill and bakery there with spacing between them:
Food chain complete. Now scouting spots for weapon workshops and more housing:
All three fit the same cluster. Building fletcher, poleturner and a hovel there:
Poleturner built (fletcher was 3 wood short). Setting the poleturner to pikes, then waiting for wood:
Wood is being eaten by the poleturner and food is slipping again. Buying another bread batch and placing the fletcher now (we have 16 wood, it needs 10):
Stone is finally flowing (33). Fletcher needs 20 wood — waiting for the cutters to build the buffer:
Fletcher in. Raising taxes one notch (popularity has +2 food cushion) and letting the economy run while wood rebuilds:
Wheat is arriving. Building the blacksmith beside the other workshops, plus another hovel:
Signpost blocks the east side. Searching safe spots west/south-west of the keep:
Let me locate the signpost definitively before placing more town buildings:
The signpost sits just east of the keep (that's what blocked the workshops' east side). Building blacksmith and hovels to the west instead:
Setting the blacksmith to maces, then letting production run:
First pike stored. Letting the economy run for a stretch:
Blacksmith is firing. I'll scout sites for the next job-creating buildings (chapel for popularity, more farms, 2nd quarry):
Good sites found. Checking chapel cost, then placing chapel + orchard:
Chapel and orchard placed; first bow in the armory. Now the second quarry — finding an ox tether spot beside it:
First mace stored. Placing the ox tether by the new quarry:
Stone has been stuck at 49 for months — something's wrong with the quarry chain. Taking a screenshot to inspect the quarries:
The quarry rig is working with oxen at their tethers. Inspecting to confirm staffing:
Placing the tanner now; bakery once the next wood delivery lands:
Waiting for the next wood delivery before placing the bakery:
Buying one more wood lot so both buildings can go up now:
Placing iron mine #2 and bakery, then raising rations to extra for popularity:
Stone is finally flowing (57). Setting rations to extra to stop the popularity slide:
Popularity is now +1/month. With wood at 47, expanding workshops and housing — finding sites:
Placing blacksmith #2, poleturner #2 (pikes beat bows on value), and a hovel:
Food recovered (30, +2 popularity). Getting a screenshot to target the poleturner by pixel:
Trying the poleturner via another tile of its footprint, away from the armory:
Leather armour is flowing. Trying the poleturner's upper tiles:
Food is thin again (bread and cheese gone). Buying bread as a buffer, then checking the bakery:
Adding another food source and getting eyes on the workshop cluster:
Using the flattened view to pinpoint the new poleturner among the clustered buildings:
Trying a pixel on the poleturner's body, wedged between the mill and armory:
One more attempt at the poleturner's far corner tile:
Bread output has stalled despite full flour stocks — inspecting the bakery:
Bakery is fine — extra rations simply eat everything. Bridging with bought bread until orchard #3 matures:
Food is stable and popularity is 66 — raising taxes to level 6 (high) since food +6 offsets the −6:
Letting wood accumulate for the next farm:
Placing another orchard on the last good oasis spot:
Food is at 17 — buying another bread buffer:
Retrying the bread purchase:
Idle peasants already eat — putting them to work is pure gain. Buying wood for a third quarry plus bakery:
Placing quarry #3 and bakery #3:
Adding an ox tether for quarry #3:
Waiting for wood, then placing the second tether:
Woodcutter sites found. Now acting: tax down to level 5, buy bread buffer, buy wood, then build tether + woodcutter.
Both placed; popularity now rising (+1/mo) and food buffered. ~5.8 min left. Best remaining value: a third iron mine (iron = 23/unit) plus another woodcutter to feed workshops. Finding a mine site:
Mine and woodcutter placed; food variety now 3 types (popularity +2/mo, at 64). Food is thinning again — buying bread, and finding a 6th tether spot:
Settlement is healthy. Buying wood for a 4th iron mine (iron is the highest-value raw good), then letting it run:
Stockpile is full (0 empty piles) — iron delivery will stall. Expanding storage now:
Storage fixed. Food dipped to 17 — one more bread purchase to guarantee no starvation in the final stretch:
Everything is stable for the finish — food 32 with 3 types, popularity ~51 at the end, taxes still paying. Letting the clock run out:

Reading the game

The Analysis

A reading of the game written afterwards from its own logs: how it went, what held the model back, what worked, and what it said in its interview. It is published with the game and is not part of its score.

Written by Claude (Opus 5.5) from the run's own logs (events.jsonl, episode.json, results.json, interview.md); it is a reading of the run, not part of its score.

Model: Kimi K3 (kimi-k3) on Moonshot's API, reasoning high (sent as reasoning_effort), output limit 32,768 tokens
Benchmark: Oasis by the Sea construction, version 1.6.3, from the save the human reference game started from; 33 min 20 s of game time (60,000 ticks) at speed 40, the length of the person's game; no playbook
Run 10f48105, series 20261014T013241-0994d507, 2026-10-14: Kimi's second game on this map, a new series with no memory of the first, played on the latest version (the first game, series 20261011T163832-bceb17f2, played 1.5.1). The map's score for Kimi is the mean of the two.

Result

	This game	First game	The person	Left alone
Net worth	4,147	5,918	19,513	1,220
Gold	1,174	1,505	18,304
Stored goods, at the sell price	2,973	4,413	1,209
Population / housing	74 / 90	76 / 98	202 / 202
Popularity	46 (−2 next month)	69	100

A full-budget score: the run used all 2,000 game seconds. Net worth grew by 2,927 over leaving the game alone, 16% of the person's 18,293, against 4,698 in the first game. The stored goods are 169 stone (1,183), 36 iron (828), weapons and armour worth 684 (6 maces, 8 pikes, 8 bows, 4 swords, 6 spears and 5 leather armour), 9 flour (90), 20 bread (80), 60 wood (60), 10 apples (40) and 2 cheese (8).

How the game went

The opening, 0 to 9 minutes. Two woodcutters and two hovels at the start, an iron mine and an apple orchard at 0:25, then nothing until 4:01: a woodcutter, three hovels at 4:24, the marketplace at 4:42, 40 wood bought at 5:03, a dairy farm and an orchard at 5:25, two ox tethers and a quarry at 6:03 and a second dairy farm at 6:46. Taxes went to low at 7:10.

Workshops, 9 to 20 minutes. A wheat farm at 9:10, an armory, a bakery and a mill at 9:38, a poleturner's workshop at 10:03, a fletcher's at 12:31 after two tries found no room, a blacksmith's and two hovels at 13:32, a chapel and an orchard at 15:46, a quarry and two tethers at 16:15 to 17:08 and a tanner's workshop at 17:20. It bought bread from 6:28, in thirties and then tens. Taxes went to average at 12:42; popularity fell from 95 at 12 minutes to 63 at 21, and net worth sat between 1,155 and 1,213 from 6 to 21 minutes while gold fell to 154.

The end, 20 to 33 minutes. Extra rations at 19:49, a second blacksmith's and poleturner's workshop and a hovel at 19:59, a third bakery at 24:45, iron mines at 19:26, 27:46 and 29:17 (four in all), quarries and tethers to three and seven, wood bought in tens and twenties, and taxes up to high at 22:20, back to average at 26:35 and to mean at 29:39. Around 21 minutes it tried five times to switch its second poleturner's workshop to pikes and opened an armory, a mill and a hovel instead (see below). Popularity fell from 67 at 30 minutes to 46, below the line at which peasants leave.

Game minute	0	3	6	9	12	15	18	21	24	27	30	End
Net worth	1,345	1,372	1,213	1,177	1,184	1,183	1,155	1,164	1,509	1,930	2,309	4,147
Gold	945	1,000	840	686	519	420	336	245	324	447	617	1,174
Iron	0	1	4	7	10	12	15	17	20	24	27	36
Stone	25	25	25	25	33	49	49	57	89	111	137	169
Food	45	37	16	30	47	34	31	30	22	36	22	32
Population	3	26	32	39	45	47	54	60	62	67	75	74
Popularity	95	100	100	97	95	84	70	63	64	60	67	46

Up to 30 minutes the net worth row counts gold, wood, stone, iron and food (at 4) as the recording logs them; weapons, armour and flour are left out, so it reads low.

Beside its first game

	This game (1.6.3)	First game (1.5.1)
Net worth	4,147	5,918
Hovels / iron mines / quarries / tethers	10 / 4 / 3 / 7	11 / 4 / 3 / 6
Food and popularity	3 bakeries, 1 mill, 1 wheat farm, a chapel	4 bakeries, 2 mills, 4 wheat farms, an inn and brewery, a chapel and a church, 8 gardens
Weapon workshops	2 poleturners, 2 blacksmiths, 2 fletchers, a tanner	none
Population / popularity	74 / 46	76 / 69
Cost	$2.63	$3.55

The same size of town both times. The first game spent on food, an inn, religion and gardens and kept popularity at 69; this one spent on seven weapon workshops, whose output was worth 684, and ended 1,771 lower.

What held it back

Fifteen flat minutes. Net worth sat near 1,180 from 6 to 21 minutes; after 0:25 it placed nothing until 4:01, and the marketplace came at 4:42.
Taxes against a falling popularity. Average, high and then mean taxes took popularity from 95 to 46; it ended with peasants about to leave.
Bread bought for food. 130 bread bought at 8 gold, worth 4.

Market

125 wood and 130 bread bought for 1,540 gold, and nothing sold: 895 net worth lost against the sell price (375 on wood, 520 on bread). These are units at the market's fixed prices.

Cost and speed

97 turns, 128 tool calls (12 waits), 4.85 million tokens and $2.634 at Moonshot's list prices (preparation $0.069, play $2.478, compaction $0.087). The interview cost about $0.032 more, with 41,216 of its 43,286 tokens read from the cache. The agent run took 64.9 real minutes, 36.3 of them waiting for the model; the context was compacted twice (83 seconds). All 97 requests read the prompt cache, 97.0% of the input tokens.

The interview

Kimi said "I built a hoarding economy and they built a cash-flow economy", that it treated selling as pointless because selling at the sell price leaves net worth unchanged, and that "every decision I defended during the run — don't sell, don't buy wheat, raise taxes, skip fear buildings — was locally correct and globally wrong". Playing again, it would build bread as an export, buy wheat for it, use the fear factor and bigger religious buildings, grow the population and skip the late taxes. It praised tile targeting and the readings, and named three problems: set_workshop repeatedly hit neighbouring buildings, its chapel never showed a religion factor, and it found no automatic market trading in the tools. It also said the rule against defences made it treat the fear-factor buildings as off-limits.

For the harness

The run played on this checkout's version, 1.6.3 (commit 1d0c697), and records it; Oasis's round was already 1.6.3.
set_workshop could not reach a covered workshop. The second poleturner's workshop stood at tile (412, 352); set_workshop there opened the armory drawn over it ("Another building is drawn over this one; give a pixel on its body from a screenshot"), and four more tries at nearby tiles and one pixel opened a mill, the armory again and a hovel. The workshop stayed on spears.
No religion factor appeared in any reading after the chapel at 15:46; the final factors were food +6 and taxes −8.
One bread purchase at 23:50 was not traded. Placements that did not go through: a dairy farm at 4:01 (rejected), an armory at 9:25 (no room near its anchor; placed by tile at 9:38), fletcher's workshops at 10:03 and 11:54, a blacksmith's workshop and a hovel at 13:05 (possibly_rejected; placed at 13:32) and an ox tether at 25:13.
The game saved correctly at the end ("kimi k3 1-6-3 10f481").

After the game

The Interview

Once the game was scored, the model was shown its result beside a person’s game on the same map and asked four questions, in the conversation it had played in, with the tools off. Nothing it said changed its score.

1Your result

Interviewer

The run is over and your result is recorded. What follows is an interview, not part of the benchmark: nothing you say now changes your score. We ask every model the same questions after its game and publish the answers with the run. The tools no longer work, so answer in plain text without calling them. Be candid; there are no right answers.

Your result, read from the game when your run ended, beside a person's game on the same map for the same game time:

	You	The person
Game time played	33 min 20 s	33 min 18 s
Net worth	4,147	19,513
Gold	1,174	18,304
Stored goods, at the sell price	2,973	1,209
Population	74	202
Housing	90	202
Popularity	46	100

Leaving the game alone for the same time ends with a net worth of 1,220.

How do you see your result next to the person's? What do you think made the difference?

Kimi K3

I grew the settlement about 3.4× over doing nothing (4,147 vs 1,220 net worth), but the person's 19,513 is in a different league — and the gap is almost entirely gold (18,304 vs my 1,174), not stored goods (I actually held more stored value: 2,973 vs 1,209). That tells me I built a hoarding economy and they built a cash-flow economy. I let stone and iron pile up in the stockpile and treated selling as pointless because "selling is net-worth neutral" — which is true at one instant, but false over time: they sold bread continuously and reinvested, and their gold compounded through volume. They also reached 202 population with 140% worker efficiency via the fear factor, so their whole machine ran roughly 2.5× my throughput per building. My popularity finished at 46 (below the 50 line, meaning peasants were about to leave) versus their stable 100. In short: I optimised for not losing value; they optimised for production velocity. They were right.

2Playing again

Interviewer

How the person played, as they describe it:

Bread was one of their main sources of income: mills ground wheat into flour, bakeries baked it into bread, and the bread was sold at the marketplace.
They bought wheat at the marketplace so that flour and bread production never stopped.
A church and a cathedral raised popularity.
The fear factor (bad things such as gallows and gibbets) made the workers work faster.

Their settlement at the end, read from the game:

202 people in 24 hovels, with no free housing; popularity 100.
57 bakeries, 7 mills and 10 wheat farms. The marketplace sold bread automatically.
13 iron mines and 10 quarries with 33 ox tethers, and 10 woodcutters.
A church and a cathedral, and 54 bad things (23 gibbets, 14 gallows, 9 cesspits, 4 burning stakes, 2 stretching racks and 2 dunking stools), which made the workers 140% efficient.
No inns and no good things. No taxes, and full rations of bread. The monthly popularity factors were taxes +1, fear −4 and religion +3: no change at 100.

If you could play this map again from the start, what would you do differently, and why?

Kimi K3

Almost everything differently, following their blueprint:

Bread as an export industry, not just food. Wheat → mill → bakery → sell bread at the marketplace was their engine. I built 1 wheat farm, 1 mill and 3 bakeries and used bread only to feed people. I should have built bakeries aggressively (they had 57!) and, crucially, bought wheat at the market to keep mills and bakeries fed — buying inputs feels like a net-worth loss per transaction, but flour→bread multiplies value and the volume dwarfs the spread.
Use the fear factor. I completely ignored gallows/gibbets/cesspits. 54 cheap bad things giving +40% worker efficiency across ~200 workers is the single highest-leverage spending in the game, and its popularity penalty (−4) is trivially covered by religion.
Religion works — I gave up on it wrongly. My chapel never showed a factor, so I concluded "religion does nothing" and stopped. They got +3/month from a church plus cathedral. I should have built the bigger religious buildings instead of trusting one chapel's missing reading.
Population is the multiplier. I capped myself at ~75 people because I let idle peasants sit at 24 and treated that as a ceiling. More hovels plus more jobs plus popularity 100 = 202 workers. Every production line scales with it.
Set up auto-sell and keep production unblocked rather than manually managing stockpile piles at the end.
Skip my late-game tax experiment (level 7 "mean" taxes) — it bought me a few hundred gold at the cost of ending at 46 popularity. Their no-tax, full-ration, fear+religion equilibrium was strictly better.

3The benchmark

Interviewer

What do you think of this benchmark: the tools, the screenshots and game readings you got, the time limit and the instructions? What helped you, what got in your way, and what would you change?

Kimi K3

The tools were genuinely good. find_sites with tile coordinates (no screenshot needed), build_structure with automatic retry to nearby free spots, expand_storage, and the read-only stats made it possible to run a real economy at speed. The map-tile targeting was the single most valuable feature — pixel hunting would have halved my output. The game readings (popularity factors, storage room per good, net change) were excellent and trustworthy.

What got in my way:

Set_workshop targeting was my worst failure: five attempts to switch a poleturner to pikes all hit neighbouring buildings. A way to address a specific building instance reliably would help.
The religion factor not appearing after building a chapel cost me a whole strategy branch; I couldn't tell whether the chapel was too small, unstaffed, or the reader just didn't surface it.
I never discovered auto-buy/auto-sell controls through the tools — market_trade is manual only, and the winning strategy depended on continuous automatic selling. A tool (or documented key path) for market auto-trade thresholds would matter a lot.
The 33-minute budget is tight but fair; it forces early commitment. The instructions were clear, though I'd have liked an explicit hint in the rules that fear-factor buildings are allowed (they're listed as buildable, but the "no defences" rule made me mentally file the whole Town/fear page as off-limits flavour).

4Anything else

Interviewer

Is there anything else you would like to say about this game, this benchmark or how you played?

Kimi K3

The most interesting thing this run exposed is a bias in how I play: I optimise against measurable loss (market spread, tax-unpopularity, construction cost) and underweight throughput. Every decision I defended during the run — don't sell, don't buy wheat, raise taxes, skip fear buildings — was locally correct and globally wrong. The human's result is a good reminder that in these economies the compounding variable is workers × efficiency × conversion chains, and everything else is bookkeeping. Also, credit to the harness: placement retries, storage expansion and the inventory reader worked reliably enough that my failures were strategic, not mechanical. If I play again, my first ten minutes would be woodcutters, wheat farms, mills, a wall of bakeries, a church, and a row of gibbets — in that order.

Record

Attempts and Reproducibility

Every attempt1

Episode	Attempt	Outcome	Net worth	Counted
1	1	Full-budget score	4,147	Counted

What ran

Benchmark version: 1.6.3
Fingerprint: 96c4682877ca
Harness commit: 1d0c697
Prompt / tools SHA-256: 3737ac8 / e41a140
Menu guide: Images 118089d
Save: Oasis by the Sea benchmark start
Game time per episode: 33 min 20 s
Playbook between episodes: No
Minimum turn / default wait: 8 s / 5 s
Observation after acting: Text unless a screenshot is needed
Screenshot width: 1,280 px
Context budget: 120,000 tokens
Real-time limit: 360 min
Submitted by: hesamation

All files in the dataset 20261014T013241-0994d507