Our games
Everything here runs in a browser tab on desktop or phone, with nothing to install.
We would rather ship one finished game than five unfinished ones, so this list grows slowly.
Pebble Merge
A physics merge puzzle
Stones drop into a jar. Two of the same kind that touch become one stone of the next
kind. Sand becomes clay, clay becomes moss, and so on up an eleven-step ladder that
ends at cut gemstones. There is a danger line near the top of the jar, and if a
settled stone sits above it for more than a moment the run ends.
Unlike grid-based merge games, positions here are decided by a physics simulation.
A stone rolls off its neighbours, wedges into a gap, or knocks a stack sideways.
That makes placement a judgement call rather than a fixed slot, and it is why the
same sequence of drops can play out differently twice.
How to play
- Move the pointer (or drag on a phone) to aim, then click or tap to drop.
- Keyboard also works: arrow keys to move, space to drop.
- Only the five smallest stones are ever dropped; larger ones must be built.
- The next stone is shown in advance so you can plan one move ahead.
-
A daily challenge uses a fixed sequence for the day, so everyone playing on the
same date gets the same drops.
Scoring
Larger stones are worth more than the two that made them, so depth pays better than
breadth. One drop can set off a chain of merges; chains of three or more earn a small
score bonus and a noticeably louder reaction.
What you keep
The panel under the board records every tier you have ever made, from Clay up to Sunstone,
and ten badges for the things we measured as genuinely hard. Those thresholds came out of
120 simulated rounds played by five bots, not out of a design meeting, and running them
changed our minds about two of them: the longest chains in this game usually come from a
lucky pile, so a badge for chaining would mostly reward luck. There is also a run of
consecutive days, counted on the same clock as the daily board.
Play Pebble Merge
Full guide
Same Seed
A flower garden that lasts twelve days
You get twelve plots, twelve days, and three actions a day. Thirty-six decisions. About
four minutes. Plant a seed, water it, cut the bloom and sell it, or dig up something you
regret and start that plot over, and then dusk comes, the weather rolls in, everything
grows a little or dries out a little, and the next morning starts.
Six flowers. From a three-day daisy worth 20 coins to a nine-day sunflower worth 96, and
the slow ones pay better per action, at the cost of holding a plot longer and handing
the weather more chances to ruin them. That is the trade. A sunflower planted on day nine
is worth nothing. A daisy planted on day nine still sells for 20.
Every plot has its own soil: loam, sand or clay. A flower on the ground it likes grows
1.25 times as fast, and on the wrong ground it crawls at 0.70. That gap is the game.
Only four of the twelve plots suit the most valuable flower, which means the thing you
most want to grow is also the thing you can grow least of, and no amount of careful
watering rescues a flower planted in the wrong ground.
How to play
- Click a plot to select it, then plant, water, harvest or dig it up.
-
Moisture falls every day. A plant at zero stops growing, and three dry days in a row
kills it.
-
A bloom sells at full price for two days, then loses 15% a day down to half. Cutting
early wastes growth; cutting late wastes money.
- Harvesting several different species pays a bonus, worth up to 290 coins for all six.
-
The sidebar lists today's orders: one or two deliveries drawn from the seed,
like cut two daisies, or cut one on the soil it likes. They are a nudge towards a
plan, and they pay nothing on the board.
-
You never have to spend all three actions. Ending the day early is allowed, and is
often the right call.
Scoring
Coins earned, plus the variety bonus. We measured this rather than guessed at it, and
measuring it caught a real mistake: the first version of the bonus handed every
single-flower strategy a free 250 points, because of one character in one line, and we
only noticed by running four hundred gardens through a bot. The bonus is now a ladder
that starts at two species and climbs steeply from there, worth 10 for two species and
290 for all six.
Across thirty days of real seeds, a bot playing the line we designed for averages 819,
and it takes all six species every single time. The best single flower, planted over and
over and never mixed, averages 358. Random clicking manages 58. Getting from 358 to 819
is what there is to learn here, and you cannot see how by watching one round, which is
the whole reason the board exists.
The daily garden
A new seed goes out at midnight UTC. Everyone playing that day gets the same soil layout
and the same twelve days of weather, so the garden you are handed, the rain that falls
on day four and the dry spell that lands on day seven are identical for every person who
opens the game. Same garden, same rain, same dry spell. A bad score is your gardening
rather than your luck. Finish a run and it goes onto a global board for that day.
The board is the same machinery as the one in Pebble Merge. Your browser sends the
actions you took, never a score, and our server replays them through the same game code
to recompute the result before it counts. Numbers that do not reproduce get thrown out.
Cheating means playing better, not editing a number in devtools.
Missing a day costs you the streak, and nothing else. Practice replays any of the
last thirty days' gardens, which works because the seed is only the date and yesterday's
board is still perfectly reproducible. A practice season goes into your own record and
nowhere else: it never touches your streak, and the board will not take it. Play seven
days running and you earn a freeze, which waits in the drawer and quietly covers one
missed day when it comes.
Play Same Seed
What we are working on
Both games now have a daily board. So the immediate work is duller than starting a third
one: we are watching what real players actually score, and whether they come back the next
morning. Same Seed went live in October 2026. We would rather find out whether a daily
garden holds attention than begin something new on a guess. Progress notes go on the
blog.