Page 2 of 2
Re: CHALLENGE: Find the longest-lived GOL pattern w/ the fewest / most alive cells as its initial state
Posted: August 17th, 2026, 7:38 am
by hash0
Gentlepeople, adjust your miners and clear your kernels! I am opening up this challenge for toroidal grids of sizes above 16x16: Let's explore the sparse and dense patterns on 20x20, 24x24, 28x28, and 32x32!
- sparse-dense___nxn.png (99.56 KiB) Viewed 400 times
I am convinced that we are going to see more interesting patterns emerge, maybe even new Agars like the p314 that was discovered on the 16x16.
Happy hashing!
Re: CHALLENGE: Find the longest-lived GOL pattern w/ the fewest / most alive cells as its initial state
Posted: August 17th, 2026, 7:50 am
by Vort
Great, thank you. I think 32x32 has massive potential (maybe unreachable for now, but anyway).
Can you please fix changing of rule during manual submission in [ GLYPH FOUNDRY ] ?
I'm trying to submit
Code:
Select all
x = 10, y = 9, rule = B3/S23:T16,16
7bo$8b2o$7bo3$7bo$2bo4b3o$2ob2o2bo$2bo!
, but website stubbornly changes rule to B3/S23:T10,10.
Re: CHALLENGE: Find the longest-lived GOL pattern w/ the fewest / most alive cells as its initial state
Posted: August 17th, 2026, 7:53 am
by hash0
Huh... that's an issue with the RLE parser, I'll look into it. Can you submit your entry in binary for the time being?
Re: CHALLENGE: Find the longest-lived GOL pattern w/ the fewest / most alive cells as its initial state
Posted: August 17th, 2026, 8:07 am
by Vort
I found better workaround.
Usually I press Ctrl+A, Ctrl+C in Golly. Then wrong parsing happens.
But if I select everything manually (including empty areas) and press Ctrl+C, then parsing works fine.
Re: CHALLENGE: Find the longest-lived GOL pattern w/ the fewest / most alive cells as its initial state
Posted: August 17th, 2026, 2:46 pm
by hash0
I've been running a miner half a day for 100 million plus patterns on the 20x20 torus but could not find any new Agar (or the known Agar p314, for that matter). The currently longest-lived pattern there runs for 2,813 unique generations (owned by Vort, no surprise there...).
Is it possible that Agars only manifest on the 16x16 but not on higher-order tori?!
Re: CHALLENGE: Find the longest-lived GOL pattern w/ the fewest / most alive cells as its initial state
Posted: August 17th, 2026, 3:05 pm
by Vort
hash0 wrote: August 17th, 2026, 2:46 pm... on the 20x20 torus but could not find any new Agar ...
I saw this thing, but I don't know if it is new or not:
Code:
Select all
x = 20, y = 20, rule = B3/S23:T20,20
7$13b2o$13bobo$3b3o9bo$5bo7b3o$3bobo$3b2o!
hash0 wrote: August 17th, 2026, 2:46 pm... Is it possible that Agars only manifest on the 16x16 but not on higher-order tori?! ...
At least it is possible to put 4 copies of 16x16 P5024 into 32x32 (at the cost of 4x population increase of course).
Re: CHALLENGE: Find the longest-lived GOL pattern w/ the fewest / most alive cells as its initial state
Posted: August 20th, 2026, 1:07 am
by Vort
bobinou wrote: August 19th, 2026, 6:26 pm
Whilst mining for this i have found the currently longest running non agar submitted to the Registry, with a runtime of 2382. This is almost 100 iterations more than my prev. record holder:
Fifty Trillion
Code: Select all
x = 16, y = 16, rule = B3/S23:T16,16
b3o5b2ob2obo$bob2ob3obobo3b$2obo5b2obob2o$obo3b6o2bob$2b2ob3obo3b3o$4bobobob2o2bob$o6bo2bobob2o$3b3obo2bo2bobo$bo2bobob3o4bo$3b2o2b2ob3obob$3bo3bo3b3obo$3o6b5obo$2bo2b2obo2b2obob$bob2o2bobo3b2ob$obob3o7bob$2bo3bo2b3o3bo!
It means
2312 gen is possible with just 21 pop!
upd: Loaf can be rewound as well, allowing for 19 pop result.
Re: CHALLENGE: Find the longest-lived GOL pattern w/ the fewest / most alive cells as its initial state
Posted: August 20th, 2026, 3:54 am
by bobinou
We might be reaching a limit on whats possible on the 16x16 with the current miners. Finding the 2300+ took a really long time.
Re: CHALLENGE: Find the longest-lived GOL pattern w/ the fewest / most alive cells as its initial state
Posted: August 20th, 2026, 4:23 am
by Vort
bobinou wrote: August 20th, 2026, 3:54 am
We might be reaching a limit on whats possible on the 16x16 with the current miners. Finding the 2300+ took a really long time.
I suspect filling grid with random dust may be not the best way of making progress.
Can 2312 gen 19 pop result be rediscovered faster with better algorithms?
For example, by position randomization of known small patterns (gliders / blinkers / ...).
Re: CHALLENGE: Find the longest-lived GOL pattern w/ the fewest / most alive cells as its initial state
Posted: August 20th, 2026, 4:49 am
by bobinou
Def. worth a try. If i am not mistaken there is a database of the known patterns, taking those and scattering them on the board might work. I will give that a try today
Re: CHALLENGE: Find the longest-lived GOL pattern w/ the fewest / most alive cells as its initial state
Posted: August 20th, 2026, 5:17 am
by Vort
bobinou wrote: August 20th, 2026, 4:49 am... If i am not mistaken there is a database of the known patterns ...
For stable patterns and oscillators, yes, Catagolue.
However, it is important to include evolutionary sequences as well:
Teardrop:
Code:
Select all
x = 4, y = 5, rule = B3/S23
b3o2$o$o$o!
Lumps of muck:
Code:
Select all
x = 4, y = 3, rule = B3/S23
b3o2$3o!
T-tetromino (predecessor):
Code:
Select all
x = 3, y = 3, rule = B3/S23
3o2$bo!
... and many others.
It may be possible to get them from 3g octohash database, but I'm not sure about it. Also there may be better source for them.
upd: I know that for densely packed still lifes + spark result is about the same as for dust. But I did not tried sparse packing and did not tried to include evolutionary sequences.
Re: CHALLENGE: Find the longest-lived GOL pattern w/ the fewest / most alive cells as its initial state
Posted: August 22nd, 2026, 9:58 am
by hash0
... if you guys continue with this, I'll soon have to implement a submission throttle to protect my API endpoints!
Re: CHALLENGE: Find the longest-lived GOL pattern w/ the fewest / most alive cells as its initial state
Posted: August 22nd, 2026, 5:10 pm
by bobinou
well, with so many new slots to fill we need to submit a ton of new patterns
Re: CHALLENGE: Find the longest-lived GOL pattern w/ the fewest / most alive cells as its initial state
Posted: August 23rd, 2026, 3:28 am
by Vort
hash0 wrote: August 22nd, 2026, 9:58 am
... if you guys continue with this, I'll soon have to implement a submission throttle to protect my API endpoints!
bobinou wrote: August 22nd, 2026, 5:10 pm
well, with so many new slots to fill we need to submit a ton of new patterns
By the way, it is not great to wait 2-3 minutes until MANUAL_SEQUENCE_ENTRY UI verifies ~5k solution.
It effectively forces people to use "hacks", which is better to be avoided.
(But I'm not saying that people should be forced to use UI either)
Re: CHALLENGE: Find the longest-lived GOL pattern w/ the fewest / most alive cells as its initial state
Posted: August 23rd, 2026, 3:53 am
by bobinou
An official API would be the best middle ground. People with external miners can easily use it, and people mining via the website won’t even find a pattern this high to have to wait minutes.
Re: CHALLENGE: Find the longest-lived GOL pattern w/ the fewest / most alive cells as its initial state
Posted: August 23rd, 2026, 4:01 am
by hash0
I hear you, gentlepeople, and I think you are right. I'll place an API (and a background checker that bypasses the runtime output to the browser so that you don't have to wait) on my feature wishlist.
Re: CHALLENGE: Find the longest-lived GOL pattern w/ the fewest / most alive cells as its initial state
Posted: August 23rd, 2026, 6:29 am
by Vort
bobinou wrote: August 23rd, 2026, 3:53 am... and people mining via the website won’t even find a pattern this high to have to wait minutes. ...
Oh, I thought 32x32 were available for website mining. Looks like they are not.
Also during my attempts to find how to do this, I found several problems with 32x32 display: at least
REGISTRY and REGISTERED_GLYPHS_LOG display them incorrectly.
Re: CHALLENGE: Find the longest-lived GOL pattern w/ the fewest / most alive cells as its initial state
Posted: August 25th, 2026, 1:01 pm
by hash0
Hey, Vort! Admit it: You're an alt account for Grigori Perelman, right?
Re: CHALLENGE: Find the longest-lived GOL pattern w/ the fewest / most alive cells as its initial state
Posted: September 5th, 2026, 2:11 am
by hash0
Gentlepeople and GOL enthusiasts,
I have updated the Glyph Registry
https://lifehashes.net/login/registry.php to make filtering by Grid Size possible:
- registry_gridsize_update.png (163.17 KiB) Viewed 143 times
With the introduction of the Sparse-Dense Challenge (
https://lifehashes.net/login/sparse-dense-challenge.php) that plays on torus grid sizes 16x16, 20x20, 24x24, 28x28, 32x32, it became necessary to distinguish these distinct "classes" more easily. Hope this makes the entire experience more convenient for you.
Also, API for submitting patterns is in the works, but please don't nail me to a release date for this feature.
Re: CHALLENGE: Find the longest-lived GOL pattern w/ the fewest / most alive cells as its initial state
Posted: September 5th, 2026, 2:16 am
by hash0
Update on the current state of the affairs:
The currently longest-lived known pattern on the torus is "it_took_soooooooooooo_long_help" (32x32), submitted by Phnx and which runs for 9,786 unique generations (no configuration previously held). This is remarkable for two reasons:
1. Given the discovery of the p314 Agar on the 16x16 torus (where the currently longest-lived pattern is "Gentle Hypergiant" by bobinou with 5,629 generations), I had assumed that bigger tori dimensions would yield other previously unknown Agar patterns. This does not appear to be the case (so far).
2. I had naively assumed that increasing the torus gride size would massively (exponentially?) increase the longevity of patterns - this, too, appears to be false. Notice that the longest-lived 16x16 pattern runs for 5,629 generations but the longest-lived 32x32 pattern "only" for 9,786 generations. Despite having four times as much space to live and breathe, the increase in unique generations is not even 100%. Go figure...
Re: CHALLENGE: Find the longest-lived GOL pattern w/ the fewest / most alive cells as its initial state
Posted: September 5th, 2026, 3:17 am
by Vort
hash0 wrote: September 5th, 2026, 2:16 am
... I had assumed that bigger tori dimensions would yield other previously unknown Agar patterns. This does not appear to be the case (so far). ...
... I had naively assumed that increasing the torus gride size would massively (exponentially?) increase the longevity of patterns - this, too, appears to be false. Notice that the longest-lived 16x16 pattern runs for 5,629 generations but the longest-lived 32x32 pattern "only" for 9,786 generations. Despite having four times as much space to live and breathe, the increase in unique generations is not even 100%. Go figure...
Such strangeness appear because you compare agar and non-agar variations.
Non-agar 16x16 2382 gen has about 4 times lower lifespan than 32x32 9786 gen.
As for exponentiality, I expect it as well (in fact it is proven for larger grid sizes).
But it is probably just out of reach yet for 32x32.
upd: regarding agar on larger grids: looks like no one rediscovered
32x32 P5024 yet with random searches.
I wonder what random will reveal first: new 32x32 agar or "old" P5024.