Blockic Seeds

[knightlife]

Even faster LWSS:

Code: Select all

x = 59, y = 24, rule = B3/S23
bo$2bo$3o5$48b2o$48b2o3$2b2o4b2o34b2o$2b2o4b2o34b2o2$53b2o$5b2o46b2o2b
2o$5b2o8b2o40b2o$15b2o$54b2o$54b2o$11b2o$11b2o30b2o$42bobo$44bo!

[knightlife]

Clean LWSS + G seed with just 4 blocks:

Code: Select all

x = 30, y = 17, rule = B3/S23
6bo$7bo$5b3o16b2o$24b2o4$28b2o$28b2o5$2o$2o$16b2o$16b2o!

Still would like to find a clean *WSS by itself from a 4-block seed or even a 3-block seed!

[knightlife]

Well, I didn't expect to find it so soon.
3-block seed for a clean fwd LWSS:

Code: Select all

x = 22, y = 20, rule = B3/S23
bo$2bo$3o2$12b2o$12b2o10$20b2o$20b2o2$14b2o$14b2o!

It is very compact and fast just as I envisioned it.
Found it with the Seeds of Destruction JAVA program.

EDIT:
Mod yields extra glider:

Code: Select all

x = 22, y = 20, rule = B3/S23
bo$2bo$3o7b2o$11bo$11bobo$12b2o10$20b2o$20b2o2$14b2o$14b2o!

[calcyman]

Congratulations.

[Extrementhusiast]

Four blocks delay a glider by 45 gens:

Code: Select all

x = 27, y = 30, rule = B3/S23
13b2o$13b2o5$2o12b2o$2o12b2o2b2o$18b2o8$6b3o15b3o$8bo17bo$7bo17bo9$13b
2o$14b2o$13bo!

[dvgrn]

Well, I didn't expect to find it so soon.
3-block seed for a clean fwd LWSS:

Beautiful! I was seeing lots of LWSS signatures when I looked at three- and four-block seeds with the SODGame, but couldn't come up with even a four-block clean version. This one is really impressive!

It's too bad that one of the blocks barely obstructs the alternate glider lane from the NE that could set off the pi-heptomino explosion, to produce a backward LWSS seed. But there are certainly lots of other possibilities for building that pi -- two blocks, or some other still life, instead of the initial target block. (Haven't had time to play around with that yet.)

This seed will make it much easier to build LWSSes with a construction arm. h4tter has been working on this offline for a while, with good results. So far I've seen 41-glider and 33-glider recipes for backward and forward LWSSes, respectively:

Code: Select all

backward_LWSS_seed_41:E9 E29 E23 E23 O27 E23 O11 O19 E49 E17 E37 E31 E45 E43 E17 E47 E29 E27 E47 E-3 E1 E17 E-7 E41 E47 O67 O51 O61 O25 E33 E53 E29 E35 E39 E29 O37 E77 E49 E69 E59 E-23
forward_LWSS_seed_33:E9 E29 E23 E41 E27 E7 E23 E35 E27 E47 E23 E17 O35 E39 O37 E47 E67 E61 E67 E83 O61 E59 E57 O75 E61 E81 E75 E73 O101 E75 O97 E53 E-19

Here's an RLE version, with one of the recipes in mirror-image:

Code: Select all

x = 6023, y = 10936, rule = B3/S23
2o$2o23$20b2o$20bobo$20bo158$170b2o$170bobo$170bo145$320b2o$320bobo$
320bo148$470b2o$470bobo$470bo150$621bo$620b2o$620bobo146$770b2o$770bob
o$770bo142$921bo$920b2o$920bobo152$1071bo$1070b2o$1070bobo163$1220b2o$
1220bobo$1220bo132$1370b2o$1370bobo$1370bo158$1520b2o$1520bobo$1520bo
145$1670b2o$1670bobo$1670bo155$1820b2o$1820bobo$1820bo147$1970b2o$
1970bobo$1970bo135$2120b2o$2120bobo$2120bo163$2270b2o$2270bobo$2270bo
139$2420b2o$2420bobo$2420bo147$2570b2o$2570bobo$2570bo158$2720b2o$
2720bobo$2720bo123$2870b2o$2870bobo$2870bo150$3020b2o$3020bobo$3020bo
156$3170b2o$3170bobo$3170bo136$3320b2o$3320bobo$3320bo172$3470b2o$
3470bobo$3470bo151$3620b2o$3620bobo$3620bo158$3771bo$3770b2o$3770bobo
140$3921bo$3920b2o$3920bobo153$4071bo$4070b2o$4070bobo130$4221bo$4220b
2o$4220bobo152$4370b2o$4370bobo$4370bo158$4520b2o$4520bobo$4520bo136$
4670b2o$4670bobo$4670bo151$4820b2o$4820bobo$4820bo150$4970b2o$4970bobo
$4970bo143$5120b2o$5120bobo$5120bo152$5271bo$5270b2o$5270bobo168$5420b
2o$5420bobo$5420bo134$5570b2o$5570bobo$5570bo158$5720b2o$5720bobo$
5720bo143$5870b2o$5870bobo$5870bo107$6020b2o$6020bobo$6020bo103$12b2o$
12b2o19$36b3o$36bo$37bo148$196b3o$196bo$197bo148$343b3o$343bo$344bo
148$502b3o$502bo$503bo148$645b3o$645bo$646bo148$785b3o$785bo$786bo148$
943b3o$943bo$944bo148$1099b3o$1099bo$1100bo148$1245b3o$1245bo$1246bo
148$1405b3o$1405bo$1406bo148$1543b3o$1543bo$1544bo148$1690b3o$1690bo$
1691bo148$1850b2o$1849b2o$1851bo148$2001b3o$2001bo$2002bo148$2151b2o$
2150b2o$2152bo148$2305b3o$2305bo$2306bo148$2465b3o$2465bo$2466bo148$
2612b3o$2612bo$2613bo148$2765b3o$2765bo$2766bo148$2923b3o$2923bo$2924b
o148$3063b2o$3062b2o$3064bo148$3211b3o$3211bo$3212bo148$3360b3o$3360bo
$3361bo148$3520b2o$3519b2o$3521bo148$3662b3o$3662bo$3663bo148$3822b3o$
3822bo$3823bo148$3969b3o$3969bo$3970bo148$4118b3o$4118bo$4119bo148$
4283b2o$4282b2o$4284bo148$4419b3o$4419bo$4420bo148$4581b2o$4580b2o$
4582bo148$4708b3o$4708bo$4709bo148$4822b3o$4822bo$4823bo!

These syntheses could probably have been improved even while still building the old five-block LWSS seed -- so the construction salvo for the three-block seed will be much shorter.

Note: in the old format the above would have looked like

Code: Select all

backward_LWSS_seed_41:9,29,23,23;27,23;11;19,49,17,37,31,45,43,17,47,29,27,47,-3,1,17,-7,41,47;67;51;61;25,33,53,29,35,39,29;37,77,49,69,59,-23
forward_LWSS_seed_33:9,29,23,41,27,7,23,35,27,47,23,17;35,39;37,47,67,61,67,83;61,59,57;75,61,81,75,73;101,75;97,53,-19

-- but it got to be too annoying to have to distinguish commas from semicolons on a laptop screen (at least without cleaning the screen a little more regularly!)

I've updated the slow-salvo generator script so that it accepts the new format, and plan to use the (E)ven/(O)dd format going forward. Generally, P2 slow-salvo recipes will be normalized to start with an "E", though swapping all E's and O's would also produce working recipes.

I'm also shifting the recipe format for elbow operations -- i.e., glider pairs and singletons aimed at an elbow. In that case, all you need to specify for a P2 slow salvo is the parity and the relative timing of the two gliders. I'm using lowercase 'e' and 'o' as the standard parity markers for these recipes; with any luck that will make it easier to tell the two types of recipes apart. Here's the new version of the script to build glider-pair salvos.

[dvgrn]

It's too bad that one of the blocks barely obstructs the alternate glider lane from the NE that could set off the pi-heptomino explosion, to produce a backward LWSS seed. But there are certainly lots of other possibilities for building that pi -- two blocks, or some other still life, instead of the initial target block. (Haven't had time to play around with that yet.)

Just had a quick look at other common still lifes that can produce a clean pi explosion when hit by a glider. Ponds, ships, boats, beehives, and blinkers are the ones I know about. Many of them can be trivially substituted in for the block -- just line up the two-cell part with the top of the block, and it will generally work just the same.

Most of these don't improve on the original block, though -- the mirror-image input lane is blocked in exactly the same way if you use a boat, beehive, ship, or pond -- and some have a bigger spark in one or both orientations and the pi insertion doesn't work anyway.

The exception is the blinker! Oddly enough, the original input direction doesn't quite work out, but a glider from the northeast does:

Code: Select all

#C G+2B4+B3 -> backward LWSS
x = 10, y = 14, rule = B3/S23
2o$2o$8bo$7bo$7b3o7$8b2o$8b2o$2b3o!

Maybe there's a pure Blockic backward LWSS seed with four blocks -- I haven't tried any two-block constellations as targets yet.

[codeholic]

Maybe there's a pure Blockic backward LWSS seed with four blocks -- I haven't tried any two-block constellations as targets yet.

I've tried some glider+2 blocks collisions that result in a pi-heptomino, but none of those pi-heptominoes fit into this reaction. But a block+hive worked well.

Code: Select all

x = 13, y = 20, rule = B3/S23
3b2o$3b2o7$6bo$6bobo$6b2o$2o9b2o$2o9b2o4$6bo$5bobo$5bobo$6bo!

[codeholic]

Is this one known? MWSS from 3 blocks.

Code: Select all

x = 9, y = 9, rule = B3/S23
6bo$2o4bobo$2o4b2o5$b2o4b2o$b2o4b2o!

Found with gencols (1st collision at gen 10, 2nd between 10 and 40, 3nd between 10 and 80, result taken at gen 200). No other 3 blocks collisions resulting in spaceships except the LWSS synthesis already mentioned above were found.

By the way, I think, that such things would be much better to do in the wiki. At least, it would be easier to organize stuff and find it later when needed.

@Sokwe Could you suggest where in the wiki it would be the best to do it?

[knightlife]

Is this one known? MWSS from 3 blocks.

Code: Select all

x = 9, y = 9, rule = B3/S23
6bo$2o4bobo$2o4b2o5$b2o4b2o$b2o4b2o!

Ultra compact (and fast)! The reaction "explosion" just barely exceeds the 8 x 9 bounding box!

A clunky 5-block seed for a backward LWSS until a better one is found:

Code: Select all

x = 18, y = 22, rule = B3/S23
10b2o$10b2o3$6b2o$6b2o$bo$2bo$3o7$16b2o$12b2o2b2o$12b2o3$15b2o$15b2o!

EDIT:
This one is "better" (posted elsewhere):

Code: Select all

x = 20, y = 22, rule = B3/S23
bo$2bo$3o12$17b2o$12b2o3b2o$12b2o$8b2o$8b2o$18b2o$3b2o13b2o$3b2o!

[dvgrn]

Is this one known? MWSS from 3 blocks.

Wow! That certainly looks like something that ought to have been known by now, but I can't recall having seen it before.

G-to-MWSS and G-to-HWSS converters are especially good for building self-destruct chain reactions. MWSSes in particular are the cheapest moving objects with significant sparks. They can clean up multiple objects in passing, but it's also possible to "Heisenburp" an MWSS or HWSS and spin off multiple output gliders. You can add and subtract outputs without affecting the *WSS at all:

Code: Select all

x = 27, y = 85, rule = B3/S23
14bo$8b2o4bobo$8b2o4b2o5$9b2o4b2o$9b2o4b2o14$7b2o$7bobo$b2o5b2o$o2bo$b
2o16$7b2o$7bobo$8b2o2$5b2o$4bobo$5bo11$25b2o$19bo5b2o$19b3o$22bo$21b2o
11$22b2o$22b2o7$13b2o$13b2o2b2o$17b2o!

With an MWSS I don't think there's a pure-Blockic way to do this kind of thing, though; the two-blocks-plus-eater bait is as close as I could come. It's easier with an HWSS, if I recall correctly...! Might a four-block gencols search for HWSS outputs be vaguely within the realm of possibility?

EDIT: You can manage a Blockish MWSS Heisenburp, if you allow bi-blocks -- I think that's cheating a little, but really bi-blocks are pretty easy for a constructor arm to build:

Code: Select all

x = 23, y = 77, rule = B3/S23
6bo$2o4bobo$2o4b2o5$b2o4b2o$b2o4b2o15$11b2ob2o$11b2ob2o4$14b2o$14b2o$
20b2o$20b2o10$21b2o$21b2o$11b2ob2o$11b2ob2o4$14b2o$14b2o25$5b2o$5b2o2b
2o$9b2o!

EDIT 2: Well, never mind bi-blocks, I think -- it is technically possible to find a pure-Blockic Heisenburp for an HWSS, but they'd be expensive to construct:

Code: Select all

x = 36, y = 89, rule = B3/S23
bobo$o$o3bo$o3bo$o$o2bo$3o2$10b2o$10b2o2$6b2o$6b2o2b2o$10b2o2b2o$14b2o
$7b2o$7b2o2b2o$11b2o6$17b2o$17b2o15$14b2o$14b2o18b2o$34b2o5$11b2o$7b2o
2b2o$7b2o2$10b2o$6b2o2b2o$6b2o9$9b2o$9b2o25$4b2o$4b2o!

There are many other versions, of course, but they all seem to require the same first four blocks. Anyone see a less messy all-block Heisenburp for a passing HWSS?

[Extrementhusiast]

Five-block seed for a down-loaf with tail plus other debris:

Code: Select all

x = 75, y = 42, rule = LifeHistory
7$47.2A$47.2A2$64.2A$64.2A5$67.2A$67.2A4$61.2A$61.2A5$51.2A$51.2A5$
24.2A$23.A.A$25.A!

The "unusual still life" can actually be made with just four blocks:

Code: Select all

x = 40, y = 33, rule = LifeHistory
29.2A$29.2A3$38.2A$38.2A2$29.2A$29.2A5$27.2A$27.2A16$.2A$A.A$2.A!

[codeholic]

A clunky 5-block seed for a backward LWSS until a better one is found:

But on the other hand the underlying reaction is quite easy to trigger, e. g.

Code: Select all

x = 421, y = 26, rule = B3/S23
18b2o37bo48bo49bo49bo49bo45bo49bo49bo$18b2o35bobo46bobo47bobo47bobo47b
obo10b2o31bobo47bobo47bobo$56b2o47b2o48b2o48b2o11b2o35b2o10b2o32b2o48b
2o48b2o$2bo167b2o46b2o$obo11b2o154b2o$b2o11b2o$18b2o34b2o$18b2o34b2o
98b2o48b2o48b2o$72b2o30b2o48b2o48b2o48b2o$72b2o30b2o$50b2o$50b2o98b2o
48b2o48b2o$100b2o48b2o48b2o48b2o$100b2o71b2o48b2o48b2o$65b2o49b2o46b2o
7b2o39b2o7b2o39b2o7b2o33b2o48b2o2b2o42b2o4b2o$65b2o49b2o46b2o48b2o48b
2o42b2o48b2o2b2o42b2o4b2o$6b2o306b2o50b2o$6b2o3b2o107b2o192b2o6b2o42b
2o$11b2o49b2o51b2o3b2o200b2o85b2o$62b2o51b2o292b2o8b2o$368b2o49b2o$
311b2o5b2o48b2o$311b2o5b2o$415b2o$364b2o49b2o$364b2o!

[knightlife]

B-to-Pi works but leaves that pesky block:

Code: Select all

x = 11, y = 18, rule = B3/S23
b2o$b2o6$2o$2o3$9b2o$9b2o3$3bo$2b3o$2bob2o!

[codeholic]

A fast and compact backward MWSS with 3 hives.

Code: Select all

x = 15, y = 18, rule = B3/S23
2bo$obo$b2o7$3b2o4bo$2bo2bo2bobo$3b2o3bobo$9bo2$13bo$12bobo$12bobo$13b
o!

[knightlife]

This old pattern is easier to construct? (backward LWSS)

Code: Select all

x = 17, y = 24, rule = B3/S23
bo$2bo$3o13$6b2o$6b2o3$2b2o$2b2o11bo$14bobo$14bobo$15bo!

[dvgrn]

This old pattern is easier to construct? (backward LWSS)

Code: Select all

x = 17, y = 24, rule = B3/S23
bo$2bo$3o13$6b2o$6b2o3$2b2o$2b2o11bo$14bobo$14bobo$15bo!

Probably a little more expensive than three blocks and a little cheaper than three beehives, yes -- but an LWSS may not be quite as useful as an MWSS, anyway, so I think the three-block MWSS seed might win the competition for now.

Really until we have a working compiler, or a search utility, for unidirectional slow salvos building small constellations, it's very hard to tell for sure what's most efficient. For example, a recipe that just happens to build a couple of blocks or other still lifes simultaneously will cut a huge percentage off the total cost.

Out there somewhere, of course, there are great piles of recipes like that. But organizing a billion-entry "unidirectional fluff" database is quite an undertaking. Just coming up with a workable optimizing algorithm for this project seems to be an NP-hard problem, as Paul Chapman has noted -- never mind actually having to run the search until the Sun goes cold...!

The only rule of thumb I really have so far is that still lifes with more possible orientations are correspondingly more difficult to construct (see the "score" suggestions, third paragraph from the end).

Mind you, depending on the problem you're trying to solve, a perfectly good compiler may be available -- i.e., for Geminoids with two construction arms. I'm stubbornly focused on one-arm construction at the moment, so it's not on my priority list to come up with an efficient design for one of these, or to solve any of the associated timing and encoding problems... but such a design will probably produce a significantly smaller Geminoid (in terms of population or total length, though the replicator units will be much bigger). And there's already a compiler script available...!

[Sokwe]

@Sokwe Could you suggest where in the wiki it would be the best to do it?

I would suggest doing it under your user page (e.g. User:Codeholic/blockic_seeds). Also, congratulations on the 3-block MWSS seed.

The 3-block LWSS seed almost looks like it could be used to make a stable pi-to-LWSS converter, but unfortunately, the reaction gets too close to where the first block was. It seems almost surprising that no simple stable H-to-LWSS converter has yet been found.

[knightlife]

11 gliders from 7 blocks, ratio 1.57

Code: Select all

x = 46, y = 58, rule = LifeHistory
.A$2.A$3A2$12.2A$12.2A18$6.2A$6.2A12$24.D19.2A$22.D.D19.2A$22.3D$22.D
2$5.2A4.2A$5.2A4.2A$39.2A$39.2A12$18.2A$18.2A!

This uses a new component: a blockic Herschel splitter.

Code: Select all

x = 20, y = 20, rule = LifeHistory
.2A$.2A12$19.A$17.A.A$17.3A$17.A2$2A4.2A$2A4.2A!

[codeholic]

3-block glider duplicator seeds.

Code: Select all

x = 214, y = 127, rule = B3/S23
159b2o$159b2o6$108b2o$108b2o2$205b2o$205b2o2$3b2o4bo49bo49bo49bo49bo$
3b2o4bobo47bobo47bobo47bobo47bobo$9b2o48b2o48b2o48b2o48b2o$53b2o48b2o
48b2o48b2o$53b2o48b2o48b2o48b2o2$2o$2o$56b2o102b2o$56b2o2b2o98b2o$60b
2o51b2o$113b2o2$206b2o$206b2o$20b2o$20b2o15$117b2o$117b2o12$6b2o$6b2o
5$9bo49bo49bo49bo49bo$9bobo47bobo47bobo47bobo47bobo$9b2o48b2o48b2o48b
2o12b2o34b2o$3b2o48b2o48b2o48b2o18b2o28b2o$3b2o48b2o48b2o48b2o48b2o2$
68b2o$68b2o$110b2o100b2o$110b2o42b2o56b2o$5b2o147b2o$5b2o2$208b2o$65b
2o141b2o$65b2o13$16b2o$16b2o17$58b2o$58b2o3$59bo$9bo49bobo$9bobo47b2o$
9b2o42b2o$3b2o48b2o$3b2o4$10b2o$10b2o2$65b2o$65b2o!

[dvgrn]

3-block glider duplicator seeds.

Very nice collection! I recognize a few of the ones I've been using, but most are new to me.

So... are there no G->3G reactions from 3-block constellations, according to gencols?

[codeholic]

So... are there no G->3G reactions from 3-block constellations, according to gencols?

Well, I thought about that, but it would be quite tricky to find this out without modifying gencols' code, because it cannot filter by objects, if the pattern consists of several of them. For finding those G->2G reactions I filtered for asynchronous output with exactly 10 cells, but still I had to filter out G->G+boat manually. With 15 cells there will be much more variants of G+still lifes, I'm afraid.

[dvgrn]

Well, I thought about that, but it would be quite tricky to find this out without modifying gencols' code, because it cannot filter by objects, if the pattern consists of several of them. For finding those G->2G reactions I filtered for asynchronous output with exactly 10 cells, but still I had to filter out G->G+boat manually. With 15 cells there will be much more variants of G+still lifes, I'm afraid.

False positives I generally don't mind so much... as long as it's possible to filter for 15-cell aperiodic patterns, it seems like it should only take half an hour or so to write a Python script to post-filter the gencols output, to flag all the candidate patterns with no still-life component.

The script will be embarrassingly inefficient and will take hours to run if there are many megabytes of output, but I have a spare computer and lots of spare curiosity...! What exactly are you calling gencols with for this search?

[codeholic]

The script will be embarrassingly inefficient and will take hours to run if there are many megabytes of output, but I have a spare computer and lots of spare curiosity...! What exactly are you calling gencols with for this search?

This is the script I used for finding G+3blocks->2G collisions

Code: Select all

./gencols -pat obj/block.life obj/glider_sw.life -tc 10 10 > glsplit/gb.col
./gencols -pat glsplit/gb.col obj/block.life -tc 10 40 > glsplit/gb2.col
./gencols -pat glsplit/gb2.col obj/block.life -tc 10 100 -gen 200 -filt a -leq 10 -geq 10 > glsplit/result.col

[codeholic]

Having said that, I realized, that I missed a large class of ouputs, namely gliders on parallel tracks. I'm correcting myself now. Sorry the list is unsorted and may contain duplicate collisions or patterns where a glider could not come from infinity.

Code: Select all

x = 257, y = 917, rule = B3/S23
10b2o71b2o77bo74bo$10b2o71b2o77bobo72bobo$162b2o73b2o$6bo149b2o67b2o4b
2o$6bobo141b2o4b2o67b2o4b2o$6b2o142b2o$2o169b2o73b2o$2o79bo89b2o73b2o$
81bobo$81b2o$75b2o$75b2o6$8b2o75b2o$8b2o75b2o57$13bo76bo75bo74bo$13bob
o74bobo73bobo72bobo$13b2o75b2o74b2o73b2o$7b2o75b2o74b2o73b2o$7b2o75b2o
64b2o8b2o73b2o$150b2o73b2o$2o20b2o51b2o22b2o74b2o48b2o23b2o$2o20b2o51b
2o22b2o74b2o73b2o68$17bo72bo77bo75bo$17bobo70bobo75bobo73bobo$17b2o56b
2o13b2o76b2o74b2o$11b2o62b2o7b2o64b2o10b2o74b2o$11b2o71b2o64b2o10b2o
61b2o11b2o$225b2o$26b2o71b2o76b2o74b2o$26b2o71b2o76b2o74b2o$2o$2o66$
19bo72bo76bo75bo$19bobo70bobo74bobo73bobo$19b2o54b2o15b2o75b2o74b2o$
13b2o60b2o9b2o75b2o74b2o$13b2o71b2o75b2o60b2o12b2o$225b2o$2o26b2o71b2o
75b2o74b2o$2o26b2o71b2o75b2o74b2o2$150b2o$150b2o65$20bo75bo73bo56b2o$
20bobo73bobo71bobo54b2o$20b2o74b2o72b2o$2o12b2o74b2o72b2o$2o12b2o74b2o
72b2o2$29b2o44b2o28b2o72b2o$29b2o44b2o28b2o72b2o50bo$231bobo$231b2o$
225b2o$150b2o73b2o$150b2o8$233b2o$233b2o54$12bo74bo75bo76bo$12bobo6b2o
64bobo6b2o52b2o11bobo6b2o51b2o13bobo6b2o$12b2o7b2o64b2o7b2o52b2o11b2o
7b2o51b2o13b2o7b2o$2o4b2o73b2o74b2o75b2o$2o4b2o67b2o4b2o74b2o75b2o$75b
2o70$16bo74bo73bo59b2o$16bobo6b2o64bobo6b2o63bobo6b2o49b2o15bo$2o14b2o
7b2o64b2o7b2o63b2o7b2o66bobo6b2o$2o8b2o63b2o8b2o72b2o81b2o7b2o$10b2o
63b2o8b2o72b2o75b2o$150b2o84b2o$150b2o69$18bo56b2o92bo74bo$18bobo6b2o
46b2o73b2o17bobo6b2o64bobo6b2o$18b2o7b2o65bo55b2o17b2o7b2o64b2o7b2o$
12b2o80bobo6b2o58b2o60b2o11b2o$2o10b2o80b2o7b2o58b2o60b2o11b2o$2o86b2o
$88b2o69$17bo77bo74bo75bo$17bobo6b2o67bobo6b2o64bobo6b2o44b2o19bobo6b
2o$17b2o7b2o67b2o7b2o64b2o7b2o44b2o19b2o7b2o$11b2o62b2o12b2o73b2o74b2o
$11b2o62b2o12b2o59b2o12b2o74b2o$2o148b2o$2o69$2o81b2o73b2o75b2o$2o81b
2o73b2o75b2o2$231bo$20bo210bobo$20bobo6b2o200b2o$20b2o7b2o50bo143b2o$
14b2o65bobo72bo68b2o$14b2o65b2o73bobo$75b2o79b2o$75b2o73b2o$150b2o6$
160b2o71b2o$160b2o71b2o2$77b2o$77b2o54$2b2o79b2o67b2o71b2o$2b2o79b2o
67b2o71b2o3$156bo$156bobo$81bo68b2o4b2o73bo$6bo74bobo66b2o79bobo$6bobo
72b2o142b2o4b2o$6b2o67b2o148b2o$2o73b2o$2o4$157b2o71b2o$157b2o71b2o4$
8b2o67b2o$8b2o67b2o54$7b2o66b2o78b2o70b2o$7b2o66b2o9bo68b2o70b2o$86bob
o$80b2o4b2o$80b2o74bo74bo$90b2o64bobo72bobo$6bo83b2o58b2o4b2o67b2o4b2o
$6bobo141b2o73b2o$2o4b2o$2o6$2b2o146b2o80b2o$2b2o146b2o80b2o59$7b2o71b
2o68b2o73b2o$7b2o71b2o68b2o9bo63b2o$161bobo$155b2o4b2o$81bo73b2o$81bob
o81b2o$6bo68b2o4b2o82b2o64bo$6bobo66b2o154bobo$2o4b2o217b2o4b2o$2o223b
2o6$2b2o71b2o153b2o$2b2o71b2o153b2o!

A large number of reactions is based on these G+2 blocks->2G+B-heptomino reactions and they differ only in the way how the last block collides with B-heptomino and destroys it:

Code: Select all

x = 67, y = 30, rule = B3/S23
12$48bo$48bobo$48b2o$14bo27b2o$14bobo6b2o17b2o$14b2o7b2o$8b2o47b2o$8b
2o47b2o!

Since B-heptomino may react with another block to produce a glider, there is a positive answer to your G+3 blocks->3G question:

Code: Select all

x = 33, y = 11, rule = B3/S23
22bo$22bobo$22b2o$16b2o$16b2o2$31b2o$31b2o2$2o$2o!

EDIT: Actually they are already listed in knightlife's post here: viewtopic.php?f=2&t=1065

EDIT: These two are interesting because they send gliders onto the same track:

Code: Select all

x = 60, y = 23, rule = LifeHistory
2.2A$2.2A54.2A$58.2A5$6.A49.A$6.A.A47.A.A$6.2A48.2A$2A48.2A$2A48.2A9$
8.2A$8.2A42.2A$52.2A!