HBK as Caterpillar

[biggiemac]

Looking at the "Parallel HBK" in comparison to the Waterbear, a thought came to mind.

The mechanism used for current HBKs is a glider traveling in series through the half bakeries. (I would personally call both designs serial HBKs, but I realize switching to parallel slow salvo construction was a major factor in reducing the size). In contrast, the Waterbear gets its speed because the base reaction occurs every 79 generations (born from necessity). The reaction is effectively going on in parallel throughout the entire ship. In a truly parallel HBK, then, repeating the base reaction more quickly could cause an immense speedup. This is entirely possible, two things can climb as long as they are in relative motion.

Caterpillar: Crawler climbing (stationary) oscillators.
Waterbear: Crawler climbing gliders (probably the hardest combination).
HBK caterpillar: Half bakery (usually still life, turned into crawler here) climbing gliders.

Due to the intermediate stationary phase, we can adjust the period as necessary. The main mechanism has a repeat time of 62, but the glider-producing combination has a much slower 123. I build a (receding!) (6,3)c/123 helix (n=1, best I could do was 7 *WSS, haven't done anything to multiply tracks), and here is an example of what all of this means.

Code: Select all

x = 367, y = 699, rule = B3/S23
47$263b3o13$258b2o$254bo3bobo$253b3o2bo$253bob2o5bo$254b3o2bo$254b3o3b
3o$254b2o4$257bo8bo$256b3o6b3o$249bo6bob2o4b2obo$248b3o6b3o4b3o$248bob
2o5b2o6b2o$249b3o$249b3o$249b2o3$225bo$223b2o$224b2o10$242b3o$241bo2bo
$244bo27b3o$244bo26bo2bo$241bobo30bo$270bo3bo$274bo$271bobo5$267b3o$
267bo2bo$267bo$267bo3bo$187bo79bo3bo$187bobo77bo$187b2o79bobo16$259b3o
$259bo2bo$259bo$259bo3bo$259bo$260bobo4$262b3o6b3o$150bobo13b3o93bo2bo
4bo2bo$150b2o102b3o5bo10bo$151bo102bo2bo4bo10bo$254bo8bobo4bobo$254bo
3bo$254bo$255bobo7$161b2o$157bo3bobo$156b3o2bo$156bob2o5bo$157b3o2bo$
157b3o3b3o$157b2o$249bo$248b3o$248bob2o27bo$160bo8bo79b3o26b3o$159b3o
6b3o78b2o27bob2o$152bo6bob2o4b2obo108b3o$151b3o6b3o4b3o109b3o$114bo36b
ob2o5b2o6b2o109b2o$113bo38b3o$113b3o36b3o$152b2o2$274bo$128bo144b3o$
126b2o144b2obo$127b2o143b3o$272b3o$272b3o$273b2o7$145b3o$144bo2bo$147b
o27b3o$147bo26bo2bo$144bobo30bo$173bo3bo$177bo$174bobo2$266bo$78bo186b
3o$76b2o186b2obo$77b2o91b3o91b3o$170bo2bo90b3o$170bo94b2o$170bo3bo$90b
o79bo3bo$90bobo77bo$90b2o79bobo95bo8bo$268b3o6b3o$261bo5b2obo6bob2o$
260b3o4b3o8b3o$259b2obo5b2o8b2o$82bo176b3o$81bobo175b3o$80bo2bo176b2o$
79bob2o$78bobo$77bo2bo$78b2o3$63bo$63bobo$63b2o97b3o$162bo2bo$162bo$
162bo3bo$40bo121bo$40bobo120bobo$40b2o212b3o$254bo2bo$254bo29b3o$165b
3o6b3o77bo29bo2bo$165bo2bo4bo2bo78bobo26bo$157b3o5bo10bo107bo3bo$157bo
2bo4bo10bo107bo$157bo8bobo4bobo109bobo$157bo3bo$157bo$27bo130bobo$26bo
bo$25bo2bo250b3o$24bob2o250bo2bo$23bobo255bo$22bo2bo251bo3bo$23b2o252b
o3bo$281bo$278bobo2$26bobo$26b2o$27bo2$152bo$151b3o$19bo131bob2o27bo$
18bobo131b3o26b3o$17bo2bo131b2o27bob2o$16bob2o162b3o$15bobo164b3o$14bo
2bo164b2o$15b2o2$271b3o$270bo2bo$177bo95bo$176b3o90bo3bo$175b2obo94bo$
175b3o92bobo$175b3o$175b3o$176b2o$274b3o6b3o$273bo2bo6bo2bo$266b3o7bo
6bo$265bo2bo7bo6bo$268bo4bobo8bobo$264bo3bo$268bo$265bobo8$169bo$168b
3o$167b2obo$167b3o$167b3o$168b2o$261bo$260b3o$259b2obo28bo$172bo8bo77b
3o28b3o$171b3o6b3o77b2o27b2obo$164bo5b2obo6bob2o105b3o$163b3o4b3o8b3o
105b3o$162b2obo5b2o8b2o107b2o$162b3o$162b3o$163b2o2$286bo$285b3o$285bo
b2o$286b3o$286b3o$286b3o$286b2o7$157b3o$157bo2bo$157bo29b3o$157bo29bo
2bo$158bobo26bo$187bo3bo$187bo$188bobo2$278bo$277b3o$277bob2o$182b3o
93b3o$181bo2bo93b3o$184bo93b2o$180bo3bo$180bo3bo$184bo$181bobo97bo8bo$
280b3o6b3o$273bo6bob2o4b2obo$272b3o6b3o4b3o$272bob2o5b2o6b2o$273b3o$
273b3o$273b2o9$174b3o$173bo2bo$176bo$172bo3bo$176bo$173bobo$266b3o$
265bo2bo$268bo27b3o$177b3o6b3o79bo26bo2bo$176bo2bo6bo2bo75bobo30bo$
169b3o7bo6bo107bo3bo$168bo2bo7bo6bo111bo$171bo4bobo8bobo105bobo$167bo
3bo$171bo$168bobo2$291b3o$291bo2bo$291bo$291bo3bo$291bo3bo$291bo$292bo
bo6$164bo$163b3o$162b2obo28bo$162b3o28b3o$163b2o27b2obo$192b3o$192b3o$
193b2o3$283b3o$283bo2bo$189bo93bo$188b3o92bo3bo$188bob2o91bo$189b3o92b
obo$189b3o$189b3o$189b2o$286b3o6b3o$286bo2bo4bo2bo$278b3o5bo10bo$278bo
2bo4bo10bo$278bo8bobo4bobo$278bo3bo$278bo$279bobo8$181bo$180b3o$180bob
2o$181b3o$181b3o$181b2o$273bo$272b3o$272bob2o27bo$184bo8bo79b3o26b3o$
183b3o6b3o78b2o27bob2o$176bo6bob2o4b2obo108b3o$175b3o6b3o4b3o109b3o$
175bob2o5b2o6b2o109b2o$176b3o$176b3o$176b2o2$298bo$297b3o$296b2obo$
296b3o$296b3o$296b3o$297b2o7$169b3o$168bo2bo$171bo27b3o$171bo26bo2bo$
168bobo30bo$197bo3bo$201bo$198bobo2$290bo$289b3o$288b2obo$194b3o91b3o$
194bo2bo90b3o$194bo94b2o$194bo3bo$194bo3bo$194bo$195bobo95bo8bo$292b3o
6b3o$285bo5b2obo6bob2o$284b3o4b3o8b3o$283b2obo5b2o8b2o$283b3o$283b3o$
284b2o9$186b3o$186bo2bo$186bo$186bo3bo$186bo$187bobo$278b3o$278bo2bo$
278bo29b3o$189b3o6b3o77bo29bo2bo$189bo2bo4bo2bo78bobo26bo$181b3o5bo10b
o107bo3bo$181bo2bo4bo10bo107bo$181bo8bobo4bobo109bobo$181bo3bo$181bo$
182bobo2$303b3o$302bo2bo$305bo$301bo3bo$301bo3bo$305bo$302bobo6$176bo$
175b3o$175bob2o27bo$176b3o26b3o$176b2o27bob2o$206b3o$206b3o$206b2o3$
295b3o$294bo2bo$201bo95bo$200b3o90bo3bo$199b2obo94bo$199b3o92bobo$199b
3o$199b3o$200b2o$298b3o6b3o$297bo2bo6bo2bo$290b3o7bo6bo$289bo2bo7bo6bo
$292bo4bobo8bobo$288bo3bo$292bo$289bobo8$193bo$192b3o$191b2obo$191b3o$
191b3o$192b2o$285bo$284b3o$283b2obo28bo$196bo8bo77b3o28b3o$195b3o6b3o
77b2o27b2obo$188bo5b2obo6bob2o105b3o$187b3o4b3o8b3o105b3o$186b2obo5b2o
8b2o107b2o$186b3o$186b3o$187b2o2$310bo$309b3o$309bob2o$310b3o$310b3o$
310b3o$310b2o7$181b3o$181bo2bo$181bo29b3o$181bo29bo2bo$182bobo26bo$
211bo3bo$211bo$212bobo2$302bo$301b3o$301bob2o$206b3o93b3o$205bo2bo93b
3o$208bo93b2o$204bo3bo$204bo3bo$208bo$205bobo97bo8bo$304b3o6b3o$297bo
6bob2o4b2obo$296b3o6b3o4b3o$296bob2o5b2o6b2o$297b3o$297b3o$297b2o9$
198b3o$197bo2bo$200bo$196bo3bo$200bo$197bobo4$201b3o6b3o$200bo2bo6bo2b
o$193b3o7bo6bo$192bo2bo7bo6bo$195bo4bobo8bobo$191bo3bo$195bo$192bobo
14$188bo$187b3o$186b2obo28bo$186b3o28b3o$187b2o27b2obo$216b3o$216b3o$
217b2o5$213bo$212b3o$212bob2o$213b3o$213b3o$213b3o$213b2o16$205bo$204b
3o$204bob2o$205b3o$205b3o$205b2o4$208bo8bo$207b3o6b3o$200bo6bob2o4b2ob
o$199b3o6b3o4b3o$199bob2o5b2o6b2o$200b3o$200b3o$200b2o!

So in the same spirit as the waterbear, a (6,3)c/123 spaceship could perhaps be built. The back end could be the same blinker trail as terminates the current HBKs. The period could still be adjusted if it is advantageous to do so, just by building a new helix.

I think it could be done.

[codeholic]

I thought about it too. But I had different geometry in mind. First of all, it might make sense to try a glider helix, because it's easier to synthesise. I'm not sure if it will work, though. Secondly, the only way to synthesise anything would probably be to collide LWSS and glider streams. Fortunately gliders can be emitted by half-bakeries, but LWSS should be synthesised in a way similar to how HBK generated NE gliders (by gliders colliding from opposite directions).

EDIT: Added a blueprint. Please don't take it too seriously, it is not well-thought-through. I haven't even checked if geometry is correct.

[codeholic]

chris_c, do you still have your script that you found a 3-glider collision for a NE glider with? We need a LWSS

P. S. Well, of course, one can build a LWSS seed with slow monochromatic salvo, but it would be nice to produce LWSS directly, if it's worth it.

[chris_c]

chris_c, do you still have your script that you found a 3-glider collision for a NE glider with? We need a LWSS

P. S. Well, of course, one can build a LWSS seed with slow monochromatic salvo, but it would be nice to produce LWSS directly, if it's worth it.

I think I just used gencols and then scanned for recipes that looked like they might sneak between the HB's manually. I certainly didn't know about Golly scripts back then. What exactly are you looking for here?

[codeholic]

Oh, I see. I'm looking for a LWSS recipe (I guess, we need a northbound of westbound one, if my geometry is correct), that can be made with least gliders coming from half-bakeries (or rather with least rows of half-bakeries, because SE gliders are 1,5 times more expensive than NW ones).

[codeholic]

Well, of course, one can build a LWSS seed with slow monochromatic salvo, but it would be nice to produce LWSS directly, if it's worth it.

Oh, no, I think you can't. There is too little space for slow salvo construction (the step is too tiny).

[codeholic]

dvgrn wrote a Golly script for finding such recipes, but for some reason it failed to find chris_c's reaction producing a NE glider with 3 gliders. I wonder, why.

[simsim314]

Oh, no, I think you can't. There is too little space for slow salvo construction (the step is too tiny).

Tiny step only means that your trigger can't be a single glider. You need to send 4-5 gliders (per HB "row"), the first one will trigger the glider mechanism, and the other 3-4 will just move the HB. Obviously all the 5 glider will need to be shoot from the helix in one "cycle".

[codeholic]

Then it means that you would need to build 3-4 more LWSS, otherwise the helix would be more sparse than the glider flow in the half-bakeries, right? I'm not sure if it is a good trade.

[simsim314]

Then it means that you would need to build 3-4 more LWSS

I really don't see the trade of. You will need a helix that moves (6N, 3N)/P instead of (6, 3)/P that's it. You will also need more complex fanout device of course. I really don't think that combining the (6, 3) helices is the best solution.

But first of all, we need to understand what is the minimal distance that allows slow salvo? Then we'll search for helix with that step.

[codeholic]

You'll need fanout devices anyway.

I guess, the upper limit for direct LWSS synthesis is 5 NW-SE gliders, 3 of them make a NE one, and then you just take a classic 3-glider synthesis. I haven't checked if it works, though. Anyway, in assumption that it works, you would need to produce only 2 more HB glider pairs with fanouts, compared to 3-4 times more (9-12).

[simsim314]

I was saying it's possible to use slow salvos with HB. But you probably right, using direct recipes like waterbear, without slow salvo at all, is probably better approach.

I'm not so sure why do we use slow salvo's in the first place, instead of simply colliding two orthogonal streams. Was it some sort proof of concept, or there is good reason for that?

[codeholic]

You don't have orthogonal streams in case of HBK. You've got only NE gliders. The best you could do is to collide *WSS with NE gliders, but that would need a *WSS emitter.

[dvgrn]

dvgrn wrote a Golly script for finding such recipes, but for some reason it failed to find chris_c's reaction producing a NE glider with 3 gliders. I wonder, why.

The reason was not very interesting. I picked what looked like a reasonable range for the search, but didn't take the time to find the exact edge of the range, and missed a good combination. The script could be fixed pretty trivially to find the 3-glider NE G, or maybe other useful items.

[biggiemac]

Well, with the HBs in constant motion, there should be a few more options that weren't accessible in the HBK. Is there a way to interact a SE glider with the spark to produce a NE glider? There's more clearance, and that would save 3 HBs on the current recipe.

[HartmutHolzwart]

Wouldn't it be time to revive this thread?

What is the current idea on the geometry of the caterpillar?

[dvgrn]

Apologies in advance -- this posting is not going to be about an HBK-Caterpillar design, just maybe a faster simpler HBK. However, some of the pieces, especially the glider edge-shooters, might come in handy in designing an HBK-Caterpillar as well.

Wouldn't it be time to revive this thread?

What is the current idea on the geometry of the caterpillar?

Seems as if there are quite a number of interesting options out there. When I look at the available mechanisms now, I'm wondering if there isn't an alternate design for (6,3)/{fairly large P}.

EDIT: Now I don't see how to make this work as described -- see footnote at end.

It would be more or less a simple triangle, very long and thin, bounded on one long side, let's say the southeast, by a double row of biloaves. The northwest side would be a slow-salvo series of intermediate targets, very similar to what can be found along the edges of a Centipede, or simsim314's upcoming engineless Caterpillar. Every now and then one of the targets would produce an edge-shooting 90-degree glider -- probably on the far side of the target, though it's just barely possible the near side could be made to work instead.

Let's assume it's the far side. The NE-traveling slow-salvo lanes can be either color and either parity, and they will generally creep northwestwards as the triangle widens, because the gliders from the wide end of the triangle have to have a clear path past all the intermediate targets from closer to the point of the triangle. The lanes don't have to strictly increase northwestwards, though, as long as edge-shooters can be found that have several lanes' worth of clearance.

The slow salvo will construct and trigger a seed for however many gliders are needed to run the two half-bakery lanes. If the width between slow-salvo targets is 36 cells, for example -- see the next post below -- then 24 gliders will be needed.

-- But only two of those gliders have to be precisely synchronized with each other! The others could be sent singly down one half-bakery lane or the other, perfectly safely as long as they're not too close to any other glider. In fact, the slow salvo could produce 22 gliders directly, one at a time, then build and trigger a seed for the last two.

If a width narrower than 36 turns out to support slow-salvo edge-shooter constructions, then the number of gliders needed would be proportionally less -- always two thirds of the width, with two gliders synchronized and the rest just having to be generated on the correct lanes.

EDIT: I don't think I had thought this design through carefully enough. First it occured to me that it clearly wouldn't be a (6,3)c/P spaceship exactly. With a slow-salvo width of 36 it would have to be a (72, 36)c/P ship, because each half-bakery would move twelve steps with each cycle.

That makes another obvious problem even more glaringly obvious: the line of slow-salvo targets to the northwest can't possibly be made to move (72,36) with every cycle of the ship. That's why the parallel lines of HBs were needed in the original design.

-- So I'm definitely out of practice thinking about this kind of thing. Next I'll try thinking harder about an actual HBK-Caterpillar design, and with any luck when somebody corrects the mistakes I make there, the result will be something interesting and workable.

[dvgrn]

But first of all, we need to understand what is the minimal distance that allows slow salvo? Then we'll search for helix with that step.

There's been very little investigation so far of what spacing will still allow for efficient edge-shooting glider recipes. The 31-lane spacing from the shield-bug/centipede spaceship project was certainly plenty. If we wanted to make it easy, we might go a little wider -- e.g., 36-cell spacing, allowing for very simple Herschel-based inserters:

Code: Select all

x = 210, y = 158, rule = LifeHistory
55.A$54.A.A$55.2A16$37.A$36.A.A$37.2A16$19.A$18.A.A$19.2A16$.A$A.A$.
2A3$3.2A$3.A.A$3.A30$71.2A$71.A.A$71.A30$139.2A$139.A.A$139.A30$207.
2A$207.A.A$207.A!

[Imagine each glider is from a separate cycle of the spaceship -- just one glider will hit one of the boats in each period P.]

I haven't figured out for sure whether it would be okay for a single glider strike to affect two adjacent targets. This would tend to happen a lot more as the width got down below, say, 20 cells.

It would certainly be okay sometimes. Let's say a glider strikes target N and produces new target N+1 -- but a spark from the conversion reaches over to the previous target N+1, and (say) removes a block. If you were going to have to shoot down that block anyway, then you've just saved a glider...!

Much more complicated interactions are possible -- a spark reaching forward to pre-change target N-1, or making large changes to target N+1 and not just a small deletion.

Reaching forward will often, but not always, result in paradoxes, where the part of the target that created the spark would no longer exist, or would no longer react in the same way. Maybe it's simplest to just disallow that possibility in searches. Really, though, it seems as if a search utility could test for paradoxes with no problem, and it might well significantly increase the search space, at least for very narrow widths.

[Sphenocorona]

I was thinking... maybe we don't need a raw receding helix or such (expensive), but instead we could bounce backward gliders off a retreating fuel rod of *WSSes or gliders. I made a crude design concept diagram, which is attached to this post. It kinda went off the side, but the idea is that we can have two sets of tracks that allow construction of spaceships in between them, where we can build streams of gliders or *WSSes that will pass through the front track, and then collide somewhere outside that salvo builder space in order to do useful work like create the fuel rod thing.

My concept drawing is pretty bad, but the idea is that we can probably do the same things in the ship as possible with a helix in at least half the cost, if not less.

Thoughts?

crude example that is probably wrong in some way.

HBKcaterpillarConcept1.png (6.66 KiB) Viewed 18292 times

[FWKnightship]

(6,3)c/1024 spaceship:

6,3c1024.mc.gz

(468.88 KiB) Downloaded 86 times

[dvgrn]

(6,3)c/1024 spaceship:
6,3c1024.mc.gz

I feel like this pattern needs a round of applause, or something bigger than "likes" anyway. It's a really impressive construction!

The .mc.gz file is 123938x112640, minimum bounding box 123938x112538, population minimum 346912 and maximum 356383 -- smaller than the original half-baked knightship but bigger than the Parallel HBK.

It's easy to reduce those numbers a little bit. If nothing else, several people on Discord have mentioned that the cleanup at the tail end can be trivially tightened up. See the attached version -- 110025x105406, population minimum 341572 and maximum 351082. I think I can see ways to get it down below 110000 width, but it's more work.

6,3c1024-smaller.mc.gz

110025x105406 corollary-snipe reduction of knightship

(458.89 KiB) Downloaded 34 times

@FWKnightship, do you know of any other places where there are obvious ways to tighten things up? I tried a couple of experiments. But I haven't worked out the math for the internal kickback mechanisms yet. Maybe they're adjustable, maybe they have a hard-coded size -- I'm not smart enough to know that on a first glance.

Compression and synthesis
Unfortunately post lengths don't quite support the attached knightship in a form that can be copied straight out of a code block and into Golly -- both macrocell and RLE format are about half again too large.

However, a Lua script that builds one of these high-speed caterpillar knightships should fit in a forum post really easily. Anyone need some Lua coding practice?

This also seems like a very easy spaceship (for its size) to build a glider synthesis for, or a gun like the old HBK gun. Maybe we should start a guessing thread for what the synthesis cost will come out to... guesses of "15" are perfectly true but will be studiously ignored.

[EvinZL]

Changed tail cleanup, minpop<=330,566

[biggiemac]

What a wonderful design! Absolutely nothing like I envisioned many years ago, but very understandable with a bit of experimentation.

Because the spacing of the glider "tracks" themselves encodes an entire recipe, and that recipe can be printed off as many times as is necessary by just instantiating another block of paired half bakeries, the dependency chain is clear. I wonder what the modular arithmetic constraints are, if there even were any. It looks like the same recipe works identically for the whole construction, but maybe the initial search was within a restricted space because of some constraint I am not immediately seeing.

Congrats on a great construction!

[HartmutHolzwart]

Does that mean that other periods would be possible with a similar design? If yes, what would be the minimum period? And how does the period affect the size?

How did you construct this? With helper scripts? Or rather manually?

[dvgrn]

Does that mean that other periods would be possible with a similar design? If yes, what would be the minimum period? And how does the period affect the size?

How did you construct this? With helper scripts? Or rather manually?

I'm very interested in FWKnightship's answer to this.

I can say, based on tracing through the inner workings and writing up the HBK caterpillar article, that it looks like period 1024 was probably chosen just to be HashLife-friendly -- there's nothing particularly magical about 1024 otherwise.

However, a fairly nightmarish number of internal glider stream crossings make it a highly non-trivial exercise to adjust this for most other periods, at least in the absence of a very clever helper script with a complexity similar to simsim314's Caterloopillar builder script.

I was thinking for a moment that if it just so happened that all of the nested rectangles had an even number of signals traversing them, then maybe I could just remove every other signal and end up with a period-2048 knightpillar. But even if the even-number-of-signals thing happens to be true, it's not that simple! If you take half of the 20G activation flotillas out of the main spine, the main spine will end up having to be adjusted to a slightly different angle -- and I'm not sure yet whether that change might cause collisions at the various internal glider stream crossings.

Glider synthesis ideas so far
Just as a side note, there are some similar tricky issues involved with making a glider synthesis of this spaceship, even without changing the period. The main idea for an "easy" synthesis is to settle the spine down into a "quiescent" state, simply by letting it run for a long time without feeding it any new 20G activation flotillas, until the gliders are all cleared out. That will change the angle of the spine by the maximum amount.

Theoretically we could then feed in 20G activation flotillas at any period we might want, to start the spine up again. The basic mechanism is completely period-agnostic. It's "just" a matter of re-connecting up the twenty nested rectangles, such that each one has a total signal path length that's a multiple of the target period. But if some of the inner rectangles have to get a little bigger, that's likely to have a ripple effect and require a change to all of the outer rectangles... Definitely an automated builder script would be really helpful here!

Anyway, for synthesizing the existing p1024, there turn out to be some minor problems with shutting down the spine. The spine itself reaches a quiescent state with no problems at all -- but then there are catastrophic explosions at the south corners of the rectangles.

The problem is basically that there's a different number of signals traveling through each of the rectangles, so the shutdown can't possibly happen cleanly -- the activation flotilla that produces the first glider in a NEward glider salvo is not the same as the one that produces the second and third gliders, so there end up being some one- and two-glider partial salvos flying around and doing unexpected things.

So the shutdown needs a whole lot of babysitting (and/or cleverness that I don't have available yet) ... to produce a quiescent-state spine and targets in all the right locations so that it can be started right back up again -- just by sending a whole lot of 20G flotillas at p1024 back in to the top of the spine. Once a quiescent-state knightpillar pattern is ready to go, I think there will just be some relatively small problems with supplying various missing gliders with "support gliders" coming in from the northeast, southwest and southeast, until all twenty rectangles are completely re-filled again.

Those input lanes can all easily be left clear of obstructions until it's time to build the half-blockades or blocks to get those corners back to normal as the signals from the re-activated spine start coming in. P1024 is plenty slow enough that there shouldn't be any trouble building a half-blockade after the last support glider has gone by, let alone building a single block.

EvinZL's Tail Reduction
This is just another temporary corollary snipe -- it looks like the tail can be shortened by another 100 cells or so, just by moving the left half further toward the main spine.

hbk_reduced_a_bit_more.mc

(1.43 MiB) Downloaded 21 times

Now minimum population = 329071 at T = 590. Min box = {107061, 105334} at T = 281:

hbk_reduced_a_bit_more_still.mc

(1.43 MiB) Downloaded 21 times

That was just to prove that the southeast edge of the trapezoidal tail could be moved northwest. It can be moved quite a bit farther, actually, but it will be necessary to tighten up the spacing between the half-bakeries in the tail to their minimum distances ... and there might be a better trick to be found that will make all of the kind of arbitrary-looking complexity in the tail unnecessary.

Of course the biggest reduction will happen if it turns out that the nested rectangles in the main body can be compacted. It looks to me like they probably can, but I haven't attempted to prove that yet!