Conway's Life: Work in Progress

Miscellaneous topics in Conway's Game of Life -- unfinished projects of all kinds and conditions

10 March 2018

It Happened One Knight

On 6 March 2018 the first member of a new class of Conway's Life spaceships was discovered. This is Sir Robin, the first elementary spaceship that travels in an oblique direction. Its displacement is two cells horizontally and one cell vertically (or vice versa) every six generations, which is the fastest possible knightship speed. The name is a reference to Monty Python's "Brave Sir Robin", who bravely runs away as fast as possible.

Code: Select all
#C (2,1)c/6 knightship found by Adam P. Goucher,
#C based on a front end originally found by Josh Ball,
#C rediscovered and extended by Tomas Rokicki,
#C using a SAT solver-based search
x = 79, y = 31, rule = B3/S23

The new knightship was found by Adam P. Goucher based on initial results by Tom Rokicki, after about a month of automated searching. The program that completed the knightship was icpx, a "multithreaded hybrid of LLS and gfind".

A detailed summary of the discovery process is now available.

15 October 2016

14-, 15-, and 16-bit still life syntheses

A week or so ago, a better recipe was found for the last still life on Mark Niemiec's list of expensive 14-bit syntheses. Now all 14-bit still lifes can be constructed with less than 14 gliders -- less than 1 glider per bit, as the old saying goes.

Catagolue results continue to be very useful in finding new recipes.

Code: Select all
#C 12-glider synthesis for the last 14-bit still life,
#C snake bridge snake / 12.105, which had previously cost at least
#C one glider per bit.
#C Goldtiger997, 6 October 2016, optimized by Mark Niemiec on 7 October.
x = 79, y = 71, rule = LifeHistory

UPDATE: The next challenge along these lines was to similarly reduce 15-bit still life costs to below 1 glider per bit. The process started later in the same forum thread, and was completed on November 19, 2016, with the following 14-glider synthesis:

Code: Select all
#C 14-glider synthesis for the last 15-bit still life
#C which had previously cost at least one glider per bit.
#C Extrementhusiast, 19 November 2016
x = 48, y = 38, rule = B3/S23

UPDATE 2: The next project involved a similar reduction for 16-bit still life recipes. The official project kickoff was on December 16, 2016, when 443 of the 3,286 16-bit still lifes had no synthesis in less than 16 gliders in Chris Cain's database. It concluded successfully on May 24, 2017.

12 June 2016

New spaceship speed: 3c/7

Tim Coe has found a symmetrical spaceship with a new speed, 3c/7 (left, below) after a series of searches that took a total of "one or two months". At 29 cells wide, it is the minimum width odd symmetric spaceship -- an exhaustive width 27 search was run some time ago by Paul Tooke. The author seems to have officially chosen a name of "Spaghetti Monster" for the new 3c/7 spaceship.

Matthias Merzenich has pointed out that two of these spaceships can support a known 3c/7 wave (right, below).

Code: Select all
#C 3c/7 FSM spaceship: Tim Coe, 11 June 2016
#C Period-28 3c/7 wave found by Stephen Silver on Feb. 2, 2000
x = 187, y = 139, rule = B3/S23

This is the twenty-second spaceship velocity constructed in Conway's Life -- counting each of the four infinite families of spaceships (Gemini, HBK, Demonoid, Caterloopillar) as one velocity each.

10 March 2016

New c/10 "copperhead" spaceship

Reposted with permission from Alexey Nigin's blog:

The day before yesterday (March 6, 2016) forums saw a new member named zdr. When we the lifenthusiasts meet a newcomer, we expect to see things like “brand new” 30-cell 700-gen methuselah and then have to explain why it is not notable. However, what zdr showed us made our jaws drop.

It was a 28-cell c/10 orthogonal spaceship:

An animated image of the spaceship

To explain why this is such a groundbreaking discovery, I should first tell you that Life spaceships can be loosely divided into two categories. Engineered ships are the ones that consist of various small components. They often have adjustable speed. However, the population of tens of thousands to millions of cells causes these spaceships to have no practical value.

There is much more incentive in hunting for elementary spaceships, which can be used for complex constructions. They are found using programs such as gfind or WLS. The algorithms behind these programs are beyond the scope of my article, but the important thing is that the search time goes up exponentially as the period of the ship grows. It is most interesting to find spaceships of new speeds, and the number of speeds that low-period ships can have is unfortunately limited:

Orthogonal Diagonal
c/2 Yes Impossible
c/3 Yes Impossible
c/4 Yes Yes
c/5 Yes Yes
2c/5 Yes Impossible
c/6 Yes Yes
c/7 Yes Yes
2c/7 Yes Impossible
3c/7 No Impossible
c/8 No No
3c/8 No Impossible
This table does not include oblique speeds, which causes little inconvenience because no elementary oblique ships are known.

The table above shows that ships exist for most of possible speeds, and it seems obvious that the speeds for which there are no ships have been searched by numerous people with good knowledge of search programs, powerful computers and lots of patience. As for higher periods, even the smallest searches would take years on modern computers. It appears that low-hanging fruit have been harvested clean during the 46 years of Life research… or, more precisely, it appeared so before zdr’s post.

The idea we all missed is that if the ship is really microscopic, it can be found in reasonable time despite its high period. After zdr boldly went where no man has gone before, Josh Ball set up the corresponding search in gfind and refound the spaceship in a little over an hour. zdr said that their program found it in a matter of 19 seconds.

To be frank, similar event did happen before when the aforementioned Josh Ball pulled loafer out of a hat. However, zdr’s spaceship (which is now called Copperhead, as proposed by muzik) is much more awesome for a number of reasons:
  • Loafer is not so mind-bogglingly high-period.
  • Copperhead was much easier to find, so it is more surprising that nobody found it before.
  • Copperhead’s tail is relatively strong and can interact with other objects without breaking down.

The discovery of a new spaceship speed immediately opened a few new areas of research, which are being explored now.


Aidan F. Pierce came up with a Copperhead synthesis only 10 hours after the completion of the spaceship. The synthesis was inefficient, and a few people discovered better ones. The current best synthesis, made by Chris Cain, requires only 22 gliders. Its repeat time is 375 ticks, which means that a gun can start constructing the second spaceship 375 ticks after the first one. There is a 23-glider synthesis with a better repeat time of 373 ticks.

Incremental 22-glider synthesis of the copperhead

The synthesis can be substantially improved if we find this spaceship crawling out of a random soup. Adam P. Goucher has written a wonderful program called apgsearch, which is perfectly suited for this task. While the current version may be too slow to find a soup in reasonable time, highly anticipated version 3.0 can probably do the trick. Once it is found, it will appear here.


Once the synthesis was complete, building a gun was nothing but corollary-sniping. The first copperhead gun was created by myself, and a video of it is available here. It was put together in a hurry and is therefore extremely inefficient. In particular, skilled gun builders can spot a silly mistake in the Northeast.

gmc_nxtman then made another gun with an almost optimal period of 376 ticks.


simeks found two eaters for this ship, the better of which is shown below:

A copperhead eater

It is now time to search for a good copperhead-to-something-useful converter. The only existing one is clumsy and slow.


Sawtooths often work by sending a flotilla of fast ships towards a slower ship. The more is the difference in speed, the less is the expansion factor of a sawtooth. Since expansion factor is proportional to how boring the sawtooth is, increasing the speed difference is a good thing. Dean Hickerson collided c/2 standard spaceships with c/10 copperhead to get a sawtooth with expansion factor 6:

Hickerson's sawtooth

He then made another sawtooth with expansion factor 10/3.

Puffers and rakes

Suppose a c/10 flotilla is hit by a glider. The glider turns into loads of mess, but all copperheads somehow survive and move on. The mess releases a glider, which flies into strategically placed second flotilla that is identical to the first one. Gliders continue to bounce back and forth between flotillas leaving mess behind them, and a c/10 puffer is complete!

Unfortunately, this cool technique doesn’t work out easily in our case. There are no interesting interactions between a glider and a single copperhead, and it is unclear how one can place two or more copperheads so close to each other that a glider interacts with all of them. Assuming we figure it out, we can try to make a rake by perturbing the mess with copperheads so that it evolves into gliders, but that seems even less likely.

However, all this hand-waving can be turned it real puffers if we find…


Tagalongs are small things that are attached to the back of a spaceship and move with it. Here is an example tagalong, called the Schick engine:

Schick engine pulled by two LWSSs

Finding a tagalong for the copperhead (or two copperheads) will be really nice. We can also try searching for pushalongs, but they are generally rarer.

Other patterns

There are a few other areas of Life exploration where the copperhead can be useful. For example, universal constructors often need to create an elbow still life very far away. It can be done by producing a copperhead, waiting for some time, and then shooting the copperhead down with a LWSS. At the moment I do not see why the copperhead can be better than the loafer in this aspect, but who knows?

17 June 2014

New Wrinkles in the Slow-Salvo Construction Game

Slow-salvo constructions are starting to gain some traction as more uses are found for them. Most recently, a (6,3) knightship has been built in Conway's Life using a small synchronized glider salvo that activates long chains of half-bakeries, which cooperate to create a slow salvo that then re-creates the synchronized gliders at the correct offset. (See item #17 below.)

sample slow salvo
Slow salvo from 31c/240 spaceship project
-- click the image to get the pattern RLE.

Quick Review

A "slow salvo" is a series of gliders all traveling the same direction. A "slow salvo construction" involves aiming a carefully designed slow salvo at a target object, and incrementally converting that target into some new object using a series of controlled collisions. Such a slow salvo is often described as a "recipe" for the new object.

Some recipes may recreate the original target in a new location. Slow salvos can be proven to be universal, so other recipes might produce output gliders at 90 degrees from the original salvo, or larger spaceships, or any possible glider-constructible object.

In a slow salvo, arbitrarily long delays can be added between any two gliders. In other words, slow-salvo glider collisions must always settle into stable ash before the next glider arrives. In this context, "stable ash" might mean actual period-1 still lifes, but in "P2 slow salvos", low-period oscillators such as blinkers, beacons, or toads are also allowed as intermediate targets.

In the P2 case, of course, gliders can only be delayed by an even number of ticks. P3, P4, P5, P6, etc. slow salvos are technically possible, with restrictions analogous to the P2 case, but they don't seem to offer any practical advantages over P2 slow salvos.

New Developments

Here's a list of known uses for slow-salvo technology, in rough chronological order:

1) Conway's Universality Proof for B3/S23
Unidirectional salvos were a significant component of John Conway's proof of construction universality in Life. The original proof was completed shortly after Gosper's glider gun was discovered (in 1970). Only an outline of the proof was ever published, many years later, in Volume 2 of Winning Ways for Your Mathematical Plays (1982). Part of the idea was to aim one or more "flotillas" -- long series of gliders all traveling in the same direction -- at a faraway simple object, or a faraway opposing flotilla produced by "double side-tracking". The glider collisions would gradually convert the target object into a complex piece of period-30N circuitry capable of performing calculations... including, eventually, the calculations needed to construct copies of the original flotillas.

These were unidirectional salvos, but not necessarily slow salvos. Some of Conway's construction recipes required multiple nearby gliders to be precisely synchronized with each other.

2) Schroeppel's Speculations on Sparse Life Universe Fate
By 1992 Rich Schroeppel and others had worked out the likely long-term fate of an infinite Life universe in which each cell had a very low but nonzero probability of being ON initially. After a few ticks, such a universe would consist predominantly of blocks and blinkers, since these only require three initial ON cells near each other to persist. Gliders would be very much rarer since they need five coordinated ON cells to get started -- but whenever a rare one did appear, it would travel until it eventually collided with something.

A glider hitting a blinker, or a block that has already been struck by another glider, can cause a chain reaction that releases multiple additional gliders. Schroeppel showed that arbitrarily long unidirectional salvos would eventually be created by chains of these kinds of collisions. Any glider could follow any other glider on nearby lanes, with few or no limitations -- except that most sets of closely-spaced gliders were much less likely.

Following Conway's universal-construction ideas, Schroeppel conjectured that a replicator would eventually appear that consisted of a long unidirectional flotilla of gliders aimed at a single faraway block. Here again this flotilla may have been visualized as containing some synchronized gliders, as long as they could be produced by gliders interacting with common random ash objects.

However, the key term "slow-unidirectional-salvo" appeared more than once in 1992. Also, Schroeppel's replicator blueprint used gliders of only a single color, analogous to the squares a bishop can reach on a chessboard. It appears that these monochromatic gliders were known or assumed to comprise a universal construction toolkit; this has recently (quite possibly not for the first time) been proved by construction. The concept of "slow synthesis" was introduced by Alan Wechsler and discussed in depth by Mark Niemiec in 1996.

3) Nick Gotts and Sparse Life
Nick Gotts introduced the term "slow salvo" into more general use in 1997, leaving "unidirectional" unstated (since that's implied by the definition of "salvo" in any case). Building on Schroeppel's earlier investigations, Gotts showed the feasibility of using pure slow salvos to construct arbitrarily complex objects, including rare infinitely growing objects. One early result was a 53-glider slow-salvo recipe for a block-laying switch engine. The primary research focus was on the emergence of increasing levels of complexity starting from random initial conditions, as opposed to universal construction or replicator design.

The appearance of actual construction recipes made it clearer that a complete universal toolkit could be built up from slow salvos alone, aimed at some very simple object -- a block, or pretty much any small common ash object. Such objects could be placed at arbitrary distances by various "pusher" reactions, or by collisions between fast and slow glider-constructible spaceships. (Various c/12 Corderships were known by 1997.) This avoided a lot of worries about the slow and complicated side-tracking guns that had been part of earlier universality arguments.

4) The Prototype Universal Constructor
In 2004 Paul Chapman built a prototype Conway's Life universal constructor. It was designed to read recipe data from a static tape and convert it into a slow salvo. Theoretically it was capable of being programmed to construct anything glider-constructible, up to and including itself. It did not include any mechanism for duplicating construction data to a new copy of itself, however, so it was not a candidate replicator.

The prototype was programmed to slow-construct a single eater. A small modification allowed the pattern to loop indefinitely, constructing an oblique line of eaters. But it was apparently too difficult to program -- no other slow-salvo recipes have ever been generated for it in its original form (!).

5) Spartan Universal Computer-Constructor
In 2009 Adam P. Goucher incorporated Chapman's prototype universal constructor into a single programmable pattern that was both a universal constructor and a universal computer. Here again, universal construction ability was theoretically provided by a single construction arm... but in practice, the UCC was never programmed to construct even as much as a line of eaters.

6) O(sqrt(log t)) Diametric Growth Pattern
In April 2010, just one month B.G. (Before Gemini), Adam Goucher also completed a pattern that grows at the slowest possible rate for a Conway's Life pattern -- or any other 2D Euclidean CA rule, for that matter. Of course it's possible to slow it down further by adding various delay mechanisms, but the new version would still have an O(sqrt(log t)) growth rate. Anything slower than that is provably impossible. can be downloaded from within Golly 2.6 via Help > Online Archives > Very Large Patterns.

This rather mysterious and under-documented pattern uses two slow salvos, 28 gliders and 10 gliders respectively, to read and write bits in an unbounded triangular grid. Basically it counts in binary, using two-dimensional arrangements of boats on the Life plane separated by 16 full diagonals. The pattern uses copies of various pieces of the prototype constructor. The slow salvos were constructed manually using the original P1 block-move table, so they are very far from optimal! A tighter bit spacing could now be found with many fewer gliders.

Honeyfarm targets for the salvos are created by a secondary arm that produces single gliders perpendicular to the primary slow-salvo shotgun direction. This initial collision is very similar to the construction method used by the (completely independently designed) Gemini spaceship -- see below.

7) The Gemini Spaceship
In May 2010 Andrew Wade's self-constructing Gemini spaceship was published, again based on Chapman's 2004 prototype. The Gemini broke new ground in many areas, including the use of two construction arms firing coordinated 90-degree slow salvos to build Herschel-based signal circuitry. The Gemini's construction gliders are paired so that each glider in one slow salvo is synchronized with a glider in the other salvo. But as usual, successive gliders coming from the same direction don't have to be synchronized with each other -- only with their opposite number in the other salvo.

In most cases, construction by two perpendicular slow salvos is clearly much more efficient than construction with a single slow salvo. On the other hand, it takes extra circuitry to process the information for two separate slow salvos, and nontrivial signal-crossing problems can also show up in many cases. Single-salvo construction tends to minimize the complexity of the circuitry required for a universal constructor, at the cost of increasing the amount of construction data -- see #22.

8) Universal Construction with Intermittent P30 Glider Streams
Between 2009 and 2011, Frank Hoetmer put together a universal construction toolkit using collisions between slow salvos of *WSSes and gliders. The *WSSes and gliders were produced from 180-degree collisions of long streams of gliders. The two colliding streams were on fixed lanes, and were reflected with p30 technology, so the only variation available was the presence or absence of gliders in each p30 stream. Surprisingly, this turns out to be enough to allow construction universality, including the construction of new precisely-timed p30 gun/reflector components far away from the originals.

Like the Gemini spaceship, this was an independent project that showed up "out of the blue" and broke a lot of new ground. Intermittent P30 construction is impressively difficult -- among other things, very large recipes are needed, and child replicator components have to be constructed in exactly the right phase relative to the parent. As a result, this line of research was not continued once Geminoid technology became available.

9) Pianola Breeders
In 2010, Paul Tooke simplified and extended the Gemini's construction mechanism, producing a number of slow-salvo-constructed patterns with superlinear growth, including a Gemini-based slide-gun puffer and a series of Pianola breeders, which moved the Gemini's two construction arms to a permanent stationary platform, using fourteen glider-loop channels instead of twelve.

10) Serizawa Geminoids and "Armless" Universal Constructors
The extra complexity of multiple-arm circuitry has meant that since the original Gemini appeared, no two-arm constructor designs have been completed using Conway's Life rules. However, a self-constructing two-armed pattern has recently been completed in Serizawa, another CA rule. Design features from this pattern can be applied to many other rules; among other things, it made possible a new "armless" universal constructor design in Conway's Life that set new records for small size and population.

11) Linear Replicators in Conway's Life
The linear Life replicator, or "GoL propagator", demonstrated the universality of single-arm construction -- a single slow salvo aimed at simple targets can build anything that two arms can build. It is conjectured that either one or two slow construction arms can build anything that can be constructed with any number of colliding gliders (and/or spaceships). A proof is now within easy reach.

12) Freeze-Dried Seeds for Slow Salvos
With linear-replicator technology as a model, "freeze-dried" slow salvos -- a variant of Nick Gotts' chain reactions, but with the seed still lifes deliberately built by a universal constructor -- can be designed that allow the construction of spaceships with very small step sizes, such as (1,0) or (1,1) -- or (2,1), a true (very slow) knightship. As of January 2017 no explicit example has been completed.

13) Spiral Growth
A spiral-growth pattern was constructed using four copies of a single-arm universal constructor. More recently, a spiral-growth pattern smaller by a factor of 100 was completed, followed by one that's 5000 times smaller than that. The last pattern in the series uses an inline, 0-degree "Snarkmaker" recipe to build small lossless elbows using single-channel technology. See #23 below.

14) Slow-Salvo Construction of Loafers
In September 2013 universal-constructor based loafer seed factory was constructed, using an early version of the two-channel Demonoid (#22 below) construction arm design. This was along the same lines as Paul Tooke's Pianola breeders (#9), which were adapted from the Gemini's twelve channels.

15) 31c/240 Spaceships
In 2013 a new reburnable fuse was discovered that produced two gliders every 240 ticks, while the reaction moved forward 31 cells and the fuse moved forward 9 cells. 9 and 31 are relatively prime, and it turned out that the fuse could be modified in various ways to produce any possible slow salvo.

A new search utility has been written to locate the most efficient recipes for lightweight, middleweight and heavyweight spaceships, which can move fast enough to overtake the front of the fuse and close the loop by building more blocks to support the 31c/240 "Herschel climber" reaction -- very much along the same lines as the original Caterpillar spaceship, but with a new mechanism and new speed.

16) Caterpillar's Smaller Siblings
It appears that the original Caterpillar can now be made significantly smaller using slow-salvo technology. For example, the Caterpillar builds a large number of HWSSes that travel forward to build new blinker trails ahead of its 17c/45 pi climbers. Each HWSS recipe requires a rake about 8000 cells high, made up of two separate sets of forerakes and backrakes running on blinker trails. The rakes' gliders collide a long distance from the trails, forming the large oblique triangles that make up most of the Caterpillar's body.

Spaceship constructions that are done entirely with a slow salvo seem likely to need rakes that are only a few hundred cells high instead of several thousand, and that can be packed very closely one after another. Even when the construction site is 1500 cells from the rakes, with a slow salvo there's no need to wait thousands of ticks between rakes to allow perpendicular glider streams to converge.

17) Half-Baked Knightships
Slow-salvo technology has now been used to construct a (6,3) knightship based on interactions between chains of half-bakery reactions. The new feature of this design is that slow-salvo gliders are generated on only one color -- every other lane -- of the Life grid. As mentioned in #2 above, this is not a new idea, but the half-baked knightships are the first completed patterns to use this type of slow salvo for actual constructions.

18) Parallel HBK gun
A gun was created for the parallel version of the half-baked knightship. To date (July 2017) this gun is the only other use of the original Gemini spaceship's slow-glider-pair construction technique, besides Paul Tooke's Pianola breeders and (in a very limited sense) Adam P. Goucher's O(sqrt(log t)) pattern.

19) High-period 2c/7 "distaff" rake
In May 2014, building on 2c/7 puffer results from the previous December, Ivan Fomichev used a long chain of weekender conduits to generate gliders for a modified slow-salvo construction of a single LWSS ("modified" because it starts from two targets produced by glider-weekender collisions). The LWSS is then sent forward to the beginning of the chain to close the cycle and produce an adjustable high-period rake.

20) 'Engineless Caterpillar' (Caterloopillar) project
Starting in September 2014, Hartmut Holzwart proposed a slow-salvo-based version of David Bell's idea for an "engineless Caterpillar", from a decade ago. Michael Simkin has since worked out the necessary mechanisms for an adjustable "strange-loop" Caterpillar-like spaceship. an adjustable-period Caterpillar design that doesn't need a period-specific "climber" reaction. Instead, streams of gliders are generated by repeatable interactions between stable objects and backward-traveling fleets of passing *WSSes. These streams of gliders can be used to perform slow-salvo constructions of fleets of spaceships traveling the opposite direction, which collide with other still lifes at the front to build more backward-traveling fleets, thus closing the cycle. To allow the period to be as adjustable as possible, only P1 slow-salvo recipes can be used. If a blinker, toad or other P2 object were allowed as an intermediate target, the adjustability would be limited to either odd periods or even periods.

The current model can be fairly easily recalibrated to move at any sufficiently slow speed. For example, it can move at c/10, c/11, c/12, or any slower c/{integer}. The price of this adjustability is that only stable intermediate targets can be used; otherwise, very different construction recipes would be needed for even periods vs. odd periods. Slow construction of edge-shooter spaceship seeds, or any other stable seed object, appears to still be perfectly possible with this restriction. In fact, some preliminary searches seemed to show that P1 slow monochromatic gliders (all gliders the same color) would be sufficient, so an engineless caterpillar with an even step length would be technically possible -- e.g., moving 38 instead of 59 cells with each cycle. However, such a spaceship would be considerably longer due to less efficient slow salvo recipes.

21) (23,5)c/79 Caterpillar
Also starting in September 2014, building on discussion from the previous year, Brett Berger ('biggiemac' on the forums) steadily generated mechanisms and techniques for constructing the first oblique caterpillar knightship, based on a known (23,5)c/79 Herschel reaction. The spaceship was completed on December 28th of the same year. Constructions are a mix of slow-salvo transformations and synchronized collisions.

Unlike the engineless Caterpillar which is designed so that the period can be either odd or even, this spaceship's rate of travel is fixed and slow salvo gliders will always be the same distance apart. This means that p2 intermediate targets can be used. For example, there's a line of traffic lights near the top of the triangle in the current front-end design; every other traffic light is in a different phase, but the glider that collides with it is always in the correct phase to convert the traffic light to a beehive to support the oncoming Herschel climber.

22) Demonoid (diagonal Geminoid) spaceships
Starting in late 2012, a number of designs were proposed for a self-constructing Geminoid variant where the two halves of the spaceship would be glide-reflected mirror images of each other, instead of exact copies. This would limit the travel direction to an exact diagonal, but would cut the amount of circuitry roughly in half (because it would no longer be necessary to support input from two different directions).

Over the next three years, the amount of required circuitry was gradually decreased until a complete Demonoid needed only 48 still lifes to be built with a slow salvo, including all the self-destruct circuitry to clean up after each construction arm when its work is done.

As usual, this self-constructing spaceship has a trivial glider synthesis -- so trivial that for the first time in Life history, a gun for the 0hd Demonoid was built before the spaceship existed, and the very first 0hd Demonoid was fired by the gun.

In June 2017, a Demonoid using a single-channel recipe (see below) was constructed with only 19 still lifes per construction arm, and even more compact Demonoids were shown to be possible.

23) Single-channel construction arms Research by Simon Ekström in December 2015 proved that a universal construction arm was possible with a stream of gliders on just a single lane, with gliders separated by at least 90 ticks. Unlike in the 0hd Demonoid where gliders could follow each other as closely as 15 ticks apart on the construction lane, a single-channel recipe can be safely reflected by a Snark reflector, or even a syringe followed by signal-splitting Herschel circuitry.

This was a really radical simplification. Any spare glider output in any Herschel circuit with 90-tick recovery time could now be set up to function as a construction arm, with the addition of a single block anywhere along the output lane.

24) Pan-directional construction elbows An idea that began its development in the 10hd and 0hd Demonoids has been almost fully realized in recent single-channel libraries of elbow operations: recipes are now known that can send a glider or spaceship directly from the construction elbow in any of eight directions. The first use was in a Snark-chain-constructing triple wickstretcher. Two of these directions are used in the single-channel Demonoid. First, as in the wickstretcher, 0-degree gliders are send ahead of the construction elbow to build a Snark reflector exactly on the construction lane, as a lossless elbow. This allows the construction arm to reach around two corners instead of one, to build the next copy of the constructor/reflector. Second, backward LWSSes and MWSSes are fired from the same construction arm after the temporary Snark is removed, to destroy the previous constructor/reflector and complete the spaceship.

25) Orthogonoid spaceship
In June 2017 the first self-constructing spaceship with a slow orthogonal direction of travel was completed, using streams of MWSSes to store construction data instead of streams of gliders. It makes use of a temporary lossless elbow in its construction arm. Removing the Snark elbow allows the construction arm to function inexpensively as a destruction arm, at 90 degrees from its original direction of operation.

26) Hydra
In November 2017, Oscar Cunningham completed (+/-2sL) a single-channel self-constructing pattern called "Hydra", due to its ability to grow new "heads" at the corners of an expanding spiral pattern. Hydra is fairly similar to the 2017 spiral-growth pattern, but instead of adding an inline lossless elbow (a Snark) at every construction cycle, the Hydra adds a lossless glider splitter at the end of every active signal branch, creating more and more signal branches over time.

27) Quadratic-Growth Replicators
It appears to be possible to use slow-salvo constructions, especially with single-channel construction arms, as part of a quadratic-growth replicator in Conway's Life, small enough that Golly can successfully run it through multiple cycles. A diamond-shaped design looks likely to be workable, but at present only fairly early design stages are complete.

Another work in progress is Adam Goucher's 0E0P megacell replicator, which will achieve quadratic growth by emulating an appropriate Life-like cellular automaton. The "ground state" for this very large unit cell is empty space, meaning that each set of megacells must create new neighboring copies of their entire structure and construction-recipe data, at every location where a birth is required.

26 November 2013

New Technology from the Replicator Project

Now that the Conway's Life replicator pattern is in working order, what might the next step be?

The phase-shifted linear replicator isn't really a very satisfactory design. Each parent pattern can produce only one child pattern, which then blocks it from any further replication. It seems as if a quadratic-growth, space-filling replicator would be much more in keeping with von Neumann's (and Conway's) original purpose.

One major limitation of essentially linear designs like the Gemini spaceship and Geminoid replicator is that replication and movement perpendicular to the long stream of gliders is fairly easy, but it's very hard to make a new copy in the other direction -- just because it means constructing the far end of the new copy millions of cells away.

It's certainly not impossible to reach that far out into empty space with a constructor arm, but it's bound to be very slow -- either in terms of the absolute number of ticks, or in the amount of time that it takes to simulate a construction cycle. Even Golly's Hashlife algorithm has difficulty with very long streams of information-carrying gliders traveling next to each other in opposite directions -- the number of combinations goes up exponentially, and beyond a certain point no reasonable amount of memory can hold all the different hashtiles.

Luckily it may be possible to solve both of these problems at once with a diamond-shaped replicator. The memory loop would travel around the outside of a hollow square.

Hand blocks (or elbow blocks, if the UCs at the corners have two arms) can be trivially constructed in the correct starting locations by colliding LWSSes from one corner with gliders from an adjacent corner; glider pairs or slow salvos following the first glider will immediately have a target to work with.

To give Hashlife as much help as possible, it will make sense to adjust the reflection timings at the four corners so that the memory loop takes 2^N ticks per cycle; the spatial periodicity should also be a power of two.

The result will be a space-filling Life replicator with the same quadratic growth rate as Langton's Loops. It will be interesting to see how much memory will be needed to allow Golly to "run away" with the replicator simulation.

In the diagram at right, blue diamonds represent glider memory loops containing construction information. (A closed loop may not actually be needed, but that's another story.) Green objects are universal constructors and reflectors. The yellow arrows are eater groups that can absorb child replicators' attempts to build on top of a quiescent parent replicator. The white lines show the paths of starter LWSSes and construction gliders. The red numbers show replicators' generation number.

The New Technology

A number of the new construction and destruction mechanisms from the Geminoid project may be useful here: