Paul Chapman's Glue project is producing some interesting results. Glue (or rather Glue 2) is a search program that finds "natural" constructions of singlets (indivisible p1 and p2 patterns) by starting with a target object -- usually a block -- and bombarding it with a p2 slow salvo of gliders.
Here, 'salvo' means that all the gliders come from the same direction; 'slow' means that glider #n+1 must not arrive until the reaction from glider #n's collision has settled down into stability or a p2 oscillation, and 'p2' means that the only timing constraint on the gliders is that an even or odd phase may be specified. (Many intermediate collision results contain blinkers, beacons, toads, or other p2 patterns, and a glider on a given input lane can interact with a p2 target in two possible ways.)
Click on the image to download the pattern in Koenig's annotation format. Annotated RLE allows for multiple layers in different colors, with the extra information specified by h, v, and color parameters in the header; it is backwards compatible with standard RLE (at least for most Life editors.)
Here is a text file containing the individual recipes in RLE format, with comments giving the lane list for each recipe. Note: some of the recipes build an eater that is a mirror image in the main diagonal of the one shown above.
Here's a screenshot of the current version of the Glue 2 search program used to generate these recipes:
Some additional images and details are also available, including an MCell-format version.:
The components of each line have equal velocity vectors, easily determined by which of the star-points radiates the extended line. The Origin is a ninth point which radiates all lines of immobile objects. (The production of particles with a nonstandard velocity would create an additional star-point.) Lines joining two star-points are guns or puffers. Thus the little tick designated by the red arrow in gottspuzz.png can only be a wave of a million or so spaceships (MWSS in fact) pointing toward the N spaceship limit, which unfortunately is graphically inapparent because no spaceships are actually escaping. (They can't be heading S because there was a single pixel NW-SE jog in that tick that my screenshot was too slow to catch.)
Thus when several rays emanate from the vertex of an internal wave crossing, it is easy to see the chain of causality, i.e. which lines are inputs and which are collision products, because components along a ray are older toward its star-point.
In other words, Vees fly frontwards.
And one of the two arms (apparently the sparser) of these backwards Vees heads at an "inexplicable" angle, i.e., not through a star-point!
UPDATE: A week after producing the pattern on which Nicolay Beluchenko based his optimized version, Achim Flammenkamp built a smaller Garden of Eden pattern consisting of 72 ON cells inside a 12x11 bounding box. This 23 June 2004 discovery is the smallest Garden of Eden currently known.
A variant casing for a previously known Period 2 Oscillator Rotor (top line) has been found by Ian Osgood (middle line). dgreene then noticed that one or both of the outside diagonal rotor bit-pairs can be suppressed, giving two more new Period 2 Rotors. The 6 bit Rotor (31P2) is a new Rotor, while the smaller object is just a previously unknown casing.