Active region construction thread

[russellsprouts]

Previous threads/posts:

Optimized snarkmaker
New construction arms (Simkin gun construction)

Starting a thread to discuss "Active region construction".

Basically, constructing objects with a continuous stream of single-channel gliders hitting a reaction, without necessarily waiting for the reaction to settle before you continue.

My script https://github.com/RussellSprouts/short ... n/snark.py has various commands for finding this kind of reaction.

The script works like this:

1. Define intermediate stages of the target pattern. This is usually the slow salvo construction of the object
2. Run a search. The search will process single channel streams of gliders hitting the initial target, and save results which settle to a p2 ash with nothing escaping. It doesn't save the resulting pattern, but it saves a bunch of statistics about the ash.
   • Dimensions -- far depth, near depth, lane width, population, etc.
   • If the ash contains a partial match for one of the intermediate stages, the best partial match, and its statistics (digest of matching components, digest of extra components in the area where the target should be, separation between the target ash and other ash, rarity of the matched components vs missing components, etc.)
   • Best full match for an intermediate stage and its statistics
3. Use SQL queries to sort and filter the results to find good reactions that make progress
4. Start another search starting with the best candidate streams from the previous step as starting points.

The snark has mostly rare components, so I mostly used the slow salvo recipe to construct it by finding streams that happen to send the correct zero degree gliders. The simkin gun is very simple, so I've been using a "SLS printing" technique to build it directly. The seed for a simkin glider gun consists of:

• An eater 1 to cap one of the barrels
• Three blocks to form a B60 conduit
• Two blocks of the other B60 conduit
• A block, blinker, and beehive seed which forms the herschel and last B60 block when hit with a glider.

As long as I place the eater 1 first (thus giving me more possible depths to find it at), I can build the entire simkin gun by directly materializing the components in the right place, back to front. Even then, however, some closely spaced objects take 1-2 days of searching to find.

To combine the "send zero degree gliders following the slow salvo recipe" and "build the components one-by-one" approaches, I think we should make a database, some kind of hybrid between a synthesis component database and the octo databases. It would include n common objects that can be materialized directly + a polyplet of some size. We would save any results that affect some of the input objects and settle to a stable state.

Then, if we can't find a direct way to materialize an object, we would have the option of materializing a different object, then finding the polyplet to convert it.

• Are there statistics for the prevalence of specific polyplets?
• Is there a database like this?

[hotdogPi]

Are you aware of https://conwaylife.com/forums/viewtopic ... 656#p38773 (census of small evolving patterns)? There are the octo3g (3 gliders) and octo3obj (glider hitting 3 objects) databases.

[russellsprouts]

Are you aware of https://conwaylife.com/forums/viewtopic ... 656#p38773 (census of small evolving patterns)? There are the octo3g (3 gliders) and octo3obj (glider hitting 3 objects) databases.

Those are useful, but I would also be interested in sparks and patterns that don't reach a population of 20.

Perhaps the most relevant would be to run random streams colliding with a block, then collect statistics about all the polyplets that appear in the evolution of those patterns after the last glider hits.