RT @leothecurious: > self dox.> check my real talk on youtube.> clicks link.> speaker is actually the lich king.> knew_it.png https://t.co/….

Turns out 2d lights have a height in Godot, now is much better.

"A medieval knight in full armor riding a motorcycle through a misty jungle trail.". Try Grok Imagine, free for a limited time.

Thanks to @EveryoneIsGross fantastic job with normals we have 3d light!

Big progress I do not have the fatal cut-off anymore. Here are fifty enemies at a quite stable framerate, I do still have a few drops from the 144cap though.

Thought

Managed to keep 144fps cap with 40 enemies each running their own state machine. More than that I currently drop to ~100fps at 45 enemies and then down to 50 quickly. I am progressively moving things to precomputed tables but I do not have anymore big refactor energy.

Line 666:Nested typed collections are not supported.

Tavern Brawl

Now that the enemies are finally able to encircle the main character I can go back to posting about machine learning.

I am very satisfied. A few parameters to tune but it is starting to look good.

with three busy slots paths naturally create circumnavigation

Tactical Vectors

looking cuter with a higher dimensional tessellation

Using Dijkstra to compute navigation paths on the polar grid that can be served to the enemies with a single computation on this low dimensional graph instead of an a* runner for each enemy on the full navmesh.

Took a break from modelling the enemy AI it was converging to my day to day job. Went back to clicking and learned about auto-tiling, mix and matching 4 terrains.

Progressive de-spaghettification of the enemy ai.

Took a bunch to implement the sensor efficiently without fps drops, I was spending too much time trying to validate the polygons shape. I guess accepting some errors is better sometimes.