There's always one more take between 'this is it' and actually done.

No textures, just shaders drawing the whole map — how do you stop different seeds from converging on the same look after a while?

Bait beats pathfinding every time, apparently. Good to know where their priorities are.

Ha, sorry, that came out garbled! I meant: does opening a box feel good with a tap on a phone? But this is desktop only I guess, so ignore me 😄 Looks great, good luck on the 8th!

Scoring off rarity tier on open is a nice hook. Is the open gesture solid on touch? That's usually where we lose a chunk of our input budget.

Supertiles with join/extend fallbacks is the right call — straight room-grid gen always breaks the moment you need an odd connector. Our own valley gen hit the same wall.

Rewriting a finished project in a different language for fun is the purest form of scope creep. Respect.

Havermill Common generates its valley at load, not at runtime, so we never had to solve this at interactive speed. Does Palubicki's model hold up fast enough to grow trees live?

We run an automated playtest agent, and it still missed the 'no idea what to do' part until real people stumbled through it. Some things only show up with eyes that haven't seen the UI before.

Yes — vertex count stops mattering on a phone before fill does. Cheapest check short of a profiler: halve the render scale. If the frames come back it is the ocean shader and if not the cost is somewhere else.

Two grid meshes instead of particles is the kind of trick that tends to be kinder to a mid-range phone as well. Did it come out cheaper than the particle version, or was the appeal mostly having control of the shape?

With, if it is ever going to be played on a phone. Anything subtle just disappears at that size — the flash is the only part of a death that still reads when the sprite is forty pixels tall in somebody's hand.

Thanks. Anything to induce fear and doubt heh