iiinova — where your digital self keeps evolving

motherfluffer @latticepolytope.bsky.social · 11/09/2026

1200 pulls is an absolutely mind-boggling number! IME, 600-ish pulls is usually enough for P6, but bad luck is a thing. Wishing you good luck when Plume alter comes!

001

motherfluffer @latticepolytope.bsky.social · 10/09/2026

The next logical step is to save Royal Tokens. You don't know what Plume alter's class will be, so you need to save tokens of all operator classes!

Tokens for Kal'tsit, Orchid and Angelina.

100

motherfluffer @latticepolytope.bsky.social · 10/09/2026

If X wants to redefine "bot" as "someone who doesn't use their mobile app", then fine, I'm a "bot".

000

motherfluffer @latticepolytope.bsky.social · 10/09/2026

But that is exactly the problem! Telling gambling addicts to take a break from the casino!

110

motherfluffer @latticepolytope.bsky.social · 10/09/2026

In math, people have long complained about articles whose bibliography contains “private communication” items. I see OpenAI decided to do something about it by not citing the relevant private communications!

000

motherfluffer @latticepolytope.bsky.social · 25/07/2026

Securing the upcoming non-collab limited banners.

000

motherfluffer @latticepolytope.bsky.social · 16/07/2026

Average luck. 571 pulls, 1 Ch'en spark.

010

motherfluffer @latticepolytope.bsky.social · 16/07/2026

It's 6 minutes before #Arknights daily reset. The event hasn't started yet, and therefore I haven't bought the event packs yet.

020

motherfluffer @latticepolytope.bsky.social · 13/07/2026

Lo mismo pasó con Jessica the Liberated en Arknights original.

110

motherfluffer @latticepolytope.bsky.social · 13/07/2026

Of course, you might run into equivalence relations (e.g., two red-black trees are equivalent iff they contain the same elements), but those can be handled just fine with set-theoretic technology. No sheaves needed.

000

motherfluffer @latticepolytope.bsky.social · 13/07/2026

In ordinary computer programming, unless you willingly overcomplicate things (laziness, higher-order functions, dynamic dispatch in general), everything in sight is discrete anyway, so there are no interesting gluing problems.

100

motherfluffer @latticepolytope.bsky.social · 13/07/2026

And it's this inability to work with anything but overlapping pieces that creates the interesting problems that sheaves help solve!

100

motherfluffer @latticepolytope.bsky.social · 13/07/2026

By contrast, when you want to define a smooth function (or vector field, differential form, etc.) on a manifold, you really have to check smoothness on a coordinate chart. Unless your manifold is itself an open subset of R^n, these charts unavoidably overlap.

100

motherfluffer @latticepolytope.bsky.social · 13/07/2026

The issue isn't even whether you “can” handle overlapping covers, but whether you need it in the first place. You can define a function on Foo+Bar+Qux by gluing two functions on Foo+Bar and Bar+Qux that agree on Bar. But why would you? It's so much easier to define the second function on just Qux.

100

motherfluffer @latticepolytope.bsky.social · 13/07/2026

Of course, I should've said “overlapping pieces” rather than “overlapping opens”, because there are are sites that don't come from topological spaces.

100

motherfluffer @latticepolytope.bsky.social · 13/07/2026

If you dropped the condition that the U_j's are open, then the theorem no longer holds. You can easily check it by taking the U_j's to be all the singleton subsets of X. Then the theorem would assert that any arbitrary function X -> Y is continuous, which is obviously false in general.

000

motherfluffer @latticepolytope.bsky.social · 13/07/2026

Let X and Y be topological spaces, let {U_j} be an open cover of X, and let f_j : U_j -> Y be continuous functions such that every f_i and f_j agree when restricted to U_i \cap U_j. Then the unique function f : X -> Y that agrees with every f_j on the respective U_j is continuous too.

100

motherfluffer @latticepolytope.bsky.social · 13/07/2026

Sheaves (and Grothendieck topologies, descent, etc.) actually buy you something when your local sections *must* be defined on overlapping opens and you have no way to avoid it. They're a vast generalization of the following theorem from point-set topology.

200

motherfluffer @latticepolytope.bsky.social · 13/07/2026

A cover of a disconnected space by disjoint opens is trivial, because there's no gluing condition to check: 1. Any two functions Unit -> T “glue” to define a function Bool -> T. 2. Any two functions A -> T and B -> T “glue” to define a function Either A B -> T. And so on.

100

motherfluffer @latticepolytope.bsky.social · 11/07/2026

Can you define a meaningful, nontrivial example of a sheaf in your code? Say, the Serre twisting sheaf on a projective space.

100

motherfluffer @latticepolytope.bsky.social · 04/07/2026

All #Arknights limited operators fully maximized: P6, L90, M9, at least one mod3 whenever possible. At least until Wang.

020

motherfluffer @latticepolytope.bsky.social · 25/02/2026

It will be gone on monthly reset, but let's enjoy it while it lasts.

000

motherfluffer @latticepolytope.bsky.social · 26/01/2026

Avdotya is incredibly pretty, no matter what the background is.

110

motherfluffer @latticepolytope.bsky.social · 20/01/2026

Goldfish can't resist the call of We Many?

000

motherfluffer @latticepolytope.bsky.social · 19/01/2026

Super secret Higashinese technique for training yourself without external distractions.

100

motherfluffer @latticepolytope.bsky.social · 16/01/2026

As usual, the welfare is the hardest to max pot.

010

motherfluffer @latticepolytope.bsky.social · 13/01/2026

I really don't know myself. A tryhard friend who does max risk clears under restrictions asked for it. My guess is he already computed timings using a pot5 Ines and doesn't want to recompute them with a pot6 one. Ines with S3 gets redeployed a lot.

010

motherfluffer @latticepolytope.bsky.social · 12/01/2026

I'm looking for an #Arknights friend with a *pot5* L90 #Ines with S3M3 and mod3 for #ContingencyContract #CCB4. Pot6 won't do, I need pot5 specifically.

112

motherfluffer @latticepolytope.bsky.social · 06/01/2026

I'm very tempted to not buy those Polyester Lumps. Most useless material in existence, even worse than Sugar Lumps.

000

motherfluffer @latticepolytope.bsky.social · 01/01/2026

Arbitrary JavaScript, or a subset with the least egregious parts of it? Can it detect when an object is meant to be used as a struct vs. as a hash table?

000

motherfluffer @latticepolytope.bsky.social · 01/01/2026

Standard ML programmer here talking shit about both. (Joking, of course.)

000

motherfluffer @latticepolytope.bsky.social · 01/01/2026

Rust has too much ergonomics fluff, but its foundation is leaner than Swift's. E.g., Rust doesn't have classes. Want shared ownership? Use Rc or Arc. It's a more orthogonal design. The only thing Rust really lacks is ML-like functors, but Swift doesn't have them either.

010