see you soon!

🫢

what is this? i found this in my screenshots?

45mins....

👀 gorgeous?

usage maxxing

For everyone interested in lilac.codes, or have questions. Feel free to join discord.gg/aaFBUrCHm.

Our discord server.

the sun's rising baby.

claude please 😭

Think of a blender. Your PIN goes in and comes out as a smoothie. The bank only keeps the smoothie, never your PIN. When you type it again, they blend it and check whether the two smoothies match. You can't turn a smoothie back into fruit, so even the bank never sees your PIN!

@Shra_va_nierying `SELECT 1 FROM users WHERE username = ?` on every keystroke. at 500M users that would melt the primary database.

production systems use a 3-tier check:

1. Bloom filter at the edge:
a probabilistic bit array held in memory. 500M usernames with a 1% false positive rate requires only ~600MB of RAM.
- if the Bloom filter returns FALSE, the username is 100% available. zero downstream lookups needed.
- if it returns TRUE, it might be a false positive (~1%), so it falls through to:

2. Distributed in-memory cache (Redis cluster):
sharded by `hash(LOWER(username))`. a single in-memory `EXISTS` check resolves in sub-millisecond latency.

3. DB Unique constraint as the final safety net:
client keystrokes only check availability. when the user actually submits registration, an explicit `UNIQUE (LOWER(username))` constraint handles the in-flight race condition if two users claim the same available handle at the exact same millisecond.

sighs. 911, emergency backup required.

nah you sleeping on couch tdy

claude >> anthropic

Seeing generational levels of locked-in on the streets of SF

(face redacted to protect the innocent 🫡)

that's why i chose other AI and not claude, i ain't losing money anthropic.

welp...