๊ณ ์ ๋ ํธ์
d0nut ๐ฆ
16K posts
d0nut ๐ฆ
@d0nutptr
Moved to: https://t.co/db4RFbvJtu
San Francisco, CA ๊ฐ์
์ผ Kasฤฑm 2013
460 ํ๋ก์20.9K ํ๋ก์
@sshell_ Maybe if i see some cool people at events this weekend...
English
if you're looking to work in the intersection of AI and offensive security, come work with us!
runsybil.com/careers
English
Incredibly proud to be a part of this team. I'm consistently blown away by the creativity, talent, drive, and humanity of everyone I work with.
The product is already finding mind-blowing vulnerabilities and delivering world-class outcomes, and it's something I truly believe in
Sybil@runsybil
The way we hack is changing and we're building what comes next We've raised a total of $40M to create the AI-native platform for offensive security
English
just saw six people tag me with a potential NA live incident affecting League, appreciate the tags, sharing all of that internally.
English
@rez0__ I see, I keep hearing that the cli is the way to go. Maybe I just trolled myself.
Thanks for the tips ๐ซก
English
@d0nutptr I have exclusively used Claude code via cli and via my own discord integration (just using the cli under the hood) so I donโt think I know about either of those i guess
English
So far very unimpressed with claude code's performance via the VSCode plugin.
1. Slow (_wayyyyyyy slow_)
2. Multiple misses in a row when it tried its plan out
Might have to stick with the web UI for now. Disappointing.
English
@rez0__ I'm still absolutely blown away at the sheer difference in performance between the vscode extension and the web client. My use-case is pretty hard on the model I think, but I was still surprised at how much worse the plugin was :/
English
@d0nutptr are you not on max? if youre not, i think they use a different data center or something that is MUCHHH slower
English
@jrozner Yea I was disappointed in the route it took in planning. I expected it to grab more of the correct context relevant to answering the question. When I did that step manually in the web UI, I got an answer in 10 seconds. I waited 10+ minutes for the vscode extension to get it wrong
English
@d0nutptr Iโve been using the cli and have had similar experiences. Often more work trying to coerce it to do the right thing than just doing it myself. Lots of waiting around
English
Prompted on web _once_ and it came up with a solution faster and was correct on the very first try.
Annoying.
d0nut ๐ฆ@d0nutptr
So far very unimpressed with claude code's performance via the VSCode plugin. 1. Slow (_wayyyyyyy slow_) 2. Multiple misses in a row when it tried its plan out Might have to stick with the web UI for now. Disappointing.
English
@drewlevin @spamofthemic As a result, you get "fun" suggestions like "let's ban players above a certain rank for not grouping with their team".
I think the wide variety of winning strategies is what makes the game fun, even if we all show up on the rift with a different understanding of the game.
English
@drewlevin @spamofthemic It certainly doesn't help the discussion that people don't seem to consider that one's rank distills multiple factors (performance at distinct skills, playstyles, interpretations of the game, etc) into a single metric.
English
asked the analytics team to look into the most basic version of this, which is "if only one team had 1+ chat transgressions in a game, which team won?"
for all ranked games in 2025, you had a 22% winrate if your team had a chat transgression and their team didn't.
Drew Levin is on vacation (see pinned post)@drewlevin
A thought I had this morning: I encounter a decent amount of people trying to justify their negative chat behavior by saying that Riot doesn't sufficiently penalize soft inting, with the implication that their own use of chat is their only real recourse to getting inted.
English
@drewlevin @spamofthemic Makes sense. Anecdotally I've certainly seen negative comms come out well before the game state deteriorated, and it has been my experience that coordination between the negative party and everyone else drops a fair bit.
(Not that I'm an angel myself; I've been negative too)
English
@d0nutptr @spamofthemic it's gonna be both/and, I think there is some question over how much of the effect size is gamestate -> comms vs comms -> gamestate, but it's real hard to find any 80/20 splits on single actions
English
@drewlevin @spamofthemic for the record, I expect negative comms => lower winrate, especially in lower elos, but I can't help but shake the Bayesian interpretation of this analysis P(lose game | negative comms)
English
@drewlevin @spamofthemic Ah, I was going to ask if it was possible to look at the subset of games where, say, the gold difference was below some amount when the negative comms happened. While an imperfect filter, it might help narrow in on the association between negativity and performance
English
@drewlevin I'm okay with this in normals because if a game isn't going well, and everyone is in agreement, we should maximize fun by getting into the next match faster. This does "minimize joy" for the other team, but negative feelings are more impactful than positive ones, sadly.
English
@drewlevin Voted "good / like to ff" for normals, "ehh not great" for ranked
I would expect a portion of skilled players to have their rank suppressed by successful ffing. Would encourage more "if I don't win lane, I won't win game" play, coinciding with in more stomps.
English
question I am curious about (follow-up poll in the reply is even spicier): if we matchmade games of league NORMS where - keeping skill and role preferences as they are - we tried to put people who share FF preferences on the same team, that would be:
English