Carla Griffiths

Are ML researchers constantly sleep deprived because they're trying to submit all their jobs before they sleep

@Yunsu_Tang @sacredrain Stop @Yunsu_Tang !!

@sacredrain @carlagriffs lol

50 shades of JPMorgan: Chirayu Rana and Lorna Hajdini.

dont teach a man how to prompt. teach him how to yearn for the vast and endless latent space

what’s the best thing to do on a Monday bank holiday in London?

@somewheresy It’s still useful in some verticalized AI cases but agree most tid bits / facts just require a simple grep tool call to search

is anyone else always lol at the fact that we spent YEARS trying to figure out hybrid search, document embeddings, "RAG" and graph databases, just for the models to improve enough to wield tools against a filesystem, and the SOTA achieved by Doing Literally None Of That

I know I’ve noticed it in the last couple of years as well, reviewers see other author’s submissions as direct competition to theirs. In my icml submission the reviewers requested lengthy additional experiments outside of the scope of the paper.. (I spent $$ in extra gpu hours) only to say in the response that the main paper results still aren’t valid. It’s a huge time and environmental waste imo

@carlagriffs it's happening quite often, I was a bit confused then what is the point of rebuttals if these issues still persist.

ML conference timeline: 1) submit the paper you have been working on for several months. 2) reviewers need additional experiments and clarifications. 3) as the primary author you run all of those experiments and report those findings. 4) reviewers are satisfied, don't increase the scores and ignore you. 5) PC says those experiments cleared all clarifications but please add those to updated paper ( you can't revise during rebuttals) and submit to another venue.

@rishdotuk Nice need to look this up although I’m fatigued with everything related to LLM as a judge

@carlagriffs Wasn’t there a conference getting organised for papers written by LLM and evaluated by LLMs. Can’t recall if the organisers were from Harvard or Stanford.

i was thinking last night - just screen your conference submissions through chatgpt (not claude!) at this point because 80% guarantee the reviews, if not also meta review will just be entirely LLM-generated

@demisama_ brutal, sorry

all positive scores still got rejected by #ICML2026 😢

I sometimes forget how weird the dropout paper is.

Harvey is valued at $11B. Legora just raised at $5.5B. I built their entire web application in two weeks and I'm making it open-source and free for everyone to use. Say hi to Mike: mikeoss.com. When I got the chance to try Harvey and Legora, I was surprised by how simple they were. A thought came to mind: I could probably build something similar in no time at all with Claude. And so I did. Assistant, project, tabular review and workflows. You get it all without vendor lock-in. Mike offers law firms an alternative, where they own the application layer and aren't stuck with a vendor they're renewing forever. You can try Mike in the demo on the website, or go to the GitHub link on the site to download the code and run a local version yourself.

@dscape @f @goncalossilva @davidgomes 🙌

Running a AI night in London on June 18! Reply if you are keen to join + @f @goncalossilva @davidgomes

@rishdotuk Thanks for the appreciation was hoping someone would notice that

@carlagriffs “My Clanker in Christ”

Getting annoyed that claude code (opus 4.7) keeps on trying to dissuade me from shipping a full-stack app in a day, saying 'it's not feasible', etc...my Clanker in Christ that is the whole reason I'm paying $100/month for Max..you're here to help me ship, not to make judgments on my work/life balance

@carlagriffs @Londonmaxxing Thanks for coming Carla, I don't mind the photos 🤠 Everyone can sign-up to londonmaxxing.com for early news on what's next, and to join the the movement 💪

AI research used to be in a unique position. Unlike biology or materials science, it was not extremely resource-intensive: 1-2 GPUs could still support top-venue-level work. Unlike math or theoretical physics, it was also not purely intelligence-intensive. (1/3)

tech tennis w/ my fav person @carlagriffs dm if you wanna join fast track if you’ve got as much rizz as her

@Yunsu_Tang Another diabolical caption 😭

@Romy_Holland agreed, one of my friends took a whole MONTH of leave to grind leetcode before his first google interview..if engineers collectively redirected their time grinding leetcode into something for the public good i bet we would have found the cure for a lot of cancers by now

there are apparently applicants to google who have done 4000 leet code practice questions. even ppl who do far less practice still spend months practicing. imagine what kind of projects those people could have done in that same amount of time. google is hijacking the attention of a bunch of the smartest engineers and getting them to waste countless hours on puzzles that have no real world utility at all. they should just administer an IQ test and call it a day.