phil
@big_algocracy
Followers
74
Following
48
Media
6
Statuses
120
optimizing for impact and fun | @stanfordnlp
Palo Alto, CA
Joined August 2024
this is unintentionally funny and sad
was going through some MoE stuff for my mini project & I assume this professor is Percy Liang all I wanna say say is I owe you big time for your clear & intuitive explanation prof
0
0
0
Spotted on the streets of San Diego. The rarest cup of them all.
2
2
43
@brianryhuang would not have phrased it this way after your viral post
3
3
175
i say it every year ... i'll never understand neurips party waitlists bc truly my worst nightmare to be stuck on a boat with ai people
1
2
38
anyone who's talked to a stanford undergrad from the past 5 years knows that the accommodation grift is 90% because stanford borked its housing system and upperclassmen started getting OAEs to guarantee themselves a single, which led to a race to the bottom where you either had
Academic “accommodations” have rapidly, insanely morphed from accessibility to arbitrage. The US has built a system where ordinary stress, distraction, or a B+ in college gets medicalized, and then also buys a student extra time and softer rules. This idiocy rewards
39
71
2K
🃏 Poker night at #NeurIPS2025 – Dec 2, San Diego → Talk RL, agents, reward modeling, RLHF → Play poker (decision-making under uncertainty) → Maybe get your RL book signed by @RichardSSutton 👀 For researchers working on agent eval, preference learning, AI safety,
luma.com
A small, off-the-record poker night for RL people exclusively! Expect sharp conversation, fine eats, and excellent company. Hosted by Judgment…
1
2
3
The rhetorical style of mech interp is like nothing else. This blog (about how mech interp doesn't really work) has 45 footnotes, an appendix, a FAQ, and uses 5 levels of bulleted indentation.
mech interp is surely a field in Kuhnian crisis
5
6
206
least surprising event of 2025
Wow, ChatGPT is already showing ads? I was just talking with it about Elon on Nikhil’s podcast when out of nowhere it popped up an ad saying, “Find a fitness class, Connect Peloton.” 🤯 Wild. At least match the ad to the topic next time!
0
0
0
This might be too niche, but pre-llm researchers : newer researchers :: Frieren : Fern (LLMs are Zoltraak)
10
2
44
MATS in shambles. the world is healing
I've been saying mechanistic interpretability is misguided from the start. Glad people are coming around many years later.
0
0
2
common debate in labs atm: incremental value ($/day) of continued contribution vs. benefit of finding a nice place to ride out post-agi
1
0
1
frontier tower makerspace people ily you give me hope @frontiertower
0
0
0
I’ll be at #NeurIPS2025 this week. If you were Reviewer 2 on my paper, please come find me. I just want to talk. I promise.
37
35
772
At NeurIPS next week. Interested in post training: long horizon, better search, agent formalisms, non-scalar rewards. https://t.co/p0ipLJK194
calendly.com
Chatting about research.
17
10
284
Hosting a brunch at NeurIPS next weekend. DM me if you’re around!
9
5
79
✈️ I'll be at Neurips Tuesday - Sunday next week! Dms are open, if you're interested in chatting about multi-step RL, async pipelines, coding/computer use, continual learning, gdm, hmu for coffee
7
2
76
1/ Hiring PhD students at CMU SCS (LTI/MLD) for Fall 2026 (Deadline 12/10) 🎓 I work on open, reliable LMs: augmented LMs & agents (RAG, tool use, deep research), safety (hallucinations, copyright), and AI for science, code & multilinguality & open to bold new ideas! FAQ in 🧵
17
124
612