Arjun Guha
guha-anderson.com
Arjun Guha
@guha-anderson.com
In the early 1990s, my grandfather purchased an original IBM PC with an aftermarket HDD. I first learned to program using that PC and reading the IBM/Microsoft Basic manual.

Using the Spark for the last 2 days reminds me of that experience.
October 17, 2025 at 1:28 PM
At Gimme! Coffee -- my favorite New York State coffee chain!
August 30, 2025 at 2:40 PM
1850’s baseball.
June 22, 2025 at 6:37 PM
The recent *Your Brain on ChatGPT* paper seems cool.

When a ugrad approaches me to do research, I still have them read a prefix of PLAI (1st ed.) and demonstrate that they understand it.

I wonder what would happen if I asked them to self-study with an LLM exclusively. Has anyone tried this?
June 19, 2025 at 6:11 PM
Now that I've written more Python than I care to admit, I'm getting tired of duplicating abstractions: one for sync code and another for async. Effect polymorphism wanted.
May 21, 2025 at 4:06 PM
Looking forward to this.
April 15, 2025 at 7:13 PM
Reposted by Arjun Guha
In this episode of HUB History, Elena Palladino discusses the creation of the Quabbin Reservoir, the four towns that were sacrificed for its construction, and her book Lost Towns of the Swift River Valley. Listen now!

www.hubhistory.com/episodes/wat...
Water for Boston, part 3 - Lost Towns of the Swift River Valley: Drowned by the Quabbin, with Elena Palladino (episode 322) - HUB History: Boston history podcast
This week, we’re speaking with Elena Palladino, the author of the recent book Lost Towns of the Swift River Valley: Drowned by the Quabbin.  This book outlines the 20th century development of Boston’s...
www.hubhistory.com
March 10, 2025 at 1:55 PM
There seems to be a fundamental misunderstanding here. I don't think PhD students complete assigned tasks.

arstechnica.com/ai/2025/03/w...
What does “PhD-level” AI mean? OpenAI’s rumored $20,000 agent plan explained.
Silicon Valley may value imperfect virtual PhDs more than universities pay real ones.
arstechnica.com
March 8, 2025 at 1:47 PM
Photo taken today at @browncsdept.bsky.social. I'm glad to see that the PhD students (@genevievemp.bsky.social), furniture, and faculty seem to have not changed in 10+ years.
March 5, 2025 at 8:06 PM
Reposted by Arjun Guha
The real lesson from DeepSeek is the importance of good old-fashioned computer science. Every day this week, they've been doing open source releases. The latest is their in-house distributed file system. github.com/deepseek-ai/...
GitHub - deepseek-ai/3FS: A high-performance distributed file system designed to address the challenges of AI training and inference workloads.
A high-performance distributed file system designed to address the challenges of AI training and inference workloads. - GitHub - deepseek-ai/3FS: A high-performance distributed file system design...
github.com
February 28, 2025 at 10:07 AM
Reposted by Arjun Guha
Please help amplify ARBOR, a fantastic new research opportunity! If you’d like to start contributing, NDIF is now hosting DeepSeek R1 8B and 70B, open for all researchers to experiment on via our API.

Sign up for API access here: login.ndif.us
February 20, 2025 at 10:35 PM
O1, R1, etc. are so good that we evaluate them on “PhD-level” benchmarks. But, these benchmarks are so hard that most people can’t even understand what they are testing. We’ve built a benchmark with problems that are hard to solve but easy to verify: for both humans and models.
February 4, 2025 at 2:37 AM
Last one: there are a LOT of people to blame for this one. I think @jasvir.bsky.social is to blame for this problem in "Humanity's Last Exam".
January 28, 2025 at 6:29 PM
Ugh, who did this? @joepolitz.bsky.social ? Wait, was it @dbp.bsky.social ? Someone else from @shriram.bsky.social's group?

Also from "Humanity's Last Exam".
January 28, 2025 at 3:51 PM
OK, who is responsible for this? Is it @natefoster.bsky.social?

Source: "Humanity's Last Exam" www.nytimes.com/2025/01/23/t...
January 28, 2025 at 3:47 PM
It was fun to be a part of this. We analyze a dataset of student-LLM interactions on programming tasks and ask: what do students get wrong when prompting?
excited to say that our Substance Beats Style paper was accepted to NAACL! We investigate *why* student-written programming prompts don't work well for LLMs, and find that while students think it's because of technical vocabulary gaps, it's actually information content that matters
Substance Beats Style: Why Beginning Students Fail to Code with LLMs
Although LLMs are increasing the productivity of professional programmers, existing work shows that beginners struggle to prompt LLMs to solve text-to-code tasks. Why is this the case? This paper expl...
arxiv.org
January 24, 2025 at 11:15 PM
Counterintuitively, this shows the power of GC. IIRC I've heard of similar tricks used for high-performance OCaml at Jane Street.
January 11, 2025 at 2:26 AM
Reposted by Arjun Guha
Last week we had the last-ever lecture by Steven P. Reiss, extraordinary creator of extraordinary programming environments. As a tribute, a bunch of faculty and staff showed up and sang this song. Here's the video:
www.youtube.com/watch?v=N4Pu...
December 4, 2024 at 1:29 AM
I found this surprising and had to test it myself (works with clang -std=c99):

lemon.rip/w/c99-vla-tr...
C99 doesn't need function bodies, or 'VLAs are Turing complete'
lemon.rip
November 25, 2024 at 1:08 AM
Reposted by Arjun Guha
since more people are joining bluesky, time for a re-intro!

i work on programming languages and compilers for hardware design and will be starting at MIT next year where i will lead the Foundations of Languages and Machines Lab (flame.csail.mit.edu).

if the ideas excites you, reach out to me!
FLAME Lab @ MIT
flame.csail.mit.edu
November 19, 2024 at 5:44 PM