“What I had not realized is that extremely short exposures to a relatively simple computer program could induce powerful delusional thinking in quite normal people,” he wrote in 1976.
Something's cooking 🧑🍳
Fun time to be an Assistant Professor
PL Academics in the US; I'm looking around for postdoc positions for the coming year!
With the NSF pause, I'm assuming a lot of US faculty are reluctant to hire postdocs right now
Is it worthwhile to continue searching for positions in the US? or should I look elsewhere~
Reposts welcomed!
Re LLMs writing term papers: the other day I saw someone say that using LLMs to write is like bringing a forklift to the gym. It's not that the weights need to go up; it's that you need to do the lifting. The doing of the work is the result, not the paper being done or the weights having moved.
Typed holes and type placeholders
the best kind of commits
A and Bs quietly crying in a corner
I wonder if gradually introducing immutability xor shared data annotations to programmers will help?
The GC will always be there except when we reach Rust parity
The irony is despite seeing @lambda-calcul.us use it all the the time I avoided it. Now here I am
My new favorite language is Racket!
Congratulations!
You are the second person who I have heard say PL people will be crucial for doing good evals in LLMs and like to hear more.
I will be interested to read that post about your experiences on the StarCoder team
I think the other way round will be more useful. Library code that ships with dependent types and application code that are type checked statically to verify the library contracts.
Core libraries are used enough that their bugs are often ironed out (t&c apply)
President Musk has said AI will write all code and types do not matter — so Rust has been laid off
Excited about Proofwala, @amitayush.bsky.social's new framework for ML-aided theorem-proving.
* Paper: arxiv.org/abs/2502.04671
* Code: github.com/trishullab/p...
Proofwala allows the collection of proof-step data from multiple proof assistants (Coq and Lean) and multilingual training. (1/3)
Join us! Science Homecoming helps scientists reconnect with communities by writing about the importance of science funding in their hometown newspapers. We’ve mapped every small newspaper in the U.S. and provide resources to get you started. Help science get back home 🧪🔬🧬 🏠
sciencehomecoming.com
"Automatically Verifying Replication-aware Linearizability" has been accepted to OOPSLA 25.
Draft: kcsrk.info/papers/mrdtc...
This work is a substantial improvement over our prior work (dl.acm.org/doi/10.1145/...) on verifying MRDTs.
Big sleep semantics
It is a confusing mess and fraught with people trying to upsell their own products. One of things I did early on was to get a credit card from a company that is easier to get approvals on - like Discover, etc. It is a real credit and helps build credit score
Blog post titled "Concurrency bugs in Lucene: How to fix optimistic concurrency failures" By Benjamin Trent and Ao Li (February 7, 2025) Text reads: "Thanks to Fray, a deterministic concurrency testing framework from CMU’s PASTA Lab, we tracked down a tricky Lucene bug and squashed it"
Back to basics: Concurrency testing in Java!
Our new tool *Fray* correctly solves a 25+ year old problem for real-world software. See this feature from Elastic Labs about Fray's contributions to Lucene.
📰: www.elastic.co/search-labs/...
🔧: github.com/cmu-pasta/fray
📝: arxiv.org/pdf/2501.12618
The entire archive of CDC datasets can be found here.
HUGE shoutout to data archivists- this work is important 👏🙌🏻
archive.org/details/2025...
This statement from CRA outlines the critical importance that the National Science Foundation plays for our nation's prosperity, security, and future. I highly recommend sharing it with your colleagues, and calling your representatives in Congress to convey these benefits to them in your own words!
I am changing my class notes to render text in white on a white background
My UG student Bhavik Goplani presented our ongoing work on verifying idempotency of REST API clients under for request retries at the POPL '25 SRC.
The work involves some exciting effect analysis and a nifty translation to Dafny to automatically prove liveness and safety
Did you add it to your favorites?
What if we could only code by using examples? How can we integrate AI into actual programming languages? I made a little proof of concept:
austinhenley.com/blog/mirrorl...
