Lef @elefthei

Proofs @Rise_MSR, @MSFTResearch. Compilers, Formal Verification, Cryptography. Previous: @Penn, @Bridgewater, @MIT. elefthei.github.io Seattle, WA Joined August 2009

Tweets

646
Followers

1K
Following

1K
Likes

3K

Lef @elefthei

8 months ago

@hackerdocc If it’s wrong why does it feel so right

0 0 2 103 0

View Details

Han Wang @handotdev

8 months ago

how it feels listening to this podcast

Andrej Karpathy calls AI Agents slop "Overall, the models they are not there. And I feel like the industry [...] it's making too big of a jump and it's trying to pretend that this is amazing. And it's not—it's slop! And I think they are not coming to terms with it. And maybe

268 638 7K 3.1M 3K

36 86 5K 847K 879

View Details

Lef @elefthei

8 months ago

Thanks to everyone who came to my talk on #Ticl a “Structural Temporal Logic for Mechanized Program Verification” at #OOPSLA. Despite being in the last day, we had a pretty good turn out!

1 0 13 673 1

View Details

Lef @elefthei

8 months ago

Well this is embarrassing.

Michael Saxon @m2saxon

9 months ago

The viral new "Definition of AGI" paper has fake citations which do not exist. And it specifically TELLS you to read them! Proof: different articles present at the specified journal/volume/page number, and their titles exist nowhere on any searchable repository.

99 212 2K 471K 474

0 0 0 151 0

View Details

Sebastian Angel @sebsgangel

9 months ago

If you are at OOPSLA 2025, don't miss @elefthei's presentation of our paper "Structural temporal logic for mechanized program verification" on Saturday October 18, at 3:15 PM (Singapore time). Paper: arxiv.org/pdf/2410.14906. Code: github.com/vellvm/ticl.

1 1 4 307 0

View Details

Lef @elefthei

10 months ago

From Wawa to WA.

0 0 8 174 0

View Details

Alexander Hicks @alexanderlhicks

11 months ago

I'm hiring at the EF: jobs.lever.co/ethereumfounda… If you're into Ethereum, cryptography, and formal verification, please get in touch! 🙂

9 42 256 26K 87

View Details

CAV @confCAV

11 months ago

Emina Torlak from @AWS delivered a wonderful keynote about the Cedar authorization engine. #cav2025

0 1 12 669 0

View Details

Lef @elefthei

a year ago

@lastland0 We can definitely schedule this for next fall. I always enjoy visiting Portland.

0 0 1 37 0

View Details

Lef @elefthei

a year ago

🎓 I have successfully defended my PhD. Deeply thankful to my advisors Sebastian Angel and Steve Zdancewic for their support and mentorship throughout this journey. Next I’m joining Microsoft Research and the RiSE group to work on AI assisted verification.

13 5 135 8K 5

View Details

joomy @joomy

a year ago

excited that my team at @Bloomberg is supporting PhD students in certified programming (and other infra/sec topics too!) through a fellowship. 💻🛡️ includes stipend, tuition, and internship. timely for Rocq and proof assistant folks as science funding tightens. please apply! 📬

Tech At Bloomberg @TechAtBloomberg

a year ago

#CFP: Applications for Bloomberg’s 2025-2026 #Infrastructure & #Security Ph.D. #Fellowship are now being accepted. Submission Deadline: Friday, July 18, 2025 (11:59 PM AoE) bloom.bg/3HLdfmO #InfoSec #supplychainsecurity #ProgramAnalysis #BugFinding

0 12 13 15K 8

0 4 41 3K 13

View Details

Lef @elefthei

a year ago

Reading in the editor is not the same as reading in the rendered PDF.

0 0 1 214 0

View Details

Lef @elefthei

a year ago

Theory is saying a DAG does not have cycles. Engineering is still adding a “visited” vector to the worklist algorithm because you know you will eventually hit a bug and it’s gonna be a pain to debug an infinite loop.

2 0 3 444 0

View Details

Lef @elefthei

a year ago

People using SMT solvers don’t know you can simply implement algebraic simplifications with smart constructors in 5x the time.

0 0 2 215 0

View Details

Lef @elefthei

a year ago

@MarisaVeryMoe Yes that is precisely what I was looking for! So staging an interpreter is good, but hard work. And the performance is still not as good as writing a compiler. I am using Rust so I will need quite a few tricks^TM to make this work. I am undecided but leaning don’t do it.

1 0 1 71 0

View Details

Lef @elefthei

a year ago

An interpreter is simply dynamic dispatch on a loop. A compiler trades dynamic dispatch for static dispatch. Can we have the cake (an interpreter) and eat it too (static dispatch)?

2 0 3 502 3

View Details

Lef @elefthei

a year ago

@MarisaVeryMoe Good idea! I imagine by using a language with partial evaluation you can get quite quite a bit of pattern matching done before getting the env. Then you are essentially dispatching only on references. Are you saying that practically the benefits are not that great?

1 0 1 40 0

View Details

Lef @elefthei

a year ago

@aaroneline Idk I always saw this style in old ML papers

1 0 0 28 0

View Details

Lef @elefthei

a year ago

I used to think ASTs with just two constructors UnaryOp and BinOp where old fashioned, but I think it actually saves a ton of lines of code (especially if using copilot).