"We envision a more helpful generation of coding agents to both carry out their tasks and formally prove their implementations against strict specifications. Instead of debugging machine-generated logic, humans dictate what they want. Today, we are taking the first major step toward that vision."
https://mistral.ai/news/leanstral
#formalVerification #mistralAI #artificialintelligence #softwaredevelopment
BOOTOSHI (@KingBootoshi)
작성자가 TLA+를 접하고 에이전틱(agentic) 코딩에 매우 유용하다고 감탄하고 있습니다. TLA+로 설계 상태(state)의 모든 가능한 시나리오를 수학적으로 검증해 버그와 충돌을 예방할 수 있으며, 문제가 발견되면 에이전트가 즉시 피드백을 받아 반복적으로 수정한다고 설명합니다.
HOLY FUK I JUST LEARNED ABOUT TLA+ AND IT'S SO GOOD FOR AGENTIC CODING ur telling ME that i can mathematically fact check every possible scenario of my design STATE to prevent bugs and crashes AND IF IT FINDS SOMETHING THE AGENTS GET INSTANT FEEDBACK AND LOOP FIXING IT TILL IT
🔬 I released Benelos, a lightweight model checker for C developers.
You describe your system in plain C — state struct, transition functions, invariants, liveness goals — and Benelos does a breadth-first exploration of the full reachable state space to verify correctness.
No new language. No external toolchain. Just a single header file.
🔗 codeberg.org/cdsoft/benelos
#Exploitation in the era of #formalverification A peek at a new frontier with adacore/Spark
For decades, software vulnerabilities have remained an unsolvable security problem regardless of years of investment in various mitigations, hardening and fuzzing strategies. In the last years there have been moves to formal methods as a path toward better security. Verification and formal methods can produce rigorous arguments about the absence of the entire classes of security bugs, and are a powerful tool to build highly secure software. AdaCore/SPARK is a formally defined programming language intended for the development of high integrity software used in systems where predictable and highly reliable operation is crucial. The formal, unambiguous, definition of SPARK allows a variety of static analysis techniques to be applied, including information flow analysis, proof of absence of run-time exceptions, proof of termination, proof of functional correctness, and proof of safety and security properties. In this talk we will dive-into AdaCore/SPARK, cover the blind spots and limitations, and show real-world vulnerabilities which we met during my work and which are still possible in the formally proven software. We will also show an exploit targeting one of the previously described vulnerabilities.
Cousot on Abstact Interpretation
🚀 We’re growing! Cryspen is hiring.
Are you passionate about high-assurance software or cutting-edge cryptography? We are looking for talented individuals to join our team and help us build the next generation of secure-by-design systems.
🔹 Compiler and Verification Tools Engineer
🔹 Cryptography Engineer
Join us in our mission to make the digital world more secure.
👉 Apply here: https://join.com/companies/cryspen
#Hiring #Cryptography #FormalVerification #SoftwareEngineering #TechJobs
**Theorem proving with the real numbers**
https://www.cl.cam.ac.uk/techreports/UCAM-CL-TR-408.html
by John Robert Harrison
November 1996
The author used #HOL to formalise real numbers, including metric, sequences and limits, continuity and differentiation, power series and transcendental functions, integration.
There is also a #CAS and IEEE floating standard verification tools.
@jbz You might be joking, but this is actually not so far fetched as one might think:
Code Migration with Formal Verification for Performance Improvement of Legacy Code:
Combining AI with Formal Verification for Efficient Migration of Legacy Code:
IMHO, this is one of the most useful applications of #LLMs.
Either way, the future of #AI will belong to automated reasoning/symbolic AI (maybe in combination with LLMs).
"LLM-Verifier systems get a boost with 4/$\delta$ Bound theorem! #FormalVerification #LLM #SoftwareVerification"
The introduction of Formal Verification tools with large language models (LLMs) has revolutionized software verification, but current methods lack reliability due to the absence of a solid theoretical foundation. This research addresses this gap by developing an LLM-Verifier Convergence...
#FormalVerification #LargeLanguageModels #LLM-VerifierSystems #SoftwareVerification
M. H. Neifer
sayzard
Christophe 🇪🇺
Mike Williamson
jnpn
Cryspen
terrorist :: Feline
Jan 
N-gated Hacker News