
PhD @Yale • AI Safety & Security • Automated Reasoning • Research Scientist @AmazonScience @AWSCloud • Views my own
New York, NY
Joined July 2009
- Tweets518
- Following2.3K
- Followers519
- Likes3.9K
Ferhat Erata 🦀 retweeted
3/ Keep the interpretation with the rule. The source, the assumptions, the readings that were rejected and the tests, all reviewed before the rule ships.
Full article, discussing work by @lmthang, @EricWieser, @ferhaterata and others:
eventum.ai/blog/ai-is-discov…
Ferhat Erata 🦀 retweeted
What privileges should we provide to the teacher in On-Policy (Context) Distillation? From my internship @awscloud , we find that simple, instance-independent instructions could outperform instance-specific gold privileges especially in OOD evaluations, even though they contain much less information.
Paper: arxiv.org/abs/2609.32201
Ferhat Erata 🦀 retweeted
.@Amazon and @DARPA brought together 150+ researchers in Seattle last week, with speakers including @awscloud CEO @mattsgarman, Fields Medalist @TaoistTerence, @EPrinceton Associate Professor @BorisHanin, and @NSAGov's Michael O'Hara to explore how AI is transforming mathematical discovery and reasoning.
Ferhat Erata 🦀 retweeted
The Institute for Responsible Superintelligence (RESI): a nonprofit founded by Shafi Goldwasser, Adam Kalai, Vinod Vaikuntanathan. @ShafiGoldwasser @Vinod_MIT @AdamFungi
resi.org
Our goal: build scientific foundations needed to make superintelligence safe by design
Ferhat Erata 🦀 retweeted
This is very nice work from NVIDIA. Like all high-performing approaches on ARC-AGI-3, it uses deep learning-guided on-the-fly synthesis of symbolic world models, i.e. navigating the world by generating programs to represent what you know.
To be clear, like with several other recent claims, scoring 100% on the public demonstration set is not the same as "scoring 100% on the ARC-AGI-3 benchmark". It would be like saying you beat a videogame because you cleared the tutorial level.
Replying to @NVIDIAAI
NVIDIA AVO continuously inspects, plans, implements, and evaluates, using memory, tools, and execution feedback to build on what it learns along the way. This allows the system to sustain progress across long-running tasks rather than starting over with each model context.
Read about AVO and how we built it for long-horizon autonomous agents: nvda.ws/3S5qdS1
Congrats to Anthony Etim, Sri Nampally, and the team! Best Student Paper Award at IEEE HOST 2026 🏆. This paper applies RSRs (Goldwasser & Micali 1984) from our "Learning Randomized Reductions" (ICML 2026 Spotlight) to protect TinyML inference against fault injection attacks.
Our paper "Learning Randomized Reductions" was accepted as a spotlight at ICML 2026 🎉
We built Bitween, a neurosymbolic system that automates discovery of randomized self-reductions (RSRs). 🧵
It discovers non-linear recovery functions for 80% of our benchmark, from sigmoid and ML functions to non-associative algebras and advanced algebraic structures.
Joint work with Orr Paradise, Thanos Typaldos, Timos Antonopoulos, ThanhVu Nguyen, Shafi Goldwasser, and Ruzica Piskac.
arxiv.org/abs/2412.18134 (not camera ready)
Ferhat Erata 🦀 retweeted
A Timeline for Logic, λ-Calculus, and Programming Language Theory by Dana Scott fm.csl.sri.com/SSFT15/Timeli… [pdf]
Ferhat Erata 🦀 retweeted
Just in at #AWSreInvent: New agentic AI capabilities in @awscloud Transform enable large-scale enterprise modernizations across any legacy code and application to be up to 5x faster, reducing costs and better preparing organizations for AI. ⬇️ aboutamazon.com/news/aws/aws…
Ferhat Erata 🦀 retweeted
Happy bday to MIT prof. Shafi Goldwasser, Turing winner & pioneer in the field of cryptography: bit.ly/2zVVs8c
Ferhat Erata 🦀 retweeted
Before software engineers even begin writing code, they have to set the stage of the entire development process. This process requires engineers to make complex tradeoffs between requirements, system design, and implementations details. Current IDEs that rely on AI features, like chat and inline coding, can help engineers get the job done quickly on small development tasks. Still, engineers spend much more time on larger projects—even after the initial code is generated—by conducting rigorous testing and creating documentation. This is where today’s AI IDEs can do more to accelerate the development lifecycle—and this is why we built Kiro.
Kiro is an AI IDE that helps you go from prototype to production with spec-driven development and agent hooks. From simple to complex tasks, Kiro works alongside you to turn prompts into detailed specs, then into working code, docs, and test so what you build is exactly what you want and ready to share with your team. After a developer builds the code with Kiro, Kiro’s agent hooks help engineers solve challenging problems and automate tasks like generating documentation and unit tests. Kiro brings structure and mature engineering practices to AI coding, so you can go from concept to application while being in the driver’s seat every step of the way.
Kiro is free during preview, and supports Mac, Windows, and Linux, and most popular programming languages. We're excited for you to try it out and let us know what you think ➡️ kiro.dev/blog/introducing-ki…
Ferhat Erata 🦀 retweeted
Introducing Kiro, an all-new agentic IDE that has a chance to transform how developers build software.
Let me highlight three key innovations that make Kiro special:
1 - Kiro introduces spec-driven development, helping developers express their intent clearly through natural language specifications and architecture diagrams for complex features. This comprehensive context helps Kiro’s AI agents deliver better results with fewer iterations.
2 - Kiro features intelligent agent hooks that automatically handle critical but time-consuming tasks like generating documentation, writing tests, and optimizing performance. These hooks work in the background, triggered by events like saving files or making commits. It’s like having an experienced developer constantly reviewing your work and handling the maintenance tasks that often get delayed.
3 - Kiro provides a purpose-built interface that adapts to how developers work. Whether you prefer chat interactions or working with specifications, Kiro supports your workflow while keeping you in control of the development process.
Kiro is really good at "vibe coding" but goes well beyond that. While other AI coding assistants might help you prototype quickly, Kiro helps you take those prototypes all the way to production by following a mature, structured development process out of the box. This means developers can spend less time on boilerplate code and more time where it matters most – innovating and building solutions that customers will love.
Starting today, Kiro is available for free during preview and supports most popular programming languages.
Here’s how to get started with @kirodotdev today: kiro.dev/blog/introducing-ki…
Excited to see how developers use Kiro, and to work with the developer community to continue to shape Kiro moving forward.
Ferhat Erata 🦀 retweeted
After careful consideration and hearing the feedback from our fans whilst fully respecting their concerns, our show in Istanbul will be now postponed until 2026 so we can ensure DBL Entertainment will not be involved.
Thank you for your ongoing support, it means everything to us. See you in 2026!
Ferhat Erata 🦀 retweeted
Do join @ferhaterata's PhD dissertation defense!
🪧Title: Learning Randomized Reductions and Program Properties for Security, Privacy, and Side-Channel Resilience
📅When: 2pm, Mar. 19 (in 50 minutes!)
📷Livestream: sites.google.com/view/yacl/s…


























