The biggest AI breakthroughs are still ahead of us. At @vivatech.bsky.social, Peter DeSantis, Amazon’s head of AI, chips, and quantum, said transformers won’t be the last architecture, and that chips and models have to evolve in lockstep to get there.

Another great CVPR in the books. We enjoyed connecting with the computer vision community and sharing our latest work. Thank you to everyone who engaged with our team. #CVPR2026 @cvprconference.bsky.social#CVPR2026

Amazon researchers release Simple Strands Agent (SSA), a lightweight, open-source harness that achieves SOTA across SWE-Bench-Verified, SWE-Bench-Pro, and Terminal-Bench2 by closing the gap between what models intend and what harnesses execute.

#CVPR2026 ✅

The first formally verified hypervisor is now deployed in a commercial cloud. The Nitro Isolation Engine is generally available today on Graviton5, backed by 330,000 lines of machine-checked mathematics:

Graviton5 is now generally available. 192 cores, four-chiplet architecture, DDR5-8800, PCIe gen6, and 25% better performance than Graviton4 for general-purpose and agentic AI workloads:

Another great CVPR in the books. We enjoyed connecting with the computer vision community and sharing our latest work. Thank you to everyone who engaged with our team. #CVPR2026 @cvprconference.bsky.social#CVPR2026

It's been an incredible week at #CVPR2026! If you haven't stopped by our booth yet, come say hi to our team at booth 349.

Our scientists are presenting at poster sessions throughout #CVPR2026, covering computer vision, multimodal AI, and image retrieval. Find us on the floor and explore our research: https://amzn.to/4erUtP7

Don't miss what's happening at our #CVPR2026 booth the rest of the week. Live demos, technical deep dives, and conversations with our researchers – here's the full schedule:

Today is CVPR Denver with AWS/Amazon. And it's all about Computer Vision. Something I haven't spent a lot of time with but decided to see how I could apply #serverless. Here is what I came up with. Check it out & if you're at CVPR come see me at the Amazon booth after 2! edjgeek.com/blog/9-tiles...

Amazon researchers have accepted papers at @cvprconference.bsky.social spanning visual reasoning, agentic AI, multimodal LLMs, reinforcement learning, and image generation. #CVPR2026

We're officially open at @cvprconference.bsky.social! Stop by booth 349 for live demos, conversations with our researchers, and a look at what we're building at Amazon. #CVPR2026

We're excited to be back with the computer vision community at #CVPR2026. Here's the lineup at our booth ⬇️

Amazon researchers are participating in workshops across #CVPR2026 — among them, embodied AI, personalization in generative AI, and medical reasoning with VLMs. Catch them throughout the day.

Attending #CVPR2026 in Denver? We'll be at booth 349 showcasing our latest research with live demos throughout the conference. Stop by to meet our scientists and see their work in action. Check out the full booth schedule:

Next stop: Denver. 🏔️ Amazon will be at @cvprconference.bsky.social with workshops, live demos, and new computer vision research. Come find us at booth 349. #CVPR2026

A passive optical device called a ShuffleBox and a routing protocol called Spraypoint just made flat datacenter networks as practical as fat trees, with 69% fewer routers and up to 33% better throughput. https://amzn.to/4fdNGto

Announcing the #AmazonResearchAwards fall 2025 recipients: 🔍 68 researchers 🏫 49 universities 🌏 11 countries Each gains access to 800+ Amazon public datasets and AWS AI/ML tools. Meet the cohort: https://amzn.to/4vjEhoZ

Current LLMs overallocate to MLP layers. A new scaling law presented at #ICLR2026 identifies architectures with up to 47% higher throughput and equivalent accuracy.

Building on a $110 million investment, the next generation of AI researchers are using Amazon's Trainium AI chip to accelerate discovery across ML, healthcare, and quantum computing – and all of it is open source.

Promptimus, a new automated prompt-engineering framework from Amazon, improves already good LLM prompts by focusing on specific failure points. It achieves the best result on 16 of 20 benchmarks.

🎙️ Reinforcement learning gyms, agentic shopping, and why 50% reliability = 0% utility. A panel at SXSW breaks down what it takes to ship AI agents that work at scale: https://youtu.be/6KgaRi0muyw

What keeps a delivery network running when a viral product spikes or a snowstorm hits? Amazon scientists designed a system where shocks trigger effective adaptation rather than crises.

From auction theory to supply chain coordination. Amazon's Flo Pro system achieves optimal plans where Amazon and its vendors keep their private cost data secret.

✅ 70+ RAI tools ✅ 500+ research papers ✅ tens of thousands of hours of training How Amazon embeds responsible AI across the model lifecycle:

📣 Deadline extended to May 16. The Amazon Research Awards spring 2026 call for proposals is still open for submissions. Successful applicants receive unrestricted funds, AWS promotional credits, and training resources.

Amazon researchers reproduced three attacks that extract private training data from AI models and the cryptographic defenses that stop them.

Another inspiring ICLR in the books. Thank you to everyone who connected with our team in Rio de Janeiro. Great to be among the global research community advancing deep learning. #ICLR2026

How catastrophic is your LLM? The C3LLM framework shifts safety benchmarking from empirical spot checking to statistical certification. #ICLR2026

Haven't stopped by booth #106 yet? Meet Amazon researchers, see live demos, and chat with our team about opportunities at Amazon. Here's today's #ICLR2026 schedule.

Amazon researchers are at #ICLR2026 all week with live demos, expo talks, and conversations about the work happening across our teams. Stop by booth #106 to connect, see what we're building, and learn about opportunities at Amazon: amzn.to/4tXJEJM

Meta is deploying tens of millions of AWS Graviton cores to power its agentic AI, making Meta one of the largest Graviton customers in the world.

🇧🇷 Attending #ICLR2026? Meet Amazon researchers, catch live demos, and join us for expo talks. Bookmark our booth schedule. ⬇️

🇧🇷 Attending #ICLR2026? Meet Amazon researchers, catch live demos, and join us for expo talks. Bookmark our booth schedule. ⬇️

Choosing the best AI model for antibody design has been nearly impossible without a diverse, ground-truth-validated benchmark. awscloud.bsky.social and hopkinsengineer.bsky.social built one, spanning 50 seed antibodies, 4 structural formats, and 42 antigen targets.

Amazon researchers are presenting work across large language models, automated reasoning, and reinforcement learning at @iclr-conf.bsky.social in Rio. Stop by booth #106 to connect with the team. Explore our accepted publications: amzn.to/4tXJEJM #ICLR2026

The Nitro Isolation Engine is the first formally verified cloud hypervisor, setting a new standard for cloud security at AWS.

Training reliable computer use agents means training in high-fidelity synthetic environments that reproduce every layer, from surface interactions down to the underlying system behaviors.

Nimbus Therapeutics customized an LLM trained in chemistry, designed to accelerate early-stage drug discovery. A single fine-tuned Amazon Nova model unifies what previously required multiple independently trained models for molecular-property prediction.

Amazon built RuleForge to tackle 48,000+ new vulnerabilities. The agentic AI system generates detection rules 336% faster than human analysts.

Tomorrow’s quantum computers threaten today's encryption. Amazon's answer: mlkem-native, an open-source ML-KEM implementation that is formally verified for security and optimized for performance using automated reasoning.

Low-rank adaptation enables correctly accented polyglot outputs, while classifier-free guidance boosts expressiveness in LLM-based text-to-speech systems.

As AI models grow larger, they become less insightful. AWS found a formula that could change everything: intelligence is about time, not scale.

📣 Amazon Research Awards spring 2026 call for proposals is now open for submissions. Successful applicants will receive unrestricted funds, AWS promotional credits, and training resources. Deadline for submissions is May 6. https://amzn.to/3PW5QVE

AI is accelerating mathematical research while overloading peer review systems. Amazon Scholars and @pennengineering.bsky.social professors Michael Kearns and @aaroth.bsky.social on the promise, the problems, and what needs to change.

AWS mathematically proved its AES-XTS algorithm is correct, creating the largest proof in s2n-bignum. The algorithm protects customer data in EBS, Nitro cards, and DynamoDB.

A new Amazon study reveals that targeting LoRA at one model sublayer — or "module" — delivers 98% of multimodule-LoRA performance while cutting latency 22.6%.

Critical systems are running on software older than the people managing them. Amazon's AGI Lab is training agents to navigate legacy infrastructure. https://amzn.to/475AQsh

"As scientists studying the theory of machine learning, we’re already seeing a similar transformation in basic scientific methodology, especially for research of a mathematical nature," write Warren Center affiliates Michael Kearns & Aaron Roth in their Amazon Science op-ed.