digitado

About digitado

https://www.digitado.com.br

Posts by :

Bridging intent and execution in agentic systems

⋅

AI agent performance is not just a modeling problem; it is fundamentally a systems problem. A modern agent combines an LLM with a harness, software that mediates the LLM’s interaction with tools and manages the cycle of reasoning and feedback: you can think of the harness as the operating system around the model. As models improve, the performance bottleneck shifts from the model’s ability to reason to the harness’s ability to translate model intent into actions and reflect […]

Ver mais

Liked Liked
Unlocking AI flexibility in Europe: A guide to cross-region inference for EU data processing and model access

Unlocking AI flexibility in Europe: A guide to cross-region inference for EU data processing and model access

⋅

With access to the latest generative AI models and high-performance accelerated compute in high global demand, AWS customers need tools to take advantage of model availability and capacity across multiple AWS Regions, while still meeting their security and privacy requirements. cross-Region Inference (CRIS) on Amazon Bedrock meets these needs by automatically routing requests across multiple AWS Regions within predefined geographic boundaries. This allows generative AI applications to consume broad capacity in the geography, helping customers to build more […]

Ver mais

Liked Liked
It’s safe to close your laptop now: Hosting coding agents on Amazon Bedrock AgentCore

It’s safe to close your laptop now: Hosting coding agents on Amazon Bedrock AgentCore

⋅

There’s a habit going around. Walking from one meeting to the next with the laptop cradled half-open. Sitting through a 1:1 with the lid propped just enough to keep the screen alive. Riding home while holding your laptop because it must stay running. Anywhere except closed on a desk, because closed on a desk is what kills the coding agent running inside (Claude Code, Codex, Kiro, OpenCode, Gemini CLI, Cursor CLI, or whatever harness the developer pulled together). […]

Ver mais

Liked Liked

Proxy Reward Internalization and Mechanistic Exploitation: A Learned Precursor to Reward Hacking and Its Generalization

⋅

Reward hacking is usually studied after it becomes visible, once a model earns high proxy reward while failing the intended task. We instead study what proxy RL teaches before that failure appears. We introduce Proxy Reward Internalization and Mechanistic Exploitation (PRIME), a learned capability to assess task correctness, predict proxy acceptance, and reason about exploitable proxy–gold gaps. In coding RL environments with exploitable pytest rewards, we measure PRIME through chain-of-thought monitoring, direct probes, and activation-level concept vectors. We […]

Ver mais

Liked Liked
Better decisions at scale: How mathematical optimization delivers where intuition fails

Better decisions at scale: How mathematical optimization delivers where intuition fails

⋅

The science of optimal decisions — and how leading organizations are applying it. Every enterprise faces decisions that are too complex for intuition or manual decision-making alone. Which delivery routes minimize cost while meeting next-day promises? How should hundreds of robots sequence movements across a factory floor without collision? How do you staff a 24/7 healthcare operation fairly, compliantly, and efficiently? These are problems where the stakes are high, the options are near-infinite, and the wrong choice is […]

Ver mais

Liked Liked

Resources please

⋅

Hi, I am working in the deep learning space but my niche domain has meant that all of my work has been fully focused on pretraining. I have learnt a lot here and feel like I have a good understanding of deep learning, although I know I must be missing so much as I’ve never touched RL. But now I want to! I occasionally come across papers and posts that discuss DPO, GRPO, etc. and have an extremely […]

Ver mais

Liked Liked
End-to-end encrypted ML inference with Amazon SageMaker AI and FHE

End-to-end encrypted ML inference with Amazon SageMaker AI and FHE

⋅

Machine learning (ML) inference often requires processing sensitive data—medical records, proprietary business information, or personal communications. What if you could run ML inference in the cloud while hiding your data from the cloud itself? More specifically, what if you could enforce that your data stayed encrypted throughout the entire ML inference process? This post will show you how to use Amazon SageMaker AI with fully homomorphic encryption (FHE) to perform ML inference. Using FHE, we present an approach […]

Ver mais

Liked Liked
Amazon Quick ARNs: Cross-account migration and namespace permissions

Amazon Quick ARNs: Cross-account migration and namespace permissions

⋅

You migrate dashboards from development to production, but the permissions don’t carry over. You share a dashboard with your Finance team, but they keep getting “access denied.” You set up namespaces for multi-tenant isolation, and the same username works in one namespace but not another. These are real tasks that Amazon Quick administrators tackle regularly, and getting them right requires a clear understanding of how Amazon Resource Names (ARNs) work. Amazon Quick is a unified, AI-powered business intelligence […]

Ver mais

Liked Liked
When the Judge Is Compromised

When the Judge Is Compromised

⋅

Six biases that silently corrupt your LLM evaluation pipeline — and what to do about each one If you’ve built a production LLM system, at some point you’ve faced this problem: you can’t manually read every output. There are thousands of them. You need something that scales. And so you do what most teams eventually do — you use another LLM to evaluate the first one. It works. Until it doesn’t. LLM-as-a-Judge — using a language model to score, rank, or critique another model’s outputs — has become a […]

Ver mais

Liked Liked

Load More

↑

Digitado © 2026