Davide Gallitelli·6d agoThe Two Layers Every Enterprise Needs for Agentic AI to Actually WorkWhy a personal agent and an enterprise concierge aren’t competing approaches — they’re the same architecture viewed from two sides.A response icon1A response icon1
Davide Gallitelli·Jun 26Learning from Experiences: Give Your AI Agents a Memory That CompoundsHow to consolidate agent experience into reusable skills — a nightly “sleep” loop that turns trajectories into procedural know-how.
Davide Gallitelli·Jun 23From Prompt Engineering to Loop Engineering: The Evolution of Agentic AI System Design on AWSAn AWS-native approach to loop engineering and durable agent execution
Davide Gallitelli·May 28Your MacBook Is a Training Cluster — It Just Needed the Right Algorithm.How DiffusionBlocks quietly makes the case for on-device, continuously self-improving AI agents.
Davide Gallitelli·May 20How to Correctly Benchmark Amazon SageMaker Real-Time EndpointsYour homemade load-testing script might be lying to you — follow me in this journey where I end up building a custom benchmarking solution…
Davide Gallitelli·May 13Why Thinking Machines’ “Interaction Models” accidentally proved the Embodied AI thesisMy opinion on Interaction Models, and what they mean for Embodied/Physical AI, and interacting with robots.
Davide Gallitelli·Apr 28Three Ways to Run an Agent on AWS: AgentCore Runtime, AgentCore Harness and OpenAI Managed AgentsAWS now has three distinct ways to deploy AI agents — and choosing wrong means either rebuilding later or paying for infrastructure you…A response icon1A response icon1
Davide Gallitelli·Apr 27One Blackwell GPU Beats Four L40S: Benchmarking Qwen3.6–27B on Amazon SageMaker
Davide Gallitelli·Apr 3The Self-Improving Stack — From CLI to Platform to ParadigmOptimization loops are the new CI/CD. Serverless AWS platform, governance layer, and why the eval is your competitive moat.
Davide Gallitelli·Apr 2autoresearchctl — Ship the Loop as a CLII packaged the autoresearch loop into a pip-installable CLI: 6 verbs, binary evals, 6 mutation operators.
