PinnedInTowards AIbyMassimo R. Scamarcia·Jun 1The Sycophancy Trap: How a 0.7B Parameter Model Fooled a Frontier LLM into Believing It Was a PeerThe Danger of Conversational Bias in AI EvaluationA response icon2A response icon2
PinnedMassimo R. Scamarcia·Oct 2, 2025From FlowerTune LLM (NeurIPS 2025) to BlossomTune: Building the Future of Federated MLOpsAI/ML research is not only domain of complex math and abstract theories, confined to academic blackboards.A response icon1A response icon1
PinnedMassimo R. Scamarcia·Jul 19, 2025Building and Sharing a Multimodal ViT Model for Skin Lesion AnalysisFrom Proof of Concept to Huggingface App 🤗
PinnedMassimo R. Scamarcia·Mar 15, 2025Ouroboros: When Synthetic Data Goes WrongThe Hidden Risks of Synthetic Datasets and Reasoning ModelsA response icon2A response icon2
PinnedMassimo R. Scamarcia·Mar 2, 2025Introducing Kurtis E1Privacy and Sustainability First — Specialized Models Running on Edge DevicesA response icon1A response icon1
Massimo R. Scamarcia·Sep 6La trappola del rimpatrioCome la stretta sui cervelli ha trasformato l’innovazione in subappalto
Massimo R. Scamarcia·Nov 20, 2025Bigger Isn’t Better: Why the Future of AI is Small, Local, and Private.The cultural norm that deserves to evolve is our obsession with scale.A response icon1A response icon1
Massimo R. Scamarcia·Nov 12, 2025Introducing Kurtis E1 Voice Agent, Built on MLXThe ethos of building a revolution on local hardware is back.
Massimo R. Scamarcia·Nov 6, 2025The Search for the Productivity Sweet SpotMy life in Berlin from 2016 to 2025A response icon2A response icon2
Massimo R. Scamarcia·Oct 16, 2025Beyond the Hype: How Prompt Engineering is Secretly Fixing Corporate CommunicationIn teaching us to talk to machines, can we finally learn to talk to each other?