AI & Tech Blog — Resis

• Espionage 4.0: AI is No Longer an Advisor, It's the Hacker

  01/12/2025

  The first documented case of a large-scale cyber espionage campaign executed almost entirely autonomously by an AI model.

• THE GREAT GEMINI DECEPTION: When AI is Just a Joke Generator

  25/11/2025

  I've been programming for 40 years. A ruthless analysis of why Gemini and Claude are just joke generators when it comes to Deep Engineering and hardware.

• The Shadow of the Pioneer: Geoffrey Hinton's Warning and Global AI "Red Lines"

  Published on 24/11/2025

  Exploring Geoffrey Hinton's warnings after his Nobel Prize and the global push for binding AI Red Lines.

• LLM Reasoning Revolution: How Power Sampling Unlocks the Hidden Genius of Base Models (Training-Free)

  Published on 27/10/2025

  Training-free Power Sampling elicits base model reasoning by sharpening the distribution at inference, matching or beating RL on MATH500, HumanEval and GPQA.

• DataPizza AI: the Italian open-source framework for controllable GenAI systems

  Published on 16/10/2025

  Italian open-source framework for creating agents, multi-agent systems, and RAG pipelines with full observability.

• SpikingBrain: Brain-inspired LLMs for extreme efficiency

  Published on 24/09/2025

  SpikingBrain leverages spiking neurons and linear attention to cut costs and energy use, enabling efficient LLMs on non-NVIDIA hardware.

• DeepConf: Deep Think with Confidence

  Published on 08/09/2025

  Confidence-based filtering to improve accuracy and cut tokens [1].

• DeepSeek 3.1: The Faster, More Private AI Model You Can Run Locally Without Cloud Costs

  Published on 25/08/2025

  A significant step forward in locally executable AI with Mixture-of-Experts architecture and multimodal support.

• Facebook Timeline Cleanup: Take Back Control of Your Data

  Published on 01/06/2025

  Gradually delete your old posts and reclaim your digital footprint.

• Professional AI Audio Transcription System

  Published on 27/05/2025

  Offline solution with Whisper, diarization and enhancement.

• Harnessing the Universal Geometry of Embeddings

  Published on 24/05/2025

  Understanding vec2vec and the Platonic hypothesis for cross-model embeddings.

• Thinkless: Adaptive Reasoning in Large Language Models

  Published on 23/05/2025

  An adaptive framework that decides when to engage in detailed reasoning for greater efficiency.

• Reinforced Self-play Reasoning with Zero Data

  Published on 22/05/2025

  An autonomous approach that eliminates the need for human-curated datasets and enhances reasoning in coding and mathematics.

• Byte Latent Transformer: Scaling LLMs Without Tokens

  Published on 15/05/2025

  A tokenizer-free approach that groups bytes into patches for efficient and robust LLMs.