AI
RAG vs Obsidian-Based Memory: A Comparative Analysis
Two paradigms for giving language models knowledge they didn’t have at training time — one statistical, one curated. When does each one win? Contents Introduction Foundations of RAG Foundations of an Obsidian-Based Memory System Comparison When to Use Which Conclusion References 1. Introduction Large language models have a memory problem. Their parametric memory — the […]
First large-scale Blackwell B200 AI cluster goes live in Canada
Alpha Compute announced on 8 May 2026 the handover of its first large-scale NVIDIA Blackwell deployment: a 504-chip B200 server cluster located in Canada, now in final testing and ready for AI compute customers. It is one of the first commercial-scale Blackwell clusters to come online outside the hyperscalers. Blackwell-architecture GPUs pack 208 billion transistors […]
US AI Safety Institute wins pre-release access to Google, Microsoft, xAI models
The US Center for AI Standards and Innovation (CAISI), the successor body to the US AI Safety Institute, announced on 5 May 2026 that it had signed agreements with Google DeepMind, Microsoft, and Elon Musk’s xAI to evaluate their frontier AI models before public release. The deals expand a framework that already covered OpenAI and […]
Google previews ‘Gemini Intelligence’ for Android ahead of I/O
On 13 May 2026, days before its I/O developer conference, Google previewed Gemini Intelligence: a system-level layer that pushes Gemini into the core of Android. The rollout begins this summer on the Samsung Galaxy and Google Pixel lines, then expands across compatible Android devices — phones, watches, cars, glasses, and laptops — through the rest […]
Google releases Gemma 4, betting on intelligence-per-parameter
Google quietly shipped the Gemma 4 family on 4 May 2026, the next generation of its open-weight models. The release continues Google’s twin-track strategy: keep Gemini at the frontier of closed models while using Gemma to set the bar for what open-weight systems can do. Gemma 4 is explicitly engineered for advanced reasoning and agentic […]
EU strikes Omnibus deal to soften AI Act and extend deadlines
On 7 May 2026 the European Parliament and Council reached a political agreement on the AI Act “Omnibus” package, a set of amendments designed to simplify Europe’s flagship AI rulebook before its toughest provisions take effect this summer. The deal pushes back compliance deadlines for high-risk AI systems, which would otherwise have applied from 2 […]
Anthropic in talks to raise up to $50B at a $950B valuation
Anthropic is in early discussions for a new funding round that would push its valuation to as much as $950 billion, the New York Times reported on 12 May 2026. At the top of the range, the AI lab would overtake OpenAI’s last reported $825 billion valuation and become the highest-valued private AI company in […]
AI Week in Review: $700B Capex, GPT-5.5, and a Stalled EU AI Act (April 24 – May 1, 2026)
Weekly AI digest — April 24 to May 1, 2026. If a single week could capture the strange duality of the AI cycle in 2026, this was it. On the same days that Big Tech reported eye-watering capital expenditure plans north of $700 billion for the year, Brussels could not agree on whether to delay […]
How AI Can Remember: A Cognitive Architecture for Persistent Memory
The amnesia problem Every conversation with a large language model begins from zero. No matter how long, how nuanced, or how personal a previous exchange was, the next session starts with an empty context window. The model has no recollection of who you are, what you were working on, or the decisions you made yesterday. […]
IA : les grandes nouvelles de la semaine du 14 au 17 avril 2026
Par la rédaction — 17 avril 2026 La semaine du 14 au 17 avril 2026 s’inscrit dans ce que certains observateurs qualifient déjà de « sprint final vers l’AGI » : sorties de modèles en cascade, rapport annuel de référence du milieu académique, et révélations stupifiantes sur les capacités offensives de l’IA en matière de cybersécurité. Tour d’horizon […]
