i-SCOOP
Digital strategy and business in a holistic context
- Website
- i-scoop.eu · all feeds from this site
- Feed URL
https://www.i-scoop.eu/feed/
Read it in your FeedListing reader
Or add to: Feedly Inoreader Raw feed
Latest posts
-
Pika H3 Flex is open from day one, but Europe can only rent it
Pika H3 Flex renders video with sound in six steps and ships open weights, but its licence excludes the EU. Speed, real prices, the clause and what to do. The post Pika H3 Flex is open from day one, but Europe can only rent it appeared on i-SCOOP on i-SCOOP
-
Dulo’s $130 million bet on AI that designs and validates hardware, and where Europe stands
Sebastian Thrun's Dulo raised $130M for AI that designs and validates hardware. How it compares, why validation is the moat and what Europe can do. The post Dulo’s $130 million bet on AI that designs and validates hardware, and where Europe stands appeared on i-SCOOP on i-SCOOP
-
What is harness engineering? How to build the environment your AI agents work in
Harness engineering is designing and improving everything around an AI model that makes an agent reliable. Where the term comes from, guides and sensors, and eight practices. The post What is harness engineering? How to build the environment your AI agents work in appeared on…
-
What is an AI harness? Agent harnesses explained
An AI harness, or agent harness, is the software around a language model that turns it into an agent. What it does, its parts, examples and how to choose one. The post What is an AI harness? Agent harnesses explained appeared on i-SCOOP on i-SCOOP
-
DeepSeek’s Super-agent starts work on its own, and treats your team’s attention as the budget
DeepSeek's open Super-agent starts work on its own but guards your team's attention. How it works, what it costs, GDPR and six controls to copy. The post DeepSeek’s Super-agent starts work on its own, and treats your team’s attention as the budget appeared on i-SCOOP on i-SCOOP
-
Ludi 0.2 teaches robots to read human intent, and Europe’s rules already cover what they see and remember
Ludi 0.2 from Krafton's Ludo Robotics aims to make robots read human intent. How it works, what the evidence shows and which EU rules already apply. The post Ludi 0.2 teaches robots to read human intent, and Europe’s rules already cover what they see and remember appeared on…
-
Meta’s Style Sketch builds a wardrobe from your Instagram photos, and the closet becomes a data asset
Meta Style Sketch builds a wardrobe from Instagram photos. What it does, what its privacy label declares, the EU rules in play and what fashion brands learn. The post Meta’s Style Sketch builds a wardrobe from your Instagram photos, and the closet becomes a data asset appeared…
-
GPT-6 Rosalind shows OpenAI now sells access, not just models, and European labs must govern it
GPT-6 Rosalind runs on GPT-6 Astra's weights with fewer biology refusals. What it does, what approval obliges you to do, and the questions EU labs must ask. The post GPT-6 Rosalind shows OpenAI now sells access, not just models, and European labs must govern it appeared on…
-
Microsoft’s Mage-VL-Cowork fits a desktop agent into 4B parameters, and the training recipe matters more than the score
Mage-VL-Cowork is a 4B Microsoft agent that picks between clicking and code. What it scores, why the recipe matters and how European teams can pilot it. The post Microsoft’s Mage-VL-Cowork fits a desktop agent into 4B parameters, and the training recipe matters more than the…
-
Mistral Large 4 gives Europe a trillion-parameter model it can own, but sovereignty stops at the chip
Mistral Large 4 now has open weights. What changed, how sovereign it is layer by layer, what self-hosting takes and a five-step plan for European firms. The post Mistral Large 4 gives Europe a trillion-parameter model it can own, but sovereignty stops at the chip appeared on…
-
Gemini 4 Carbon tops the frontier benchmarks, and puts cyber-first AI governance to the test
Google DeepMind unveils Gemini 4 Carbon, topping the AA v5.0 index at 59. See how it compares to Argon, Opus 5.5 and Astra on cyber, cost and EU readiness. The post Gemini 4 Carbon tops the frontier benchmarks, and puts cyber-first AI governance to the test appeared on i-SCOOP…
-
Nvidia Granary v2 turns speech training into a multi-task data pipeline
Nvidia has released Granary v2, an open data curation pipeline covering 463,062 hours of speech across 48 languages. By pairing audio with multiple aligned targets, it enables promptable models that switch between verbatim and clean text. The post Nvidia Granary v2 turns speech…
-
Local AI on the M6 MacBook Pro, what it can run and when your business needs more memory
Apple's M6 MacBook Pro reads prompts 3.5x faster than an M4 but stops at 32 GB. Which local AI models run, how fast, and which setup fits which role. The post Local AI on the M6 MacBook Pro, what it can run and when your business needs more memory appeared on i-SCOOP on i-SCOOP
-
Nvidia’s RTX Spark laptops bring local AI to Windows, and memory decides who should buy one
Nvidia's RTX Spark N1X laptops run large AI models locally. Specs, euro prices, the Surface Laptop Ultra and how much memory your organisation needs. The post Nvidia’s RTX Spark laptops bring local AI to Windows, and memory decides who should buy one appeared on i-SCOOP on…
-
Nemotron 3.5 Transcribe covers 23 of the EU’s 24 languages, but tells speakers apart in only six
Nvidia's open Nemotron 3.5 Transcribe covers 23 EU languages, but labels speakers in only six. Accuracy per language, self-hosting, GDPR and what to test. The post Nemotron 3.5 Transcribe covers 23 of the EU’s 24 languages, but tells speakers apart in only six appeared on…
-
NVIDIA MotionBricks makes motion a building block for games and robots, and most of it is open
NVIDIA MotionBricks generates character and robot motion in 2 ms. How it works, what is open, the licences and what European teams can do now. The post NVIDIA MotionBricks makes motion a building block for games and robots, and most of it is open appeared on i-SCOOP on i-SCOOP
-
AI is rewriting the security agenda, and Europe’s reporting clocks are already running
AI finds flaws in hours and attacks can run in about 18. What the a16z CISO survey, the CRA's 24-hour clock and NIS2 mean for European security teams. The post AI is rewriting the security agenda, and Europe’s reporting clocks are already running appeared on i-SCOOP on i-SCOOP