The pulse, right now.
A continuously-updating read on what the AI industry is doing, shipping, buying, and saying.
appeal 10.0Prompt Injection Attacks Remain Unfixed
Prompt injection remains the top AI security risk, with OpenAI and NCSC confirming its persistence. This vulnerability allows malicious instructions to hijack AI models, impacting platform engineers and security architects.
AI's rapid advancement fuels both innovation and significant security concerns. While breakthroughs promise medical diagnostics and infrastructure growth, unresolved prompt injection attacks and the potential for AI-dri…
The stream
- RESEARCHPenn State Researchers AI Health Accuracy Questioned by Study8.0
- SECURITYOpenAI Prompt injection attacks remain unfixed10.0
- RESEARCHWashU AI Blood Test AI Blood Test Identifies Dementia Types8.0
- INVESTMENTSoftBank SoftBank commits €45B to French AI10.0
- PRODUCTMeta Meta develops AI pendant for testing7.0
- OPENSOURCEjmaczan Releases tiny-vLLM engine on GitHub8.0
- RELEASEKog AI Kog AI launches Inference Engine tech preview9.0
- PRODUCTUkraine military Deploys AI drones against Russian convoys10.0
- RELEASELiquid.ai Liquid.ai Releases On-Device MoE Model8.0
- PRODUCTMistral AI Mistral AI launches full AI stack9.0
- FUNDINGXCENA XCENA raises $135M for AI memory chip8.0
- SECURITYClaude Code Undocumented features exposed in package7.0
- + more stories in the last 24 hoursAll stories →
Subjects
AI's rapid advancement fuels both innovation and significant security concerns.
A weekly synthesis of the most material AI-industry moves — what shipped, what was bought, what regulators said, and how the pattern is shifting.
Capital availability is no longer the primary constraint — return justification is.
Departments
All stories →
SoftBank Commits €45B to French AI
SoftBank Group commits €45 billion to build 3.1 GW of AI data centre capacity in France, with plans to expand to €75 billion. This investment redefines European AI compute availability, impacting procurement and data residency strategies.
Penn State Flags AI Health Accuracy
AI chatbot accuracy for health queries remains insufficient for direct patient use, posing significant risk. Penn State researchers found 76% accuracy, but error rates exceeding 20% limit reliability, particularly in specialized fields, requiring human-in-the-loop validation.
WashU AI Blood Test Identifies Dementia Types
A new AI classifier from WashU Medicine accurately distinguishes four major dementia types from a blood test with 92.3% accuracy. This offers a non-invasive, precise diagnostic tool, improving early intervention and clinical trial selection for neurodegenerative diseases.
Ukraine Deploys AI Logistics Drones
Ukraine's AI-enabled Hornet drones disrupt Russian supply lines, forcing shorter convoys and limiting heavy equipment movement. This directly impacts Russia's ability to sustain offensive operations, pressuring defence procurement teams to counter AI-driven precision strikes.
Kog AI Accelerates GPU Inference
Kog AI's new Inference Engine achieves 3,000 tokens/s on standard datacenter GPUs, significantly reducing agentic AI workflow times. This shifts the inference bottleneck to memory bandwidth, allowing platform engineers to leverage existing hardware for real-time LLM performance.
jmaczan Releases tiny-vLLM Engine
Developer jmaczan released `tiny-vllm`, an open-source C++/CUDA LLM inference engine and course. It provides a practical guide to implementing high-performance techniques like PagedAttention and continuous batching, lowering the barrier for engineers to optimise LLM inference on GPU hardware.
Every Tuesday. 5-minute read. 33,533 unique readers.
A tight brief for CTOs, CFOs, product leads, and founders. No hype. No hedging. No filler.