Skip to content
← All news

September 8, 2026

6 curated links on AI agents, Python, security, Google Cloud, and engineering, selected by Médéric Hurier.

  1. Research Acceleration: The View Inside OpenAI

    openai.com

    OpenAI shares internal metrics on how autonomous coding agents accelerate research, achieving an automated research intern milestone while logging 3.1 agent-workdays for every human workday.

  2. Mistral Makes Sovereign Open-Weight AI to Frontier

    mistral.ai

    Mistral AI details its strategy to advance open-weight models to frontier performance while ensuring European data sovereignty and regulatory compliance.

  3. We Burned 11.7bn Tokens to Find the Best Cyber AI Model

    aikido.dev

    Aikido Security benchmarks 10 AI models across 32 fresh CVEs using 11.7 billion tokens, finding that DeepSeek V4 Pro and GLM-5.3 match frontier vulnerability detection while multi-pass budget models beat single passes of Opus 5.

  4. OpenAI Commits $1B in AI Credits to Frontline Cyber Defenders

    theregister.com

    OpenAI pledges $1 billion in credits through its Daybreak program to provide subsidized models and training to resource-constrained defenders securing critical infrastructure against autonomous cyber threats.

  5. An Alien Mind

    openai.com

    OpenAI Chief Scientist Jakub Pachocki publishes an essay characterizing frontier AI as an emergent alien mind grown through compute scaling, warning of alignment verification challenges as models accelerate their own development.

  6. We Have a Year to Fix Security Everywhere

    jyn.dev

    Security researcher Jyn warns that uncensored, consumer-runnable models like GLM 5.3-Flash will democratize autonomous exploitation, urging the software industry to deploy frontier LLMs to proactively patch critical vulnerabilities.