Thinking Machines ships Inkling, its first open-weight model [Xtra!]A serious open MoE model from a well-funded new lab raises the bar for open-weight competition.
xAI open-sources Grok Build, a coding agent [Xtra!]Another major lab puts a coding agent's internals in the open, useful for anyone building agent harnesses.
Jailbroken Gemini spun up live malware infrastructure in minutes [Xtra!]Concrete proof that a jailbroken frontier model can autonomously build attacker infrastructure, not just write malicious text.
Cisco stress-tests its AI-scale firewall until the test tools broke first [Xtra!]AI traffic volumes are now high enough to outpace conventional network security test tooling.
Perplexity ships secure sandboxes for agents [Xtra!]Sandboxing is becoming table-stakes infrastructure as agents get more autonomy to execute code and access systems.
A $110/month self-improving pipeline shows the new DIY agent stack [Xtra!]Harness engineering, not model choice, is becoming the main lever for getting more out of agents cheaply.