Latent Space interviews Katelyn Lesse and Angela Jiang about the engineering behind Opus 5.5 and the Claude Platform. The discussion covers separating model reasoning from agent actions, scaling APIs, and building reliable managed agents.
Prophecy has launched DataGPT, a service that lets business analysts run complete data workflows inside ChatGPT, Claude, and Microsoft Copilot. Answers include a visual map of the steps that produced them, making the workflow behind each result easier to inspect.
Adobe has released patches for eight critical Campaign Classic flaws that could allow arbitrary code execution, security feature bypass, and privilege escalation. The reported issues include a code-injection vulnerability, making this a relevant patch review for teams operating Campaign Classic.
Researchers say the Oracle PeopleSoft vulnerability exploited by ShinyHunters since May consists of three chained flaws, which Oracle patched under a single identifier in June. The chain begins with server-side request forgery reaching an internal management interface because PeopleSoft trusts its own loopback address.
Google DeepMind has released Genie 4 in research preview for trusted testers. The world model adds real-time audio, more realistic simulations, and text prompts that can steer a simulation while it is running.