Your daily AI digest for developers — Monday, September 07 2026
Google has open-sourced Mantis, an AI-agent framework designed to automate the software vulnerability lifecycle, from identifying and validating vulnerabilities to reducing false positives.
Figma's engineering team has documented their use of AI agents to assist their security team in investigating alerts, searching past incidents, and checking compliance.
This guide explores the rise of agent-to-agent marketplaces, where autonomous AI agents are the primary consumers of web APIs, handling complex, multi-step workflows.
This article discusses the x402 protocol, which enables HTTP-native micropayments for AI agents, addressing monetization and resource allocation challenges.
CUA-Lite is an open platform that unifies sandboxes, data, evaluation, and reinforcement learning (RL) for training and benchmarking computer-use agents.
Meta FAIR introduces AI Research Preference Models to rank machine learning experiments, optimizing resource allocation by executing only the most promising ones.
The article explores how AI is being used by job seekers to game the application process, leading to a cycle of inefficiency and frustration.
ASCII smuggling, once used to attack AI, is now being used by spammers to bypass security measures, highlighting a shift in threat vectors.
OpenAI's focus on Recursive Self-Improvement (RSI) aims to enhance their AGI capabilities, offering insights into their research acceleration strategies.
The article discusses the importance of advanced memory and storage architectures to support AI inference and real-time data processing.