AWS Kiro bots manage parallel agent sessions with persistent memory
AWS introduces Kiro bots to handle long-running tasks and parallel sessions, reducing the need for developers to manually transfer context between AI agent interactions.
Daily coverage of AI, developer tools and infrastructure. Each story explains what happened and why it matters, with a link to the original source.
AWS introduces Kiro bots to handle long-running tasks and parallel sessions, reducing the need for developers to manually transfer context between AI agent interactions.
GitHub project Sibyl provides an all-in-one private AI server using Fedora bootc, supporting AMD, NVIDIA, and Intel GPUs with standard APIs.
A new benchmark shows frontier coding models struggle with minor task mutations, often ignoring prompts to guess what hidden tests expect.
As AI agents generate over 40% of pull requests, software engineers must transition into process engineers who design the systems that build, test, and ship code.
Apiiro researchers found 1,122 public GitHub repositories exposing passwords and API keys saved by Anthropic's Claude Code assistant.
Discord's new Rust-based client launches significantly faster and uses 70% less RAM than its Electron predecessor by leveraging the GPUI framework.
Anthropic explains how Claude uses dreaming and salience judgment to consolidate memory, correcting errors and building knowledge structures during idle time.
Amazon Web Services details a serverless architecture using ECS Fargate and FastMCP to give AI agents centralized, metadata-driven access to enterprise data lakes.
Researchers found over 800 unprotected instances of the Artex AI hacking tool online, linking them to major breaches in finance and cloud services.
AWS demonstrates how to scale AI agent security using Bedrock AgentCore Gateway, Cedar policies, and Okta Cross App Access for centralized authorization.
A new wrapper tool called llama-sleep reduces model switching latency on single-GPU setups by parking CUDA state in host RAM instead of reloading weights.
ZelligeAI releases a specialized coding model that thinks in mixed CJK shorthand to reduce token usage while improving math and code accuracy through reinforcement learning.
Researchers demonstrated a root-level host escape from Firecracker microVMs by hijacking saved CPU registers during KVM exits, executing arbitrary code while keeping the VM running.
Ramp data reveals AI token volumes are rising while spending flatlines, driven by aggressive price cuts from OpenAI and Anthropic and improved cost controls by enterprises.
Blockchain analysis firm Chainalysis shut down key services after an agentic cyber attack exposed names, usernames, password hashes, and live tokens for every user.