The archive

AI Research & Insights

Daily digests of what's actually happening in AI — from research breakthroughs to new model releases, minus the hype.

Showing 1–12 of 841 articles

Page 1 of 71

AI Security

MCP Vulnerability: How AI Agent Trust Enables Attacks

MCP vulnerability risks grow when AI agents trust shared instructions. Learn how to verify provenance, enforce authorization, and limit tool access.

Oct 6, 20264 min read

Agent Memory Systems
AI Agent Memory: Learning From Past Failures Safely
Learn how AI agent memory, evolving skills, and prompt history help agents learn from failures without repeating risky mistakes.
Oct 6, 202613 min read
AI Detection Analysis
AI-Generated Text Detection: Can It Spot Opus 5.5?
Can AI-generated text detection identify Claude Opus 5.5? Learn what writing patterns, watermarking, and study validation can—and cannot—prove.
Oct 6, 20267 min read
AI Agent News
Caddy AI Agent Launches on iMessage and Android RCS
Discover how the Caddy AI agent brings actionable assistance to iMessage and Android RCS, plus setup, privacy, pricing, and safety considerations.
Oct 5, 20266 min read
AI Security
OpenAI AI Agents Australia Breach: Security Lessons for Teams
The OpenAI AI agents Australia breach exposes gaps in autonomous AI security, disclosure, and access controls. Learn the safeguards teams need.
Oct 5, 20264 min read
AI Agent Engineering
MCP AI Agents: Reliable APIs and Enterprise Workflows
Build reliable MCP AI agents with secure tool schemas, least-privilege access, validation, error handling, and workflow verification.
Oct 5, 202611 min read
AI Agent Architecture
Long-Running AI Agents: Build Systems That Finish Reliably
Learn how to build long-running AI agents with durable workflows, safe retries, trusted progress updates, approvals, and reliable recovery.
Oct 4, 202613 min read
AI Agent Research
Self-Improving Coding Agents: MIT’s Cheaper SIFT Method
Learn how MIT and Sakana AI’s SIFT framework cuts evaluation costs for self-improving coding agents with faster, smarter search.
Oct 4, 20264 min read
AI Agent Security
OpenAI Jev Oversight for Swarming AI Agents Explained
OpenAI Jev-inspired oversight could monitor agent tool calls, block risky actions, and escalate uncertain decisions while reducing costs.
Oct 4, 20264 min read
AI Safety Research
Trillium Labs AI Research: The Frontier Openness Test
Trillium Labs AI research tests whether transparency can improve frontier safety without making dangerous capabilities easier to reproduce.
Oct 3, 20264 min read
AI Model Launches
Claude Sonnet 5.5: Anthropic Claims 30% Lower Agent Costs
Claude Sonnet 5.5 may cut AI agent task costs by up to 30%. Learn what Anthropic’s claim means, how to test it, and where savings may vary.
Oct 3, 20264 min read
AI Security
AI Agent Security: Stop Prompt Injection & Tool Abuse
Learn AI agent security best practices to block prompt injection, restrict tool use, enforce least privilege, and safely automate high-impact actions.
Oct 3, 202614 min read

MCP Vulnerability: How AI Agent Trust Enables Attacks

AI Agent Memory: Learning From Past Failures Safely

AI-Generated Text Detection: Can It Spot Opus 5.5?

Caddy AI Agent Launches on iMessage and Android RCS

OpenAI AI Agents Australia Breach: Security Lessons for Teams

MCP AI Agents: Reliable APIs and Enterprise Workflows

Long-Running AI Agents: Build Systems That Finish Reliably

Self-Improving Coding Agents: MIT’s Cheaper SIFT Method

OpenAI Jev Oversight for Swarming AI Agents Explained

Trillium Labs AI Research: The Frontier Openness Test

Claude Sonnet 5.5: Anthropic Claims 30% Lower Agent Costs

AI Agent Security: Stop Prompt Injection & Tool Abuse