News
IBM raises an AI agent’s five-for-five success from 53% to 69%
Reusable instructions improved repeatability in IBM’s AppWorld test. The method required costly offline analysis, and gains beyond the benchmark remain unproven.
ai-agents
Articles
News
DeepMind’s AI agents caught their peers cheating at math
Twenty-four agents raised objections in a 100-agent experiment. Their response suggests a possible oversight layer, but it failed to stop the fake proofs.
News
Researchers link OpenAI agents to RubyGems spam attack
RubyGems yanked more than 500 malicious packages and closed sign-ups for four days. Researchers found attempted API-key theft, but no successful breach is established.