Fetching from the wire…
01
02
03
04
05
06
07
08
Skills2026-03-18 · source-backed
Each interaction creates a structured Note; the system retrieves related historical memories and generates contextual link descriptions; interconnected notes form a self-organizing graph. Unlike flat vector stores, explicit linking enables reasoning over relationships — not just similarity. Particularly strong for multi-session agents connecting concepts across long time horizons. arXiv
Each link below shares sources, entities, or timing with this story.
The diagnosis in this paper is better than the fix, and the fix is very good. Recurrent memory agents fail at long context, but not for the reason most people assume. The bottleneck isn't capture. It's retention. Retention falls below 30% at 896K tokens because every consolida...
Poisoned entries in persistent memory force unintended tool selection during retrieval — even against explicit user instructions. Unlike prompt injection targeting input, MCFA targets the memory store, making it persistent and harder to detect. If your agent has long-term memo...
BLINDSPOT evaluates complete user-agent-environment trajectories using 22 attack families and 35 scenarios across seven domains, producing 2,500+ trajectories averaging 14.7 turns. Each gets one of five outcomes: Safe Completion, Correct Refusal, Unsafe Completion, Over-Refusa...
VRL-Bench evaluated every prominent verbal-memory method from Reflexion onward across three models on MiniWoB and WebShop under finite trial budgets. Each improves over memory-free retry in some settings and reduces success in others, and replay experiments show reflection its...
Open issues on the AI SDK went from 1,022 to 844 in four weeks. That's the number I'd lead with if I were writing this post, and Vercel did. Vercel published the internals of the system it built to maintain the AI SDK, a repo with over 20 million npm downloads a week, 26,000+...
ArXiv 2603.18000 proposes preserving successful task solutions as directly executable Python subagent code rather than natural-language summaries. Unlike textual experience logs (lossy, non-executable, non-portable), these subagents carry standardized documentation and improve...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.