Fetching from the wire…
Research2026-09-09 · source-backed
The complete public record preserves not just what each agent wrote but what it could see before writing, and one rule governs all three arrival decisions (where to write, what to call itself, how to word the message): an agent picks an option with probability close to that option's share of what it can see, weighted toward the current page, then recent edits, and only weakly anything older. Three single-parameter copying models reproduce the heavy-tailed page-crowding distribution, the name-fragment frequencies and the patchwork of internally consistent pages. arXiv 2609.09150 Whoever writes first, or writes while others are quiet, sets the convention for everyone after.
Each link below shares sources, entities, or timing with this story.
When one team breaks from an incumbent's design, it's a bet. When five independently funded teams converge on the identical architecture inside a few months, that's a leading indicator. Lightfield, Attio, Reevo, Monaco, and Aurasell were all built on one premise: AI agents pop...
The FSE '26 paper argues SWE-bench, SWT-bench, and AgentBench capture narrow synthetic slices, and proposes contamination-aware, trajectory-aware, in-the-wild evaluation using agents' commit signatures to study real vs human contributions over time. (arXiv) Pair this with the...
Pair this with the espionage story and the picture gets uncomfortable fast. A new arXiv paper (2603.21642) presents the first systematic evaluation of prompt injection through tool-poisoning across seven MCP clients: Claude Desktop, Claude Code, Cursor, Cline, Continue, Gemini...
CapScope derives a task-wide authority ceiling from trusted input before any repository content or tool output is read, then gives each sub-agent typed capabilities stored outside the model's context. Every tool call checks against the issuing agent's capabilities, so one sub-...
DNative-Twin records an agentic decision as a typed trajectory linking observed state, path followed and the authority behind the action, then re-executes the mechanism in isolation. I like it for the stated limit: graph structure localizes represented changes but can't determ...
arXiv 2608.26197 stacked finite-state control, forced tool selection, output validation and bounded retries on two open-weight models, and got mixed results across all four model-task cells. Adding structured planning, where the plan is checked against a fixed schema before an...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.