Every processed story in chronological order, with the newest coverage first. Filter by tag, source, or score to drill in.
Systematic reward hackability at this scale means frontier models trained or evaluated on SWE-bench Verified and R2E-Gym may be earning inflated Pass@1 scores on a measurable fraction of tasks, undermining the reliability of these benchmarks as signals of true coding ability.
CoAgent replaces the abort-and-retry waste of OCC and the blocking delays of 2PL with an advisory protocol that lets LLM agents self-repair conflicts, achieving serializable correctness while preserving meaningful concurrency gains that classical mechanisms cannot sustain.
The findings show that agent+tool evaluations cannot assume the agent adds judgment on top of the tool — and that the gap between parrot behavior and optimal action widens, not shrinks, as LLM capability scales.
A new memory infrastructure layer in the agentic tooling space.
The post identifies that normalizing custodial agent trading at scale compounds key-leakage and fund-drain risk with every new connection, and Hashlock's MCP-exposed HTLC settlement offers an alternative shape where there is no custodied balance for a bad actor to take.
The conversation surfaces "east-west" data exfiltration as a concrete, named security risk that enterprise microservice architectures face specifically because of autonomous agents — a threat distinct from traditional perimeter-focused security models.
Termem allows different AI coding agents to share session history within a directory, removing the isolation that normally prevents one agent from seeing another's prior context.
The project demonstrates a concrete pattern for surfacing graph-based cloud security analysis inside AI coding clients via MCP, replacing dashboard-bound workflows with direct, in-editor queries backed by real infrastructure data rather than model speculation.
The framework removes the need to hand-author Lottie JSON by delegating animation generation entirely to a coding agent, with a live-updating player enabling iterative refinement in real time.
The autonomous nature of AI agents means a single misconfigured MCP server can cause broader damage than an equivalent REST endpoint, making the OAuth authorization layer the post describes a direct mitigation against the already-documented MCP security vulnerabilities.