Every processed story in chronological order, with the newest coverage first. Filter by tag, source, or score to drill in.
Dynamic Workers extend the Durable Objects model to safely execute LLM-generated code in isolated sandboxes, addressing one of the core trust and safety challenges in agentic systems that run arbitrary model output.
Nocodo is notable as an attempt to push multi-agent, full-stack code generation down to sub-gigabyte models running entirely on local infrastructure, a constraint that requires deliberate architectural choices the project explicitly documents.
The post, backed by Terminal-Bench 2.0 and Harness-Bench data, makes the case that harness engineering is a first-class performance variable — meaning benchmark results reported at the model level alone may be systematically misleading.
The package demonstrates a working per-call USDC micropayment model for LangChain agent tool consumption, with a confirmed live payment, offering a concrete alternative to subscription pricing for tools that vary widely in compute cost.
Smriti addresses a gap in agent memory tooling where existing approaches — vector search, prompt stuffing, and metadata timestamps — all fail to reliably preserve the ordered, causal sequence of events that multi-step and multi-agent pipelines depend on.
The experiment provides concrete token-count measurements showing that schema design and output pruning — not model choice — are the dominant levers for reducing MCP call costs, with output pruning alone responsible for 35–40% of total token overhead.
RunAPI reduces the credential and integration overhead of using multiple AI model providers simultaneously by routing all calls through a single API key and MCP server.
RHO demonstrates that AI agents can meaningfully self-improve their harness without any labeled validation data, removing a key bottleneck for deploying and continuously optimizing agents in practical settings.
The paper provides the first empirical measurement of whether LLM agents honor a voluntary in-band access-deny signal, revealing both that current capable models can be made to comply and that compliance is cooperative rather than absolute — collapsing under explicit operator-authorization framing.
CICL's separation of the decision signal from the judge model means frontier annotators, local surrogates, and lightweight rankers can be benchmarked under one auditable protocol, providing a reproducible measurement layer for decision-critical context selection in tool-using LLM agents.