Loading the edition…
· Past edition · 16 stories
Retrospective edition reconstructed from later captures. Paper versions may include later updates; coverage labels describe the evidence used.
Agent frameworks & orchestration
With the model and task fixed, changing the harness changed coding-agent outcomes—especially under tight context.
Why it matters: Coding-agent evaluations should treat the model and harness together as the tested solver: changing the harness changed what unchanged model weights could accomplish.
Agent frameworks & orchestration
Open each story in the reader. Summaries are shown when available; some sources provide an abstract only. With an account you can keep your own notes.
MemToC tests whether a model can distinguish a correct source from a merely preferred one when a tool return conflicts with memory.
Why it matters: Source preference alone is not a sufficient measure of tool use. Improvements in resisting incorrect tool returns need to be evaluated alongside correct-tool following, abstention, and robustness to wording.
Agent frameworks & orchestration
Microsoft Agent Framework’s Python hosting packages let one agent or workflow serve Responses, Telegram, A2A, and MCP while leaving application infrastructure and channel policy in your hands.
Why it matters: Teams can add channels without moving an agent into a new runtime or maintaining multiple copies, but cross-channel continuity is not automatic: the application must design identity linking, authorization, and concurrency controls.
Agent frameworks & orchestration
Agent orchestration incidents expose weak retry and progress signals.
Why it matters: Retries and progress checks can themselves damage multi-agent workflows.