Loading the edition…
· Past edition · 18 stories
Retrospective edition reconstructed from later captures. Paper versions may include later updates; coverage labels describe the evidence used.
Agent frameworks & orchestration
Growing Harness turns recurring agent-control decisions into shared executable code, leaving the LLM to handle task-specific semantic reasoning.
Why it matters: For streams of related tasks, moving recurring control from repeated model calls into reusable code may reduce inference effort while preserving LLM calls for task-specific reasoning.
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.
SWE-Serve measures whether agents can turn inference tasks into production-correct serving changes—not merely patches that pass local tests.
Why it matters: The benchmark makes a production correctness gap measurable: passing local tests does not guarantee that a patch works in end-to-end model serving. That distinction gives future agents a more demanding target than task completion alone.
Agent frameworks & orchestration
JAZ tests how far an agent can go with a minimal loop: one LLM-based invoke primitive, executable code, and access to the interaction history.
Why it matters: The reported results suggest developers could investigate a code-centered loop before adding specialized memory or self-improvement systems. The abstract also reports lower costs in both comparisons.