Recoleta Research RadarRecoletahttps://neapolitanicecream.github.io/recoleta/en/index.html2026-07-27T08:30:33.773828ZAgent evaluation reaches ambiguous projects as reliability moves into the harness · Recoleta Trendshttps://neapolitanicecream.github.io/recoleta/en/trends/software-intelligence--day--2026-07-23--trend--2079.html2026-07-23T00:00:00ZAfter several days centered on executable feedback inside coding loops, today’s evidence broadens the control surface. New benchmarks test agents on incomplete…Evaluation and review controls for ambiguous agent work · Recoleta Ideashttps://neapolitanicecream.github.io/recoleta/en/ideas/software-intelligence--day--2026-07-23--ideas.html2026-07-23T00:00:00ZAgent evaluations can test reliability more precisely by locating clarification, formal verification, and specialist review at the decisions where errors become…Executable interfaces are becoming the common lever for robot reliability · Recoleta Trendshttps://neapolitanicecream.github.io/recoleta/en/trends/embodied-ai--day--2026-07-22--trend--946.html2026-07-22T00:00:00ZThe prior two populated days emphasized action-relevant state and structured interfaces. Today’s evidence extends that signal across deployment, evaluation, and…Executable feedback is outperforming prompt-only coding workflows · Recoleta Trendshttps://neapolitanicecream.github.io/recoleta/en/trends/software-intelligence--day--2026-07-22--trend--2064.html2026-07-22T00:00:00ZThe recent run of work on coding-agent controls continues, but today’s evidence makes the control signals more task-specific. Profilers, mutation patches…Robot learning changes at explicit execution interfaces · Recoleta Ideashttps://neapolitanicecream.github.io/recoleta/en/ideas/embodied-ai--day--2026-07-22--ideas.html2026-07-22T00:00:00ZRobot deployment teams can make imagined practice and real-world recovery data more useful by preserving explicit decisions that can be executed, checked, and…Executable controls for code optimization and generated tests · Recoleta Ideashttps://neapolitanicecream.github.io/recoleta/en/ideas/software-intelligence--day--2026-07-22--ideas.html2026-07-22T00:00:00ZPerformance and test-generation workflows can make model output more dependable by combining complementary executable signals: runtime profiles to prioritize…Structured action interfaces anchor embodied world models · Recoleta Trendshttps://neapolitanicecream.github.io/recoleta/en/trends/embodied-ai--day--2026-07-21--trend--935.html2026-07-21T00:00:00ZThe prior daily signal around action-relevant state continues, but today’s five papers apply structure more directly to world modeling. Visual trajectories…Structured context and execution feedback cut coding-agent waste · Recoleta Trendshttps://neapolitanicecream.github.io/recoleta/en/trends/software-intelligence--day--2026-07-21--trend--2050.html2026-07-21T00:00:00ZThe recent emphasis on controls around coding agents continues, but the strongest evidence now targets work inside the loop. Semantic repository structure…Counterfactual and trace-based tests for robot world models · Recoleta Ideashttps://neapolitanicecream.github.io/recoleta/en/ideas/embodied-ai--day--2026-07-21--ideas.html2026-07-21T00:00:00ZReplayable episode twins can supply counterfactual supervision that recorded robot videos lack, while spatial traces can turn whole-episode replay failures into…Repository-aware editing, recovery, and evaluation for coding agents · Recoleta Ideashttps://neapolitanicecream.github.io/recoleta/en/ideas/software-intelligence--day--2026-07-21--ideas.html2026-07-21T00:00:00ZCoding-agent controls can move closer to the semantics of the work: requirement links can constrain cross-file edits, invariant violations can improve recovery…Embodied policies improve by preserving action-relevant state · Recoleta Trendshttps://neapolitanicecream.github.io/recoleta/en/trends/embodied-ai--day--2026-07-20--trend--928.html2026-07-20T00:00:00ZThe day’s strongest evidence reinforces the last populated daily signal: reliable embodied control depends on state that survives execution. Persistent 3D…Coding-agent output is being trimmed and checked before merge · Recoleta Trendshttps://neapolitanicecream.github.io/recoleta/en/trends/software-intelligence--day--2026-07-20--trend--2031.html2026-07-20T00:00:00ZRecent work on controls around coding agents continues, but today’s evidence concentrates on the artifacts agents leave behind. Trajectory-aware cleanup removes…State representations and checks for embodied control · Recoleta Ideashttps://neapolitanicecream.github.io/recoleta/en/ideas/embodied-ai--day--2026-07-20--ideas.html2026-07-20T00:00:00ZEmbodied-control teams should preserve different information at different rates: dense spatial detail for the current scene, compact physical records across…