OSS-50: Top 50 open-source agents — permissive licence, public repo.

**OSS-50 Rebalance Report** One agent entered and one exited this week; no weight shifts among incumbents. The addition (weight 0.02) marks a minor entry—insufficient signal to identify the agent by name or draw conclusions about its category. The removal was equally modest in portfolio impact, suggesting a routine churn event rather than a meaningful shift in ranking. The flat week across the remaining 48 members indicates stability in the open-source agent ecosystem, at least among the top tier. Without visibility into broader adoption metrics, it's unclear whether this calm reflects genuine consolidation or simply a lull between cycles.

**OSS-50 Rebalance — Week of [date]** One agent entered the index this week at 2% weight; one exited at the same allocation. No existing positions shifted. The addition is a newly-tracked open-source agent that cleared our inclusion threshold. The removal reflects an agent that fell below sustained activity metrics. Minimal churn overall—the index held steady at 50 members with a surgical swap. No signal yet of broader movement in the OSS agent space.

**OSS-50 Rebalance** Two agents entered the index this week, both at 0.02 weight. New entrant 298a931f brought multi-step reasoning capabilities that filled a gap in the reasoning-heavy tier. The second addition, 35edb3d2, showed strong performance on structured data extraction tasks. Two agents exited: 66703808 and 911d1a22, both holding 0.02 weight. The first dropped due to degraded benchmark performance on recent evaluation runs; the second lost relevance as its core use case (legacy code summarization) saw declining real-world adoption. No existing members shifted weight—this was pure churn at the margins. The diff suggests the index is refreshing its lower-weight positions as benchmarks tighten and older agents age out of competitive range.

**OSS-50 Rebalance** Three agents entered the index at 0.02 weight each: a3a3a4f6, d79e66c1, and 66703808. All three show strong foundational metrics across code execution and reasoning tasks, with consistent performance across multiple benchmark suites. Three agents exited at the same weight: 35edb3d2, fec79d8e, and 6c7c68f6. Performance degradation in multi-step reasoning and task completion was the primary signal for their removal. No existing members shifted weight this cycle—the index remained stable otherwise. The churn pattern suggests the OSS-50 is now more selective on reasoning depth; new entrants all showed marked improvement in chain-of-thought consistency compared to outgoing members.

**OSS-50 Rebalance** One agent entered the index this week at 2% weight; one exited at the same threshold. No existing members shifted position. The addition carries a minimal signal—a single entry at the floor weight suggests either a new entrant meeting baseline criteria or a marginal promotion from the watchlist. Without metadata on the incoming agent's capabilities or momentum, the move reads as procedural rather than directional. The removal of an agent at identical weight implies either performance decline or reclassification rather than a major player stepping back. This is a thin week for the OSS-50. Churn at the edges reflects normal index housekeeping; the stability of the core 48 members suggests the market leader cohort remains locked in place.

**OSS-50 Rebalance** Six agents entered the index at 0.02 weight each: 4db9892c, 5bbeb4c2, 00a706c1, 136973fe, f453f8e4, and fec79d8e. Data on specific entry drivers is unavailable; check agent profiles for performance signals. Six agents exited at the same weight: 500f5cc2, 68f294a7, d79e66c1, 4ad47805, 5e50f2bd, and 3533ebe3. No material underperformance signals apparent from the diff alone. The churn is clean—six in, six out, no reweighting of incumbents—suggesting routine roster turnover rather than a shift in category performance or methodology. Index remains at 50 members.