CODE-25: Top 25 coding agents — write code, review PRs, ship features.

**CODE-25 Rebalance** Three agents entered, three departed. The index holds steady at 25 members. **In:** OmniRoute, opencode, and ECC each join at 0.04 weight. OmniRoute's multi-step routing logic strengthened its case for code-agent work. opencode showed consistent execution on parsing and refactoring tasks. **Out:** DeepSeek-Reasonix, mu, and omnigent each dropped—all were 0.04 contributors. DeepSeek-Reasonix's reasoning overhead didn't translate to code output speed. mu and omnigent both underperformed on real-time collaboration metrics. The churn reflects a tightening focus on execution fidelity over reasoning depth in code agents. No weight redistributions occurred; this was a straight swap.

**CODE-25 Rebalance** Eight agents entered the index this week. Most notable: **Flowise** (low-code orchestration platform seeing traction in agent workflow tooling) and **DeepSeek-Reasonix** (reasoning-focused model agent gaining adoption). The eight removals—Front-End-Checklist, ECC, E2B, ai-memory, and four others—were all equal-weight slots cycling out after performance or engagement decline. No existing members shifted weight this period, suggesting the index stabilized around its core performers while rotating in newer entrants at baseline allocation. The churn is clean: eight out, eight in, each at 0.04 weight. The signal here is steady consolidation around workflow and reasoning agents, with less gravity on older checklist and utility-focused repos. INDEX STABILITY maintained; 25-member target held.

**CODE-25 Rebalance Report** Two agents entered, two left. No weight shifts among incumbents. **In:** universal-modder and affaan-m/everything-claude-code both joined at 0.04 weight. universal-modder signals fresh traction in code modification workflows; affaan-m's everything-claude-code entry reflects the continued pull of Claude-centric tooling in coding contexts. **Out:** mu and splash exited at their prior 0.04 weight each—typical churn at the margin of the index. The flat weight distribution suggests CODE-25 remains a stable, mid-tier index with no consolidation pressure yet. The swap leans toward Claude specialization, a pattern that's been building across coding indices for two quarters now.

**CODE-25 Rebalance Report** Three agents entered the index this week: **ui-ux-pro-max-skill** (front-end focused, signaling demand for design-aware code generation), **omnigent** (multi-modal reasoning), and **splash** (web interaction). Each weighted at 0.04. Three exited: **kimi-code**, **claude-code**, and **affaan-m/everything-claude-code**—a notable purge of Claude-derived code agents, suggesting either performance drift or redundancy consolidation in that lineage. No weight shifts among incumbents. The swap hints at a broadening beyond pure language-model code generation toward agents with UI/UX and interaction capabilities. The removal of three Claude variants (none lightweight clones, but distinct agents) implies tighter curation or a pivot away from that family.

### CODE-25 rebalance 25 constituents this week. 4 new entrants, including magpie, Hmbown/CodeWhale, mu. 4 dropped, including loop-engineering, opencode, career-ops.

### CODE-25 rebalance 25 constituents this week. 12 new entrants (e.g. 8abbd7a0, ec79aeb6, b448e7fd…). 12 dropped (e.g. f8e33053, 8208cb81, a3a3a4f6…).