AI developer tools · What shipped

Claude Fable 5 has achieved a new high score of 161 on the Epoch Capabilities Index (ECI), surpassing GPT-5.5 Pro by 1 point and marking the first time Anthropic has taken the lead on the ECI in over a year. Epoch AI recently began tracking 13 new evaluations on its benchmarking hub, with 7 incorporated into the ECI, and added nine new external benchmarks spanning agentic work, cybersecurity, algorithm engineering, forecasting, and research-level physics.

Composer 2 by Cursor leads the React Native Evals benchmark at 96.1%, followed by Composer 2 Fast at 94.9% and GPT-5.4 at 85.3%. React Native Evals is an open benchmark measuring AI coding agents on real-world React Native implementation tasks, with 16 models evaluated across framework-specific mobile work including navigation, animation, and async state management. The benchmark snapshot was published July 4, 2026, with quarterly refresh cadence, and currently functions as a display-only reference excluded from BenchLM's overall scoring formula.

v1.43.3 of Gentle AI shipped with fixes for Pi upgrade conflicts, Claude Code Context7 configuration path corrections, OpenCode TUI theme-change crashes, and CodeGraph setup improvements. v1.43.0 added Community Tools to the TUI with CodeGraph as the first optional community tool, including lazy per-project indexing and managed guidance sections across supported agents. v1.41.0 introduced a reworked update experience with opt-in self-update prompts, beta channel support, profile-scoped judgment-day agents in OpenCode, and a new trigger-rules review system for declarative agent behavior decisions.

Odysseus, an open-source self-hosted AI workspace, was publicly released in 2026, bundling chat, agents, deep research, documents, email and local-model workflows into a single interface. The project supports local runtimes including llama.cpp, Ollama, LM Studio and vLLM, with optional cloud API integrations, and provides Docker and native setup paths. The significance for practitioners is the shift from scattered local AI demos toward integrated, privacy-forward workspaces where model selection, agents and productivity tools are consolidated on a single auditable deployment surface, though this requires careful attention to authentication, patching, credential management and security hardening by operators.