AI developer tools · What shipped

NVIDIA released AVO (Agentic Variation Operators), a general-purpose agent architecture designed for long-horizon autonomous tasks. AVO achieved a 100.00 RHAE score on the ARC-AGI-3 benchmark, completing all 183 levels across 25 environments with approximately 12% fewer environment actions than the VISTA system using Claude Opus 5. The architecture uses persistent memory and supervision mechanisms to sustain progress over extended task sequences, demonstrated initially on GPU-kernel optimization where it produced multihead attention kernels outperforming cuDNN by up to 3.5% and FlashAttention-4 by up to 10.5%.

Google has released Gemini 3.7 Flash, a new developer-focused AI model designed for coding and agentic tasks. The model delivers improvements in software engineering benchmarks, achieving 65.3% on DeepSWE v1.1 (up from 49.0% for Gemini 3.6 Flash) and 43.6% on FrontierCode 1.1 Main (up from 34.4%), with stronger performance in code accuracy, debugging, and multi-step task execution. Gemini 3.7 Flash introduces adjustable thinking levels (low, medium, high) for flexible control over latency and reasoning depth, rolls out as the default model across Google Antigravity, Google AI Studio, and Gemini Managed Agents, and is priced at $0.75/1M input tokens and $3.75/1M output tokens during its introductory period through December 31, 2026. The model also improves web development workflows with an Elo score of 1588 on LMArena's WebDev Arena and better design adherence capabilities for UI code generation from design mocks.

AWS Glue 6.0 is now generally available with a 30% price reduction and full support for Apache Iceberg v3, newer versions of Apache Hudi and Delta Lake. The runtime upgrades to Apache Spark 4.1, Python 3.13, and Scala 2.13. New features include the VARIANT data type with automatic shredding for semi-structured data, deletion vectors for row-level updates, geometry and geography data types for spatial processing, and flexible schema evolution. Developer productivity improvements include Spark Declarative Pipelines to eliminate orchestration code, Real-Time Mode streaming for sub-second latencies, and Arrow-native Python UDFs for improved PySpark performance.