Python tooling, a new Java release and AI models trained on each company's internal knowledge.
Astral, maker of uv and Ruff, joins the Codex team at OpenAI
Astral announced an agreement to join OpenAI as part of the Codex team. Its Python tools, such as Ruff, uv and ty, add up to hundreds of millions of downloads per month and have become a foundation of modern Python development. According to the founder, OpenAI will keep supporting the open source tools, and the team will explore ways to integrate them better with Codex.
Teams that use uv and Ruff can keep using them, as they stay open, and should see deeper integration with coding agents. Standardizing the Python environment with these tools is still good practice.
How Dexpert helps: Backend & APIsSources
- Astral: Astral to join OpenAI
Java 26 brings HTTP/3 to the HTTP client and JVM performance gains
Java 26 shipped with a smaller set of features that lays the groundwork for bigger changes. New features include ahead-of-time object caching with any garbage collector, which helps reduce JVM warm-up time, G1 throughput improvements and HTTP/3 support in the HTTP Client API. The release also removes the old Applet API.
For enterprise systems in Java, less warm-up time and better throughput mean faster responses, especially in containers and environments that scale on demand. It is a good moment to plan upgrades from old Java versions.
How Dexpert helps: Enterprise Apps & SaaSSources
- Hanno Embregts: Java 26 Is Here, And With It a Solid Foundation for the Future
Mistral Forge lets companies train AI models on their internal knowledge
Mistral AI launched Forge, a system for enterprises to train models on their own data, such as internal documentation, code, structured data and operational records. It covers pre-training, post-training and reinforcement learning to align models and agents with internal policies and processes. The company says it already works with organizations such as ASML, Ericsson and the European Space Agency, and highlights that the models stay under the customer's control.
Companies with a lot of proprietary knowledge and compliance requirements get an alternative to generic models. The first step is to organize and clean internal data, which is the base of any custom model.
How Dexpert helps: Artificial IntelligenceSources
- Mistral AI: Introducing Forge
Want to put one of these to work in your company?
Dexpert builds custom systems, apps, integrations and AI solutions. Tell us what you need and get an estimate, or take the free digital maturity assessment.