Analytical databases that keep getting more versatile, a new option for hosting code designed for agents and a GPU language that opened its code.
DuckDB 2.0 arrives this fall with server mode, triggers and a new storage format
The DuckDB team previewed version 2.0, due this fall and built from more than 10,000 commits. The headline is DuckDB as a server: the Quack protocol becomes stable and the new CONNECT statement lets you query other instances and push SQL straight to PostgreSQL and MySQL. The release also brings triggers, the VARIANT type, asynchronous I/O, a new SQL parser and a new storage format, with a few breaking changes.
DuckDB now covers everything from local analysis to multi-user scenarios and integration with the company's transactional databases. Current users should plan the upgrade with the breaking changes in mind.
How Dexpert helps: Backend & APIsSources
- DuckDB: A Preview of DuckDB v2.0
Cursor launches Origin, code hosting with PRs and GitHub sync
Cursor began rolling out Origin in beta, code hosting designed for agent scale, with repositories, pull requests, code browsing and GitHub sync. GitHub repos can sync in real time, with PR comments flowing both ways between the platforms, and Cursor's agent can answer questions, make changes and update PRs right there. Integrations with Vercel, Depot and Buildkite are already available.
Developer tools are reorganizing around agents, and the choice of code platform now affects the AI workflow. Since it keeps GitHub sync, teams can try it without migrating everything.
How Dexpert helps: Dedicated Squads & Staff AugmentationSources
- Cursor: Origin Code Hosting
Mojo, a language for GPUs and AI accelerators, opens its whole compiler
Modular open-sourced the compiler and entire toolchain of Mojo under the Apache 2.0 license with LLVM exceptions. Mojo is a general-purpose language built to take advantage of GPUs, AI accelerators and other advanced hardware, and it had reached version 1.0, with source stability, the week before. The standard library and hundreds of thousands of lines of kernels were already open.
With an open compiler and a stable release, betting on Mojo for high-performance AI code is safer, without depending on a single vendor. It is an option for those optimizing inference or heavy GPU processing.
How Dexpert helps: Artificial IntelligenceSources
- Modular: Mojo is now open source!
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.