Faster code review for teams producing a lot with AI, cheaper models and an example of AI applied to software quality at scale.
GitHub launches stacked pull requests to split big changes into small reviews
GitHub moved stacked pull requests to public preview: an ordered series of PRs, each with a focused layer of the change, that can be reviewed separately and merged together in one click. Existing reviews, checks and merge rules keep working. The feature works on the website, the CLI, the mobile app and with agents such as Copilot, and TED's CTO says AI raised productivity so much that PRs got too large to review well.
With agents producing more code, review becomes the bottleneck, and splitting changes into layers makes review faster and more accurate. Teams can adopt the flow without changing tools.
How Dexpert helps: Dedicated Squads & Staff AugmentationSources
- GitHub Changelog: Stacked pull requests are now in public preview
OpenAI cuts GPT-5.6 Luna prices by 80% and Terra by 20%
OpenAI cut the price of GPT-5.6 Luna, its fastest and cheapest model, by 80%, and of Terra, aimed at everyday work, by 20%. It also launched Fast mode in the API, which makes GPT-5.6 Sol up to 2.5 times faster at twice the price. The company suggests combining models in one flow, for example using Sol to plan and Luna to implement well-specified changes and write and run tests.
High-volume tasks whose costs did not add up before become viable with AI. Defining the quality bar and using evaluations to measure where the pricier model really makes a difference is the way to optimize costs.
How Dexpert helps: Artificial IntelligenceSources
Google uses Gemini agents to find and fix flaws in Chrome's code
The Chrome security team explains how it uses language models to find, triage and fix security bugs at scale. In early 2026, a set of Gemini agents began scanning the whole Chrome codebase more efficiently and with fewer false positives, and found a problem that had existed for more than 13 years. Improvements include a knowledge base with the full Git history, SECURITY.md files that explain the system's trust boundaries to the models and a critic agent that reviews the findings.
The same approach can be applied to enterprise systems to review old code and prioritize fixes. Documenting the project's architecture and security rules helps models produce better results.
How Dexpert helps: Artificial IntelligenceWant 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.