Hey devs, the latest AI news is here! Between December 14 and December 20 , 2025, a number of important AI updates were announced that directly impact developers. Here are the highlights.
Anthropic announced that its “Skills” feature (allowing users to teach Claude repeatable workflows) is becoming an open standard so it can work across AI platforms. The company also added a directory of pre-built Skills from partners like Notion, Canva, Figma and Atlassian and made Skills easier to manage and edit.
Read more: https://www.anthropic.com/engineering/equipping-agents-for-the-real-world-with-agent-skills
OpenAI introduced GPT-5.2-Codex, a version of GPT-5.2 tuned for coding tasks. According to OpenAI, the model improves long-horizon work through context compaction, handles large code changes like refactoring and migrations more reliably, and offers improved Windows performance and stronger cybersecurity capabilities. GPT-5.2-Codex is available across all Codex surfaces and will be added to the API after safety improvements.
If you don’t know what Codex CLI is, please watch my video about it.
Read more: https://openai.com/index/introducing-gpt-5-2-codex/
Google released Gemini 3 Flash, a frontier model designed for speed at a lower token cost. Google says the model is ideal for iterative development because it can quickly reason and solve tasks in high-frequency workflows; it outperforms all Gemini 2.5 models and Gemini 3 Pro in coding benchmarks. Strong multimodal reasoning and tool-use capabilities make it suited for video analysis, data-extraction and visual-question-answering applications.
You can try Google 3 Flash inside Google AI Studio for free. This is the instruction on how to use Google AI Studio:
Read more: https://blog.google/products/gemini/gemini-3-flash/
Google announced A2UI, an open-source project that allows AI agents to generate context-appropriate user interfaces. A2UI can create bespoke UIs from a catalog of widgets based on a conversation’s context-for example, an agent that helps book a restaurant can generate an interface for entering party size, date and dietary requirements rather than asking those questions sequentially. Google says A2UI helps LLMs compose graphical, easy-to-use interfaces for specific tasks.
Read more: https://developers.googleblog.com/introducing-a2ui-an-open-project-for-agent-driven-interfaces/
Thank you for following this week’s AI developer updates. I hope you found the roundup useful! If you like this newsletter, please subscribe to stay informed, and feel free to share your own developer news or thoughts in the comments below.
Cheers! ;)