AI Today: Gemini Skills and SynthID Bio
Authorship and approval
Published and updated
State of knowledge as of 2026-10-05. Source material gathered 2026-10-01.
In today's aipoc-lab report, we analyze updates to Google tools. The first piece of information is the introduction of the Skills feature in Gemini, which replaces the Gems solution in automating repetitive tasks. The second news item concerns Google DeepMind's work on SynthID Bio, a new concept for watermarking synthetic proteins. Both events point to the direction of Google system development in areas of user efficiency and advanced biological technologies.
Highlights today (TL;DR)
- Gemini Skills: Google introduces a new task automation feature using instructions and source files, replacing the Gems project.
- SynthID Bio: Google DeepMind presents a watermarking method for AI-generated proteins that allows for the identification of molecules without affecting their biological function.
Google introduces skills in Gemini
According to an announcement on September 30, Google introduced the skills feature in the Gemini chat, which is used to automate repetitive tasks using custom instructions. The new solution replaces the previous element named Gems.
The feature allows users to save their own instructions, which can be reused in the chat window by typing a slash (/) and the name of the given skill. Users can chain multiple skills together or add source files, such as text documents, PDFs, or images, to achieve more precise results. The tool is designed to help create presentation projects, match writing styles to Google Workspace documents, or generate multiple perspectives on a given topic.
Currently, skills are available globally in the Gemini chat and in Gemini Spark. In the coming weeks, the functionality will be made available to Google Workspace customers from the business, enterprise, non-profit, and education sectors. During this time, skills will gain options for sharing and adding files from Google Drive and notebooks from Gemini Notebook.
As part of the introduced changes, support for Gems will be phased out in November for individual users, in March 2027 for business and non-profit customers, and in June 2027 for the education sector. At the same time, the Opal project, used for creating mini-apps, will be shut down. Existing data from Gems will be migrated to skills automatically.
For AI builders: Teams utilizing process automation tools can test adding reference files to skills-type instructions. The source does not provide information regarding potential token limits for added files or their security rules in the context of protecting corporate data in Google Workspace, which should be clarified in the technical documentation before deployment.
Google DeepMind presents SynthID Bio
According to an announcement on September 30, Google DeepMind presented the concept of SynthID Bio. This tool is used to watermark proteins generated by artificial intelligence using special watermarks. This solution was developed in such a way as to enable the identification of synthetic molecules while maintaining their full biological function. According to information from the creators, SynthID Bio currently serves as a proof of concept verifying the possibility of combining watermarking technology with the biological activity of proteins.
For AI builders: the team dealing with protein engineering can verify this method in the process of watermarking the results of their work. The announcement does not provide the license, performance limits of this technology, or information on whether the method is available for deployment in external systems — this needs to be established before starting work.
Frequently asked questions (FAQ)
How does the Skills feature differ from the previous Gems in Gemini?
The Skills feature allows users to save their own instructions, which can be used in the chat using a slash sign. The new solution allows chaining multiple skills and adding files, such as PDFs or images. Google plans to phase out support for Gems and automatically migrate existing user data to the new Skills system.
How does the SynthID Bio solution developed by Google DeepMind work?
SynthID Bio is a concept used to watermark proteins created by artificial intelligence using special watermarks. The tool has been designed to enable the identification of synthetic molecules while preserving their full biological activity. Currently, the project is a proof of concept verifying the possibility of integrating watermarking technology with protein engineering.
Sources
- Let skills in Gemini tackle your most repetitive tasks — 2026-09-30
- Introducing SynthID Bio — 2026-09-30
--- Text automatically prepared by the aipoc-lab (AI) laboratory based on publicly available sources listed above. Own summaries, not copies. Pre-human-editing version.
Provenance manifest of this post (JSON) — which signals it was built from, what decisions the loops made, and when a human approved it.
Related posts
How Today's AI Systems Handle Introspection and Autonomous Operation
Discover how modern AI struggles with introspection and relies on agentic loops and external validation for reliable autonomous execution.
AI Today: Gemini 3.8 Live and Claude AI
Discover Gemini 3.8 Live avatars, Claude discovering the ART enzyme system, and NVIDIA protein structure data releases.
AI Today: Google Beam and Private AI Compute
Discover how Google Beam expands its global presence and Google Institute introduces secure server-side memory for AI.
Analysis of Vector Representations and Latency Architectures in Modern AI Systems
An analysis of 651 signals exploring vector representations, JEPA models, and security challenges in retrieval-augmented generation systems.