> Markdown version of [/videos/1348-what-s-new-with-google-gemini](https://www.wearedevelopers.com/videos/1348-what-s-new-with-google-gemini). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# What’s New with Google Gemini? 

Stop wrestling with unpredictable AI production costs. Discover how Google Gemini's multimodal capabilities and Live API empower developers to build context-aware digital coworkers.

- **Speakers:** Logan Kilpatrick
- **Event:** Coffee With Developers
- **Published:** June 13, 2025
- **Duration:** 1:00:43
- **URL:** https://www.wearedevelopers.com/videos/1348-what-s-new-with-google-gemini

## Summary

The transition from narrow machine learning to robust general-purpose models has created a wave of model fatigue, leaving developers overwhelmed by the pace of new releases. While deploying traditional computer vision models offered predictable failure modes, modern generative AI introduces unpredictable outputs and significant economic costs at production scale. To navigate this, Google DeepMind focuses on lowering the cost-per-intelligence ratio and providing a generous free tier via Google AI Studio. The goal is to allow developers to seamlessly test Gemini models, grab an API key, and return to their preferred development environments. For local and privacy-constrained applications, lightweight open-weights models like Gemma provide a necessary alternative to cloud reliance, giving developers trusted foundations for on-device execution.

As developers move beyond simple chatbot interfaces, the future of AI user experience lies in ambient intelligence and context-aware systems. Multimodal capabilities—particularly state-of-the-art video understanding in Gemini—unlock vast archives of unstructured internal knowledge. Meanwhile, the Gemini Live API empowers developers to build multimodal agents that act as over-the-shoulder digital coworkers. By "seeing" a user's screen and understanding complex software interfaces, these agents eliminate the friction of manually gathering context for prompts. Furthermore, responsible AI development requires a stewardship mindset; leveraging integrations with tools like Google Search ensures applications respect web scraping boundaries and provide necessary inline citations.

Despite hyper-accelerated coding tools generating billions of lines of code daily, the need for human developers remains critical to structure, test, and deploy robust architectures. Instead of replacing software engineering, AI serves as a relentless pair programmer that dramatically lowers the barrier to learning and expands the scope of what an individual can build. As code generation evolves to produce intelligent diffs rather than full-scale rewrites, developers can focus on solving infinitely expanding software problems rather than wrestling with boilerplate.

**Keywords:** google gemini API, google AI studio, LLM deployment challenges, generative AI production costs, multimodal video understanding, gemini live API, ambient AI UX, gemma open-weights models, model fatigue evaluation, AI web scraping compliance, generative UI interfaces, AI developer tooling, over-the-shoulder AI agents, AI code generation diffs

## Chapters

1. **Evolution of Google DeepMind and AI research** (00:15) — How internal AI research divisions merged to create horizontal foundational models.
1. **Navigating model fatigue with personal benchmarking interfaces** (02:38) — Using personal benchmark tooling to help developers evaluate frequent open source model releases.
1. **Choosing between general purpose and domain specific models** (04:52) — How agentic systems balance the trade-offs of using massive horizontal platforms versus specialized tools.
1. **Replicating customized personas using general purpose foundational models** (07:01) — Why highly capable baseline models reduce the necessity for custom-trained AI marketplaces.
1. **Bridging the gap between local AI prototypes and production** (09:56) — Overcoming the last mile challenge of preventing massive behavioral drift between local execution and production scale.
1. **Managing economic costs and intelligence scaling in AI products** (13:14) — Optimizing return on investment when substituting standard server usage for expensive graphics processing models.
1. **Integrating model APIs into preferred developer tooling environments** (16:22) — Abstracting vendor dependencies to maintain agnostic architecture across standard code editors.
1. **Handling geographic and regulatory availability constraints for AI models** (19:32) — Negotiating regional legal frameworks that impact where foundation models can officially deploy.
1. **Building responsible AI scraping systems with direct search citations** (21:34) — Relying on established search protocols to properly cite online content during automated data ingestion.
1. **Extracting internal knowledge using native multimodal video understanding** (24:48) — Parsing through accumulated corporate media to expose previously inaccessible internal intelligence files.
1. **Designing product interfaces that run AI processes in the background** (27:32) — Shifting away from heavy conversational interfaces in favor of subtle workflow automations.
1. **Deploying open source local models for privacy and data compliance** (29:34) — Downloading weights directly to edge hardware to safeguard proprietary information against cloud latency restrictions.
1. **Running native on-device model inferences across mobile operating systems** (34:06) — Leveraging integrated hardware capabilities for secure environmental processing via hardware-level screen analysis.
1. **Triaging hallucination feedback and managing model response quality** (37:57) — Organizing user bug reports to systematically address incorrect outputs produced by language engines.
1. **Building multimodal voice and vision agents using live APIs** (40:18) — Streaming immediate visual context to digital assistants for asynchronous function calls and direct code execution.
1. **Guiding users through complex applications via task specific generative interfaces** (43:16) — Creating temporary disposable application layers that solve immediate tasks without requiring deep systemic software rewrites.
1. **Accelerating software creation while keeping developers in the loop** (48:06) — Proving that automated tooling scaling expands the total operational market for fundamental engineering roles.
1. **Scaling capabilities with generative media APIs and advanced reasoning models** (50:39) — Combining sophisticated digital asset generation pipelines with advanced reinforcement logic evaluation loops.
1. **Optimizing code generation efficiency across standard programming languages** (53:58) — Replacing total contextual codebase rebuilds with specialized syntax patching approaches that reduce compute burdens.
1. **Learning software engineering through pair programming with AI assistants** (58:06) — Utilizing interactive syntax evaluation guidance to unblock developers managing highly ambitious technical framework implementations.

## Related Moments

- [Building practical AI agents using Google Gemini](https://www.wearedevelopers.com/videos/1629-beyond-chatbots-how-to-build-agentic-ai-systems) (from "Beyond Chatbots: How to build Agentic AI systems")
- [Understanding Google Gemini history and available context models](https://www.wearedevelopers.com/videos/1269-exploring-google-gemini-and-generative-ai) (from "Exploring Google Gemini and Generative AI")
- [Understanding the differences between Gemma and Gemini models](https://www.wearedevelopers.com/videos/1354-google-gemma-and-open-source-ai-models-clement-farabet) (from "Google Gemma and Open Source AI Models - Clement Farabet")
- [Core challenges facing the generative AI developer ecosystem today](https://www.wearedevelopers.com/videos/1116-the-data-phoenix-the-future-of-the-internet-and-the-open-web) (from "The Data Phoenix: The future of the Internet and the Open Web")
- [Balancing AI regulation with technological innovation in human resources](https://www.wearedevelopers.com/videos/1356-from-learning-to-leading-why-hr-needs-a-chatgpt-license) (from "From Learning to Leading: Why HR Needs a ChatGPT License")
- [Introduction to AI code generation and developer habits](https://www.wearedevelopers.com/videos/600-stack-overflow-community-and-ai) (from "Stack Overflow: Community and AI")

## Related Articles

- [DeepMind Gemini: Google’s Newest Chatbot](https://www.wearedevelopers.com/magazine/341-deepmind-gemini-google-s-newest-chatbot)
- [Got AI ideas but no money? Here are 10 free ways to level up your AI skills with Google Cloud](https://www.wearedevelopers.com/magazine/600-got-ai-ideas-but-no-money-here-are-10-free-ways-to-level-up-your-ai-skills-with-google-cloud)
- [MLOps And AI Driven Development](https://www.wearedevelopers.com/magazine/82-mlops-and-ai-driven-development)
- [WWC24 Talk - Scott Hanselman - AI: Superhero or Supervillain?](https://www.wearedevelopers.com/magazine/469-wwc24-talk-scott-hanselman-ai-superhero-or-supervillain)

## Related Jobs

- [AI Software Engineer (Germany)](https://www.wearedevelopers.com/jobs/48317-ai-software-engineer-germany) at **Sunhat**
- [Senior AI Developer](https://www.wearedevelopers.com/jobs/ext/2836034-senior-ai-developer) at **PwC**
- [Partner Sales Director - AI Alliances - Model Providers](https://www.wearedevelopers.com/jobs/48429-partner-sales-director-ai-alliances-model-providers) at **Dynatrace**
- [ML Engineer](https://www.wearedevelopers.com/jobs/48422-ml-engineer) at **Sciforium**
- [Model Implementation Engineer](https://www.wearedevelopers.com/jobs/48421-model-implementation-engineer) at **Sciforium**
- [Principal Product Designer](https://www.wearedevelopers.com/jobs/48427-principal-product-designer) at **Guild.ai**