AI tools are rapidly evolving, with new capabilities for generating images in UI design and enabling agent collaboration across platforms like Slack and Notion. Updates to models like Gemini 3.5 Flash and GPT-5.5 Instant enhance their interactivity and functionality. New advancements include AI chips, hackathons for AI development, and specialized AI for legal research and speech-to-text.

• Codex can now generate images for UI assets, improving web page aesthetics.
• Claude Code can be tagged in Slack for team collaboration, maintaining context.
• Gemini 3.5 Flash can now control browsers and desktop environments.
• Figma’s new tools allow turning design layers into code, motion design, and shader generation.
• Notion’s developer platform now supports code-based workflows and integration with external agents.
• OpenAI has developed its first AI chip, ‘Jalapeño’, for LLM work.
• Runpod Flash Hack Day offers a hackathon to learn serverless GPU endpoint creation.
• Updates to GPT-5.5 Instant improve conversational abilities and intent recognition.
• Exa Connect offers web agents for querying business databases.
• Perplexity introduces Computer for Counsel for legal research.
• AssemblyAI’s Universal-3.5 Pro Realtime uses call context for speech-to-text.
• Modal Auto Endpoints simplifies running open models in production.
• Executor is an open-source gateway for connecting agents to services.
• Aside is an AI browser with local encrypted data and ChatGPT support.
• Genspark Design generates UI prototypes, videos, and code.
• Hubble is a Markdown notepad for agents with live HTML previews.
• John Jumper, formerly of AlphaFold, is moving to Anthropic.
• Engram is a lab aiming to train personal models that learn from user work.
• Harvey Labs focuses on legal foundation models and firm-owned intelligence.
• A ‘Dev Intel’ role is proposed for keeping up with AI releases and distilling information for teams.