- AI Report by Explainx
- Posts
- Open Source Rival to DeepMind’s AlphaFold
Open Source Rival to DeepMind’s AlphaFold
LiteFold simplifies protein folding, Gemini 2.5 Pro boosts AI web dev, and Pinterest’s AI visual search redefines shopping.
This week in AI and tech, major updates are reshaping how we research, build, and shop. LiteFold is making protein structure prediction more accessible with an open-source, GPU-powered platform that simplifies folding model usage for researchers and students—no complex setup required. Meanwhile, Google’s Gemini 2.5 Pro is setting a new benchmark for AI-assisted web development, making it easier to build rich, interactive apps with a single prompt and outperforming previous models in both UI generation and video understanding. On the consumer side, Pinterest is rolling out advanced AI visual search tools that let users shop and explore styles using images instead of keywords, powered by new visual language and multimodal embedding models for highly personalized results.
Explore more in this week’s roundup of cutting-edge innovation.
LiteFold: AI Protein Predictor
LiteFold is an open-source protein structure prediction and visualization server designed to make AI-powered protein research accessible to a wider scientific community. By simplifying access to state-of-the-art folding models like ESMFold, LiteFold eliminates the complexities of GPU setup and software dependencies that often act as barriers for researchers and students. The platform offers a lightweight, GPU-powered inference server, supports simple self-hosting with Docker, and provides a user-friendly interface for submitting prediction jobs, monitoring progress, and visualizing results via REST API. Future plans include support for additional models (such as AlphaFold2), RNA structure prediction, and protein-ligand docking. LiteFold aims to streamline the protein engineering workflow, allowing scientists to focus on discovery rather than infrastructure, and is open to contributions from the structural biology and machine learning communities.
Gemini 2.5 Pro: Enhanced Web App Development with AI
Google has released early access to Gemini 2.5 Pro Preview (I/O edition), an updated AI model with significantly enhanced coding capabilities-especially for building rich, interactive web apps. This version improves on UI development, code transformation, code editing, and complex agentic workflows, making it easier and faster for developers to create visually appealing and functional applications. Gemini 2.5 Pro now leads the WebDev Arena Leaderboard by 147 Elo points over the previous version, reflecting strong human preference for its web development output. It also excels in multimodal tasks, such as video understanding (scoring 84.8% on the VideoMME benchmark), and is available through the Gemini API in Google AI Studio, Vertex AI, and the Gemini app, where users can build interactive web apps with just a single prompt.
Pinterest’s AI Visual Search: Find & Shop Your Style Instantly

Pinterest is rolling out new visual search features powered by generative AI to offer a more personalized and intuitive discovery experience, especially for women’s fashion in the U.S., Canada, and the U.K., with plans to expand to other categories and regions. These features allow users to search and shop for outfits and styles using images instead of words, breaking down and decoding visual elements like color palettes, fits, and aesthetics. When viewing a Pin, Pinterest now generates descriptive keywords to help users articulate what they like, while a new animated glow and refinement bar enable further customization-such as adjusting for style, color, or occasion. Behind the scenes, Visual Language Models (VLMs) and multimodal embedding models deliver rich, tailored search results, making it easier for users to find and shop for content that matches their unique tastes without being limited by vocabulary.
Hand Picked Video
In this video, we’ll look at how an AI Nutrition Agent creates personalized meal plans based on your health data, dietary preferences, and goals—automating research to optimize your diet effortlessly!
Top AI Products from this week
Hedy 2.0 MacOS App - Hedy is the first AI tool that provides real-time, customized insights and recommendations to help you excel in your business meetings, classes, doctors appointments, and more. Favorited by thousands on iOS and Android, you can now tap into Hedy on your Mac.
Brev - A new kind of OKR platform — where agents help automate the busywork, hold teams accountable, and coach teams to success - so your team can focus on actual impact.
Sagehood AI - At Sagehood, we're redefining how investors navigate the U.S. stock market. Today, we're excited to launch our newest feature: Portfolio, a smarter, faster way to diagnose and rebalance your investments using domain expert AI Agents.
Invent by Relevance AI - Invent by Relevance AI lets you build AI agents by describing tasks in plain English—no coding needed. Describe what you need, and watch as Invent instantly transforms your words into fully-functioning AI agents and custom tools—ready to work for you.
Zapier MCP 2.0 - Zapier MCP 2.0 lets your AI securely connect to 8,000 apps & 30,000+ actions via OAuth. With tailored client integration for Claude, Cursor & Windsurf, multi-server support, private apps and more— it's the fastest way to make your AI do real work.
Pluno - Escalate Zendesk tickets to Slack or Jira in one click, with full context & key details via smart, customizable summary templates. Keep teams aligned & customers updated; no manual tracking, no missed updates. Faster resolutions, better support, higher CSAT.
This week in AI
WCAG 3.0 New Scoring for Accessibility - WCAG 3.0 shifts from pass/fail checks to graded scores, focusing on real user outcomes. It aims for better usability, not just compliance, with Bronze, Silver, and Gold levels.
AI Meal & Glucose Tracking - Oura Ring now logs meals via AI and tracks glucose with Dexcom Stelo, offering personalized metabolic insights in the app.
AI Clarity - Google’s new iOS feature, Simplify, uses Gemini AI to instantly clarify complex web text, making jargon easy to grasp without leaving the page268
AI Slide Generation Made Easy - Reverse engineering PowerPoint XML lets AI create custom, well-designed slides fast, even handling images, links, and layouts for better presentations.
OpenAI Open-Weight Model - OpenAI CPO, Kevin Weil saying planning to release Open-Weight model built on democratic values, one generation behind frontier.