Yesterday's Top Launches: 5 Tools from July 22, 2026

Yesterday brought a fresh batch of tools into the world, and while some are aimed at a broad consumer audience, a few represent genuinely interesting additions to the new developer tools ecosystem. From debugging aids to AI fine-tuning platforms, here’s a look at what launched.

AI Attractiveness Test

Curiosity often drives a lot of online behavior, and the AI Attractiveness Test is a straightforward web tool built to satisfy that. You upload a photo of a face, and it returns an AI-generated attractiveness score. The developers emphasize privacy, stating that your photo isn’t stored. Built with a modern web stack like Next.js and TypeScript, it’s a simple freemium tool that feels more like a tech demo or a weekend project than a serious product. It’s the kind of thing that might go viral on social media for a day, but its long-term utility is questionable. For developers, it’s maybe a neat example of integrating an AI vision model into a clean, user-friendly interface.

Loova Ads Studio

For marketers and creators tired of the grind of producing endless ad variations, Loova Ads Studio appears to be a significant time-saver. It’s an all-in-one AI workspace that generates everything from UGC-style videos and product commercials to static images and even avatar-led content. The core idea is to remove the bottleneck of creative production, allowing for much more extensive A/B testing. A particularly clever feature is the “Viral Ad Clone,” which analyzes the structure of successful ads and rebuilds them around your specific product.

The platform leverages some heavy-hitting AI models like Seedance, Kling, and Veo to handle the video generation. It’s freemium and web-based, with plans to integrate directly with ad platforms like Meta and TikTok. If you’re in e-commerce and spend a lot on creative, this could be a game-changer for scaling your testing efforts without scaling your budget proportionally.

Replay QA

Debugging complex web applications can feel like detective work without a complete crime scene. Replay QA steps in as a powerful time-travel debugger designed to fix that. It records entire user sessions in a special browser, capturing every click, network request, and state change. You can then replay the session exactly as it happened and inspect it using the standard developer tools you already know.

What makes it compelling is how it addresses the gap between fast AI-assisted coding and slow, traditional testing. It doesn’t just tell you a bug occurred; it shows you the precise sequence of events leading to it, which is invaluable for diagnosing elusive issues like race conditions. The fact that it’s completely free and can integrate with GitHub for continuous testing makes it an incredibly accessible tool for developers and QA teams looking to squash bugs before users ever see them.

Skippr AI

Imagine having a new employee who never sleeps, speaks multiple languages, and can guide users through your software by actually taking control of the screen. That’s the premise of Skippr AI. It deploys real-time AI agents that act as live guides within your product, handling onboarding, support, and user activation. They’re not just chatbots; they have memory, can perform actions, and are trained on your specific knowledge base.

The promise is a significant reduction in support tickets and better user adoption rates. You can embed an agent with a few lines of code or share a simple meeting link. It operates on a freemium model, starting with 200 free credits. The “human-in-the-loop” option for enterprise clients is a smart touch, ensuring the AI knows when to ask for help instead of guessing. For SaaS companies focused on customer success, this could be a powerful way to provide always-on, personalized assistance.

Tinker

The pace of open-source AI model releases is frantic, but fine-tuning these models for specific tasks remains a complex, infrastructure-heavy challenge. Tinker offers a refreshingly simple solution: a flexible API for LoRA fine-tuning. LoRA (Low-Rank Adaptation) is a technique that lets you efficiently adapt large models by training only a small subset of parameters, saving massive amounts of time and compute.

Tinker’s value is in its focus. It doesn’t try to do everything. Instead, it gives developers and researchers granular control over their data and algorithms while completely abstracting away the pain of managing GPU clusters and distributed systems. You just use the API to point it at your data and the model you want to fine-tune. Being a free service, it significantly lowers the barrier to entry for creating highly customized AI solutions, making it a standout tool for anyone serious about model customization without the DevOps headache.

Quick Links to Yesterday’s Launches: