Yesterday's Top Launches: 5 Tools from July 21, 2026
A new privacy-focused analytics tool called Kobbe has launched for website owners, offering traffic insights without using cookies.
Yesterday, July 21, 2026, brought a particularly interesting batch of new developer tools and productivity applications to the forefront. We saw everything from privacy-focused analytics to an AI runtime that pushes the boundaries of on-device processing. If you’re tired of the usual suspects and are looking for alternatives that put privacy, performance, or sheer convenience first, this digest is for you.
Kobbe
For website owners caught between needing clear analytics and respecting visitor privacy, Kobbe arrives as a compelling solution. It’s a web analytics tool that completely ditches cookies, offering a privacy-conscious way to understand your traffic, user behavior, and even revenue attribution. Instead of invasive tracking, it uses a lightweight script to gather anonymous data.
Where Kobbe stands out is in its practical integrations. It connects directly with payment processors like Stripe and Paddle to show you exactly which traffic sources are driving sales. It also pulls in data from Google Search Console to give you a clearer picture of your search performance. One of its more forward-thinking features is the ability to track referrals from AI tools like ChatGPT and Claude, something most traditional analytics platforms completely miss. It’s built for anyone from indie developers to SaaS businesses who want actionable insights without the ethical baggage of traditional tracking. Kobbe is available as a freemium web app.
BaseRT
The raw power of Apple Silicon chips has often felt somewhat untapped when it comes to running large language models locally. BaseRT is a new runtime engineered specifically to change that. It’s a high-performance LLM runtime for Apple Silicon that promises to make on-device AI dramatically faster and more practical.
The performance claims are significant: up to 6.4 times faster than llama.cpp and 3.9 times faster than MLX for prompt processing. This kind of speed isn’t just a nice-to-have; it opens up possibilities for real-time applications like always-on AI agents or voice processing pipelines that simply weren’t feasible before due to latency. The setup is designed to be simple—just a single command to install. Because everything runs locally, you get complete data privacy and eliminate the per-token costs of cloud-based AI services. For developers and tinkerers who want to experiment with open-source models on their Macs, BaseRT, which is free and open-source, looks like a game-changing piece of infrastructure.
Rewisp
How many times have you seen a piece of information on your screen—a date, a promise, a data point—only to forget where you saw it minutes later? Rewisp is a fascinating new open-source app for macOS that aims to solve this by giving your computer a persistent memory. It runs in the background, discreetly reading text from your screen to create a searchable history of your digital activity.
Its features are thoughtfully designed. If you type “I’ll send that over on Friday” in a chat, Rewisp can remember the commitment and remind you. You can ask it what has changed on a specific webpage since last Tuesday, and it will show you the differences. It can even store your contact details to auto-fill forms. A key selling point is its strict privacy model: it processes everything on-device, and screenshots are discarded instantly after text extraction, with all data stored locally in a SQLite file. It feels like a utility that could genuinely reduce the cognitive load of managing information across countless tabs and apps.
Double
The job search process is notoriously broken, often feeling like a full-time job in itself. Double is an ambitious attempt to fix this by acting as your personal AI career agent. The concept is that instead of you endlessly scrolling through job boards, Double’s AI scours the web for roles that match your profile and aspirations, then handles the application process for you.
It goes beyond just applying, aiming to cultivate your professional network and build your online reputation so that recruiters start coming to you. The system uses a memory layer to learn from its actions, theoretically getting better at finding you the right opportunities over time. You interact with it via a simple text interface, making it feel like you have a personal assistant dedicated to your career. While the promise of “doubling your income” is a bold one, the core idea of automating the administrative grind of job hunting is certainly appealing. It’s free to try and could be a huge time-saver for anyone actively looking to advance their career.
CitedSpy
As more people turn to AI assistants like ChatGPT and Claude to find product recommendations, a new visibility challenge for brands has emerged: are you being cited? CitedSpy is a tool designed to monitor your brand’s presence within AI-powered search results. It tracks how often and in what context your company is mentioned by various AI models, providing benchmarks against competitors.
This is part of the emerging field of Generative Engine Optimization (GEO). CitedSpy doesn’t rely on static data; it makes live API calls to AI models to sample responses over time, giving a more accurate picture of your visibility. It helps identify gaps—for example, if you’re never recommended when users ask for the “best project management tools.” For marketing and product teams, this data is becoming as crucial as traditional SEO metrics. It’s a clear signifier of how the landscape for discovery is shifting.
Quick Links
For more detailed information on any of these new launches, check out the project pages: