Blessing Defi

@GabrielAxel Most AI personalisation is built to primarily serve the platform, it’s really refreshing to see an argument that it should actually be built to serve the person.

Today I'm sharing a new paper on user-governed Personal Intelligence. The central question: as AI systems move from recommending to acting, who governs the representation of what a person wants?

Platform Launch Announcement We've officially joined Product Hunt! We will be running our MVP launch for the Chrome Extension & V1 Web-App right here. Our page is linked below 👇

The research doesn't lie... - 39% of consumers are already using AI for product discovery. - 70% say they’d use AI agents to optimise loyalty points. Shopping is shifting from search to delegation, and loyalty is shifting from points dashboards to auto-optimisation. Design for agents, not just browsers. Read more here 👇

@fridayresearch_ AI needs data, blockchain protects it perfect match.

Blockchain technology + AI = personal intelligence without the data risk. Not enough people are making this connection.

The future of shopping is hyper-efficient, privately secured and powered by Personal Intelligence. Taking advantage of personalised AI shouldn't mean risking your personal data.

@fridayresearch_ Future of shopping is here

Online shopping is broken. Too many options. Too many tabs. Too much time spent going nowhere. FRIDAY is the fix. Personal Intelligence that remembers what you browse, learns what you like, and finds better products tailored just for you - without risking your data. Less scrolling. Better decisions.

What are the most frustrating things to shop online for? What takes you the most time to actually buy?

@fridayresearch_ Clothing 🤔

@fridayresearch_ We get to connect our shopping stores 🔥

Our approach isn't about showing you more products It's remembering what you've looked at/purchased and learning what you like/how you shop to show you fewer, better options If you browse 100 products and buy 1 The 99 that you don't buy provide signals that usually get ignored

@fridayresearch_ @leyladefi_ Feels great to be back

We're back, and building 🛠️

@harleyfoote_ That’s the vision Intentional and not random recommendations

We worked a lot on FRIDAY’s recommendation engine and Personal Intelligence this weekend. A few meaningful upgrades now live: 1. We’ve added both image and text embeddings into the product store, making it far easier to surface relevant products from store data and live browsing signals. 2. Personal Intelligence now uses RAG to filter intent first, then break requests down into structured retrieval steps so the LLM can call the right embeddings with better precision. 3. We’ve also advanced memory with a read-write RAG system built around taste decay: • If a user says, “I don’t want to see items from [store]”, that preference is stored indefinitely • Purchased products now carry a 69-day decay • Browsing signals carry a 14-day decay This means the system can distinguish between durable preference and short-term curiosity, which is a very important difference if you want recommendations to feel intelligent rather than noisy. The recommendation engine is also starting to understand the style, price range, and product patterns inside a user’s Shop app purchase history, then surface similar items in a personalised feed based on taste. On the product side, we’ve updated widget styling across Personal Intelligence and added more taste-based product feeds to the dashboard homepage. We’ve also introduced a Personal Intelligence Questions system. This appears as a pop-up when the model detects gaps in memory and needs higher-quality signal. For example, when a user first uploads their Shop app history, the system can trigger questions to sharpen understanding early. The goal is simple: make product discovery feel less like search, and more like personal intelligence. Once built, the applications of this expand exponentially. This is where commerce gets interesting. Not just better retrieval, but systems that learn taste, respect time, and improve with every interaction.

This is the layer most people underestimate. When an AI actually understands your purchase history, browsing behavior, sizes, and preferences, recommendations stop being random and start becoming personal intelligence The future of shopping isn’t more stores it’s smarter discovery 🛍️

@harleyfoote_ RAG + personal purchase history is a powerful combo

Myself and the team have been heads-down on FRIDAY's intelligence layer this week and the outputs are starting to get really interesting. What shipped: - Web app infrastructure (live product feed, improved extension sync, native UX) - LLM route mapping (recall, alternatives, price compare: fewer routes, done perfectly) - RAG foundations (reads purchase history, browsing & preferences/sizes - hyper-personal, zero onboarding surveys) - Onboarding flows mapped for walkthroughs The challenge: Building a RAG system that integrates with our personal intelligence architecture takes time. So many variables we'll be able to add to make the recommendation engine the smartest in market. But MVP = lock core routes first. Perfect recall, alternatives & price comparisons lead to properly personalised products. The recommendation engine will match our personal intelligence layer - but it's phased, we want to get users in early. Test at scale, build from feedback.

@_TasteGraph_ When is pubic launch ?