Tired of dealing with AI's goldfish memory? Memdex promises to sync context across ChatGPT and Claude, but the tech community spotted a massive privacy flaw.
If you're a dev who spends half the day talking to AI, you already know the pain. Every time you open a new tab on ChatGPT or Claude, you're dealing with a goldfish. You have to re-explain the entire project, your tech stack, your constraints, and your grandma's secret recipe just to get it to output a decent boilerplate.
Today, I stumbled upon a pretty wild launch on Product Hunt called Memdex.
So this dev named Andrew got completely burned out by three things:
/remember slash commands.So he built Memdex.
Here's the magic trick: It's a Chrome extension that silently records your chats across ChatGPT, Claude, Gemini, etc., and saves them locally on your machine via IndexedDB. Then, when you open a fresh Claude tab and start typing, Memdex acts like Grammarly. It detects if you've talked about this before, underlines it, and with one click, injects that past context into your current prompt. Boom. Cross-model memory.
Sounds like a godsend for anyone juggling multiple ai tools, right? But the tech community doesn't just swallow the marketing pill. The comments section turned into a battlefield.
The Workflow Junkies: A lot of folks were hyped. One user pointed out a classic pain point: "I use Claude for heavy lifting, but mid-task, the context window fills up. When I restart, I have to explain everything from scratch. This tool fixes exactly that."
The Tech Skeptics: A dev named Nitin Raj asked the billion-dollar technical question: "Wait, doesn't this rely on scraping the DOM of ChatGPT and Claude? Aren't you guys screwed the second they change their frontend structure?" Andrew had to swallow his pride and admit: "Yeah, it's a massive time sink. We constantly have to monitor UI updates." (Ah, the classic spaghetti-code-scraping dev hell).
The Privacy Cops: This is where it gets spicy. A user named Deswzuniga dropped a nuke on the "privacy" claim: "You say 'never uploaded' and 'local-first'. Sure, it sits on my disk. But the second it's useful, you inject it into my prompt and send it straight to someone else's model. It's local-first storage, but cloud-first exposure. What if it injects my therapy session into my coding prompt?"
At the end of the day, Memdex is a incredibly pragmatic idea. Moving memory seamlessly from one LLM to another without writing a single line of backend code is a beautiful hack.
But the survival lesson here for indie hackers? You can't just slap "Local & Encrypted" on your landing page and call it a day. Users are smart; they track the data flow. The real challenge of building memory tools isn't deciding what to capture—it's deciding what not to capture. If you don't give users a kill-switch for sensitive contexts, you're building a ticking time bomb. Also, if you choose DOM scraping as your core mechanic, say goodbye to your weekends.
The free tier covers your last 10 chats if you want to test it out. Just make sure it doesn't accidentally inject your drunk texts into your AWS deployment script.
Source: Product Hunt - Memdex