Everyone's Building Jarvis. Nobody's Even Close.

A Swiss Army knife is a terrible knife.

It's a terrible screwdriver, a terrible bottle opener, and a terrible saw. The only thing a Swiss Army knife is genuinely good at is being small enough to carry around, and the reality of "small enough to carry" is not a strong engineering thesis.

But that's exactly what everyone in AI is building right now.. the everything assistant. The one that reads your email, manages your calendar, writes your code, books your flights, and files your taxes. The pitch sounds incredible but the product is a mediocre version of six different tools duct-taped together.

I know because I built one.

I Built the Swiss Army Knife

I spent about three months on a product called Skippy. It leveraged OCR screen capture, email ingestion, calendar sync, and pattern recognition to become an AI assistant that understood your whole digital life. It attracted a scary amount of interest from investors and Reddit before I'd even launched, people were just asking for beta access non-stop.

And I shelved it. Not because it failed, but because the more I used it the more I realized that a tool trying to do everything just ends up doing nothing well enough to actually rely on. I was playing with automations, making reservations, ordering food on DoorDash, and every single integration felt like a worse version of the thing it was replacing. You sacrifice everything for the benefit of having everything, and the benefit isn't actually that great.

If you don't believe me, go look at Killed by Google sometime. Google Inbox. Google Allo. Google Hangouts. Google Wave. Google Stadia. Products backed by billions of dollars and thousands of engineers. Dead. If Google with functionally infinite resources can't sustain multi-feature products, what makes a solo dev stitching 15 libraries together in Bali think they're going to pull it off?

Sound familiar?

Ask Claude a question and you'll get a brilliant answer, scoped precisely to what you asked. But it won't mention the related problem from last week, it won't connect the dots to the bug you introduced three months ago, it won't anticipate what you actually need versus what you literally typed. And that's not because the model is bad at reasoning, it's because it has no memory. No continuity. No accumulated understanding of you or your work. Every session starts completely from zero. Real reasoning isn't just answering your question, it's answering the pieces surrounding it too, and AI can't do that without knowing what you've been working on, what's gone wrong before, what you actually care about.

TrueMemory

That's what I built .

The bottleneck in AI right now isn't intelligence. It's that your model forgets you exist every time you close the tab. Everyone's building Jarvis and nobody's even close, because they keep building the mouth and the hands and nobody's building the brain.

Josh Adler builds persistent memory systems for AI. Research: .