I have been witness to many technical “BIG changes” in the last 40 years.
- 1980s – Computer on every desk
- 1990-1995 – AOL (Walled garden)
- 1995-2000 – Netscape (WWW Browser)
- 2006 – jQuery
- 2016 – Figma
- 2020 – Covid / Zoom
- 2023 – Artificial Intelligence
- 2026 – Open Claw
Notice the frequency of milestones is accelerating.
Introducing Open Claw
ChatGPT shows how a natural language dialog can be part of personal computing. Unfortunately, it hallucinates constantly and has a terrible memory. It also doesn’t know that much about me.
Recently, a new model was introduced called Open Claw. It holds both the promise and the risk of the future of AI.
Open Claw runs on a local machine, like a Mac Mini running on your local network. You can think of it as an independent person. It can have it’s own email, it’s own file system, it’s own signal chat, it’s own LinkedIn address. It can have everything that a person can have. It’s only limits are what is possible to do through a computer, which is an awful lot.
Imagine you had a personal assistant. You would give them access to your email and calendar and they could schedule meetings for you, right. You can give this system access in the same way. You could give it access to your brokerage account and it could trade stocks. (Dont! but its possible)
Security
Security is the real problem. Open Claw requires that you trust it with whatever you give it access to. The same way you would trust a personal assistant. They can’t respond to an email without access to your email. This raises the possibility of the agent becoming an imposter, signing up for a credit card in your name and then buying stuff without your knowledge. Keeping malicious other users (or agents) away is a major concern. It is very easy to expose more than you want.
We rely so much on our email and phone as a way to secure our lives. This opens the door to the most powerful system in the world to just take over.
1Password has a great blog post about Open Claw.
New product idea
Open Claw is extremely easy to mess up. Unless you understand network and computer security inside and out, I would not suggest using it. However, I think the fundamentals are the most important thing.
A new vendor could offer a mac mini (or similar hardware) and preinstall all of the right software onto it. You could plug it in and go through a user friendly setup and it would limit access properly. It’s still terrifying, but at least it would be configured properly.
This new system would replace Alexa and similar home smart-devices. It would be the first truly capable AI assistant. I could actually do work for you.
There is a movie from the 1984 that predicted this exact setup. Electric Dreams. The entire movie is on YouTube for free. Embedded below
We are accelerating into the future. I hope we don’t collectively face-plant.
Whatya think?