LarsenCC

Super cool!

1. Self healing means that it detects it needs some new helper function to complete a task. So, it adds it by itself while doing the task. 2. Will consider yes. 3. Nothing. Only if you decide to use remote browsers, we use the API Key to create one for you

Would be crazy if Opus 4.7 let this happen haha

Interesting way to put it. Maybe it is.

Browsers don't sign ToS, users do. Use it with care.

PW is usually easier to detect. Using raw CDP is in our opinion much better for this.

Lmaooo.

Curious why? You can just take this and run locally or deploy anywhere you'd like with any provider agent provider.

Hey! Where did you submit this exactly? Can you provide a link? Will ask others on the team also, but I am not sure what you are referring to.

Agreed, the pattern is converging across the industry. The Unikraft setup is where it gets interesting for us with sub-second boots (or sub 100ms even), scale-to-zero that suspends the VM after a few seconds of idle (frees resources), and dedicated bare metal in AWS so we're not sharing hardware.

We haven't open-sourced the control plane glue yet but it's something we're thinking about. browser-use itself is open source. The sandbox infra on top is the proprietary part for now.

Fair point, and you're right that those three steps alone aren't a security boundary. They're defense-in-depth, not the primary isolation.

The actual security model is the architecture itself: the sandbox runs in its own VM inside a private VPC. It has no AWS keys, no database credentials, no LLM API tokens. The only thing it can do is talk to the control plane, which validates every request and scopes every operation to that one session.

So even if you bypass all three hardening steps, you get a session token that only works inside that VPC, talking to a control plane that only lets you do things scoped to your own session. There's nothing to escalate to.

The bytecode removal, privilege drop, and env stripping are just there to make the agent's life harder if it tries to inspect its own runtime. Not the security boundary.

Man I was also surprised haha

Thats awesome!

So cool!

I was thinking of doing a really simple app to just capture your txs after you have paid at e.g. the grocery store. Maybe I will build it some time in the future.

A TG bot also seems interesting ngl

It's just a premade spreadsheet indeed.

Oof haha. My bad.

I wanted to sync accounts some time ago and decided to not do it at all. I lost all track of where what is, how I managed to spend etc.

Writing it down yourself is still the best option for me at this point. I feel on top of things.

I can also feel it sometimes yeah

Ah no, not linkedin!

You can check how it works in the demo. Idk, charging more than $5 for a google sheet like this felt wrong.

Its also simpler I believe

I might just make one then :D

You are right, I did edit the comment since after I read it again I found it too defensive, so I rephrased it a little. Hope you don't mind!

The reddit page you posted was made quite some time after the initial post that went "viral" (https://www.reddit.com/r/sheets/comments/178kf81/tracking_fi...). I also haven't put much effort in growing it, sadly.

I understand you might think these people are fake but I am not here to try to prove you wrong. I know what I've built, I know how many people have signed up and requested a copy. What I will do however is fix the landing page so that less people think this is some sort of trick to get people to buy.

In any case, I really do appreciate the feedback!

I still prefer Google Sheets somehow... Let's see how many people think the same

Thanks!

Wow, this is awesome! Love the setup.

And thanks for the kind words! You nailed it, it’s exactly what I wanted for people who don’t want to spend a weekend building a CLI just to track their groceries...

Sometimes a simple spreadsheet does the job!

If you inspect a page you will likely not see that there is some JS logic in the background that polls from a DB.

But will try to make it more convincing!

I tried a lot of things, I even work with AI professionally.

This is just a side project I had since 5 years ago. And it just reminded me how simple things can be sometimes. No need to put AI everywhere.

In any case, the reviews are not fake, I talked to these people personally in search for feedback. I am not going to share their personal info ofc.

But will consider improving the site to not make it feel fake!

Hope you like it!