F2F #76: Building habit - agentic browsers
Around three years ago, the world turned upside down with the release of ChatGPT. This weirdly named app took the business world by storm and changed how we work and live. I decided early on it would be the thing, so I forced myself to build a daily usage habit.
Around three years ago, the world turned upside down with the release of ChatGPT. This weirdly-named app took the business world by storm and changed drastically the way we work and run our lives.
While I wasn't one of the first users, I was also not the last. I am fairly sceptical of new trends and I always give them time to settle, to see if they're just a passing fad or even if they drop the prices (hey, I am Catalan, after all).
This time around, I remember thinking "this stuff's gonna go big". It had all the right conditions: hyper-growth despite a lot of constraints: clunky interface, web only, waiting list, hallucinations galore, lots of downtime, a terrible name... and in spite of all of them, it grew and grew.
I am a big fan of things that thrive with lots of constraints because they will grow even more when you remove those constraints.
For instance, at MarsBased, we started doing only web development with Ruby on Rails and Angular, we didn't do mobile, we didn't do other programming languages, we didn't do ads nor outbound, we were remote only, we were more expensive than the competition and we said no to a lot of projects. We received plenty of projects, and we grew even more when we started removing some of those limitations, one at a time.
Another example: Five years ago, I started a talk-show podcast with my good friend Marc Collado, called Foc a Terra. The show was audio only, no guests, no script, no sponsors, no ads, exclusively in Catalan and we only hosted it on Apple Podcasts. Despite those constraints, the show grew to thousands of listens per episode. Now, we're more flexible on all the above and it's become the #1 business and technology show in Catalan, drawing in very big names of the industry as guests.
But back to the topic: because I knew that ChatGPT was going to be a thing - or the thing - I decided to spend 30 to 60 minutes every day using it for tasks that I could do otherwise. It didn't matter if I could do them faster without ChatGPT. The point was to build the habit.
After three months, more or less, I had built the habit. I reached the point where I went to ChatGPT the first thing when I wanted to tackle a task.
I did the same with its Advanced Voice mode, as I discussed on F2F #29: Talk to me, bAIbe. Now, I do loads of stuff using this mode because I have acquired that habit.
Now, it's the time of agents. Some will say that agents have been around for a while, and that will be also true but I didn't have the time to focus on getting the basics. And while I will have to wait for Perplexity Computer to see if it's a passing fad or if they lower prices, agents have been around for a good while.
I even tested OpenClaw a couple of weeks ago without much success, but I got a glimpse into the future and I realised that it was only going to work for me if I built the habit of using frictionless agents. No installation, no complex permissions, no setup.
I've been trying Perplexity's agentic browser, Comet, for a good while and I always forget that it has an Assistant mode (cmd+A). This assistant is the agentic part of the browser.
Now, I am forcing myself to spend 30-60 minutes per day giving it tasks that it can do for me.
Here's the ones I've been able to do in the first week and a half using an agentic browser for one hour a day:
- Find discount codes for my online purchases
- Tip: you have to tell it "codes that work" so it tries them for you.
- Delete old tweets from my account.
- Check all my followers on Twitter and block bots.
- Find me providers for bus companies and contact them asking for quotes for an event I'm hosting.
- The provider I ended up choosing came from this search.
- Scrutinise all the pages on the MarsBased website, looking for broken links.
- Spellcheck and correct blog posts (including this one).
- Do edits to my google sheets documents.
- Export the content of big tables as .csv or even in image formats.
- etc.
My usage of ChatGPT has greatly diminished over the last days because I feel like I'm not only no longer forced to go to their app to make things happen, but I also can do stuff here that's not possible on ChatGPT in the first place.
I should've started earlier using agentic browsers. But I've got one final consideration for you: don't use them for private stuff. Be very mindful of what you give it access to. For me, it's no email and no finance stuff.
Build the habit today!