Writing

The beautiful thing about using agentic AI to co-work is that the limits simply seem endless. It feels like you can achieve anything you want within a day of diving deep, which is an incredible feeling. We are talking about things that would otherwise take months to understand and create.

Lately, I’ve started to fix the things that have annoyed me on my Mac for longer. The following are two examples, there are many more. From Obsidian Plugins over to Terminal CLI’s.

There are webpages that simply do things that you don’t want them to do, or look a way you don’t want them to. Claude Code is incredible at creating Chrome extensions and those are incredibly powerful if utilized in the right way.

People don’t like Artificial Intelligence. Artificial Intelligence spreads fear and worries about Jobs and the future.

What people do like though is what AI does, how it helps them. They love interacting with Artificial Intelligence, using ChatGPT, Claude and Gemini, but they hate that it exists.

And ads for AI companies should reflect exactly that. They need to show how people can do more, learn faster, build higher, discover more, and push the world forwards by using such tools.

Everybody does different things on their Mac, so their configurations are naturally different from one another. I do a creative combination of interaction design, web development, all-things development using AI tools, and studying Computer Science. Here’s what’s on my Mac, currently. The things I left out are things I couldn’t think off, meaning they have to be irrelevant.

If you get some inspiration from here, or have inspiration for me, let me know!

If you’re changing browsers a lot, this makes things easier than navigating though System Settings

OpenAI is introducing ads on ChatGPT Free and Go tiers under four pillars. The biggest one is that ads do not influence the answer AI gives you. The thing is, it does exactly that. Just not directly.

Yes, in an ideal world a ChatGPT ad at the bottom of the page has nothing to do with the content above, it’s a random display ad like in any cheap mobile game. But OpenAI is doing the exact opposite. Information about the user is not shared with the advertiser, but the advertiser is promised that ads will be shown to users who fit the profile best and are most likely to convert, based on the current conversation. If you’re thinking about what to do for dinner, you might get an ad for groceries.

And people have no idea how much OpenAI knows about you. We think Instagram, TikTok and the like know all about you to tailor the algorithm, but people simply forget what they feed daily into ChatGPT. Every little prompt tells a whole story about a totally different aspect of your life. Point is, OpenAI knows where you work, about your family, your medical issues, your interests, friends, values, motivation, depression, thoughts, projects, your lifestyle, and from that everything else follows. How your mind works, how you think, how you act under pressure, what softens you, what your trigger points are, how much confidence you have, how much structure you need.

Simply following promises won’t work here. Siri simply being able to do what they promised won’t cut it. The thing with AI integration is, it’s easy to do something, and very hard to make it magical and right. Apple’s current form of ChatGPT answers is horrible, but Google’s Gemini integration in Android doesn’t work either.

It’s a large language model, designed to write text, that is also given the ability to call certain tools and set timers, etc. But it is slow. It feels less as an assistant on your phone and more like a ChatGPT window you can slide over. Less an assistant, more ChatGPT that can control the system.

Ask current Siri to open the camera, and she will do so. Ask Gemini to open the camera, and it will process your task, and then tell you that it will open the camera, and then, only then, open the application.

If you’ve spent time chatting with large language models (LLMs) like ChatGPT, Claude, or Gemini, you’ve probably noticed that the conversation flows linearly. You ask a question, the model answers. You follow up, it continues. This might work for quick questions, but the thing is, thought and research processes don’t work like that.

We need the ability to branch off. Ask a question to a specific part of the answer of the LLM, and get the answer without being automatically scrolled down to the very bottom of the conversation which means you need to scroll back up to were you were.

Conversations with AI shouldn’t be a single thread, but a branching tree. I should be able to select a part of the answer that I don’t understand, pose a question, and get an answer in a split screen next to the main, linear thread.

It’s genius. Here’s some thoughts following on DevDay 2025.

Through the new Apps SDK, all SAAS companies are required to build an integration into ChatGPT, because if they don’t, they’ll just get completely replaced by the competition. When creating a presentation, people already spend a lot of their time in ChatGPT to assemble the contents. They used to copy-paste them out into a presentation tool, but when they can generate, edit, and export a Canva deck right on ChatGPT, there’s no need to ever visit that website.

It’s an absolutely scary, but genius move. Sooner or later, a “normal person” (meaning non-dev or so) will be able to spend 100% of their time on their computer in ChatGPT. Wait till they add a nice Music player, Video (Netflix and co.) player, and Email UI and invent a new kind of messaging service (I have some ideas that I’ll share later).

All of these launch videos follow a similar pattern, and they are absolutely . They leave you in awe really wanting that product, and I can’t describe how and why. It’s a relaxed atmosphere, elements of nature and everyday life combined, while showing the product.

Have another one that fits the mood? Send me an email, I’ll add it!

Say hi to https://t.co/VIWYU64dUI! 👋🏼🌴 pic.twitter.com/gFoZupfqZE

Video evidence is inadmissible now. It used to be proof something happened, or not. And well, no more. We need a new way of proving what’s true and what’s not and how this’ll look like? Not sure, yet.

The appearance of LLMs marked the start of artificial content. Back in 2022, when ChatGPT launched, it felt like a niche thing for tech nerds to play with. Some of us (including myself) already had an OpenAI account and were using the GPT-3 API before ChatGPT even existed. Either way, it took a year for people to understand what this meant, and another year for people to figure out how to use it to actually get work done. One thing always stood out –

We’ve accepted that we don’t know whether what we read is human- or AI-written. We’ve accepted it to the point where we don’t even question who wrote the latest Economist article, because to a degree, it doesn’t matter. Someone thought of the content and somehow got it down on paper. We’ve accepted that if an entire newsroom article announcing GPT-5 is written by an LLM, it doesn’t really matter. The beginning was hard. When ChatGPT launched, people were freaked out about exactly this – and now we accept it.

With announcement after announcement, one thing never changes: I love technology. Because it’s a field that never stands still.

And just when we think things are stagnating, we’re onto something new.

Right now, you can feel that energy. LLMs (besides being the revolution themselves) inspired new products and companies to launch things that could’ve been done before as well. Take Taya for example, a necklace you talk into, which transcribes and saves your ideas and thoughts. Absolutely beautiful.