This is less about tools and more about how my relationship with AI changed. I’m basically answering a few questions here. - What new tools did I start using this year? - What were my ‘aha’ or the ‘holy fuck’ moments? - What experiments did I try? - What's my view on AGI (Artificial General Intelligence)? - What’s up with my ’26 and what I think of it? What new tools did I start using this year? NotebookLM This honestly felt like magic. I consume quite a chunk of information every day. Substack posts, Medium articles, research papers, long PDFs, and deep research docs generated by AI on specific topics. What I started doing was simple. I take that document, convert it into a PDF if needed, upload it to NotebookLM, and tell it how I want the information presented. A few minutes later, I get a podcast. Two people talking through the material. What makes it wild is that I can interrupt them and ask questions in between. It feels less like passive listening and more like an interactive briefing. try this out: If you are going to an event focused on a specific topic, ask ChatGPT to generate a deep researched document on that subject. Upload it to NotebookLM. While traveling to the event, just listen to it. You arrive already warmed up. Agentic Browsers These were genuinely interesting. Normally, if I land on a website I have never interacted with before, I have to manually navigate it, figure out where things are, click around, and decode what matters. With agentic browsers, I just tell the AI what I want from the site, and it navigates the website in front of my eyes. Another use case surprised me. Sometimes I want to replicate something I see on a website or extract elements for a side project. I am not always good at verbalizing what I am seeing. Instead of struggling to describe it, I ask the browser agent to describe it for me. Design, layout, video, music, anything that exists on the page. Once I have that description, I tweak it and hand it off to a coding agent or ChatGPT and continue the conversation from there. AI voice transcription I was never happy with the built in transcription on iPhone, Mac, Windows, or Android. I use all of them. The problem was on both sides. On my side, I stammer. I repeat words while thinking through the next sentence. On the system side, the transcriptions had no punctuation, no paragraph breaks, and words were often wrong. I hated reading them back. AI transcribers fixed this almost completely. The stammering and repeated words are gone. The output comes back as clean, structured paragraphs with proper punctuation. It finally feels usable. Something I am actually comfortable saving in my notes or sending as a prompt. Coding agents This was a different beast altogether. I tried Cursor, Antigravity, Kiro, GitHub Copilot, and Claude Code and a few more. I used Cursor for a few months and eventually settled with Claude Code. it’s worth every dime. I never really knew how to code or build applications in the traditional sense. With coding agents, that stopped being a hard blocker. I describe what I want. A feature, a button, a scroll animation, how data should be displayed. After that, I usually ask how we are going to implement it. Once we talk through the approach, I just say go ahead and implement it, and it writes clean code with the logic in place. What surprised me was that I could do this from anywhere. Sometimes it runs locally on my laptop. Other times I do it from my phone, where it runs on a cloud server. Either way, building applications suddenly feels approachable and even fun. That said, I do hit roadblocks. Since I am not a professional developer, when something breaks, I spend more time debugging than an experienced coder would. Those moments are also where I learn the most. When an error shows up, I ask the agent to point out exactly where the issue is and explain the snippet line by line. Once I understand that part, I look at the broader structure and move forward again. It is not frictionless, and it is not cheap, but for me it has been worth it. Notion AI Notion has been my second brain for about four years now. I take a lot of notes. I started using Notion AI last year mainly for editing and grammar checks. This year, it crossed a line. Now I can just describe something vaguely like “I think I wrote something about comparing X and Y” and it finds the exact note or document. That alone is a massive time saver and strangely calming. No more digging through folders. On top of that, I can ask it to create databases, reorganize information, or restructure entire sections just by describing what I want. It feels less like a feature and more like a layer over my thinking. Poke I used Poke for a few months. It is an AI that sits inside your messaging. It responds the way you respond, picks up your tone, and maintains your vibe. It connects to email, Notion, Slack, Instagram, and more. It gives updates and lets you build automations for scheduled tasks. It made sense for me to use, but economically it's a hard pass. $50 a month is i think a lot for this. what were my ‘aha’ moments? Music generation with Suno. I was skeptical about AI generated music initially. But I wanted to try it out so I thought of having a theme for a song and writing lyrics for it. I took the theme to Claude and asked it to come up with some lyrics. It did, but I didn't fully accept them as proper lyrics, like if this song was supposed to be made manually, I wouldn't go with these. But since generating music only takes a few minutes, I went with it anyway. The magic was in the description of the music. How would it start, how would it flow in the middle, how would it end. I put all of that together into Suno and my mind got blown with what it created. Vibing an application Creating applications on a whim. Like say I wanted to learn about something, or present something, I'm writing whatever that I need from it, talking to Claude for a bit about the features that I'd want and asking it to generate me a perfect prompt including all those, making some tiny changes to that prompt and giving it to Lovable or Bolt (I got those both for free for a year) and making an application and using it for like say few hours and discarding it. What a fucking time! Video from images and text. I take a picture, give it to AI, tell it I want the person in the picture to do so and so, and I get a video of that person doing exactly that. Fuck, that's an aha moment. Isolating subjects in pictures like selecting a subject and isolating it in a picture. Simple but crazy when you think about it. Generating podcasts on the fly & interacting with them. I give NotebookLM a PDF of an article or a research paper or some deep researched doc that AI made for me and tell it how I want the information presented. Few minutes later, I get a podcast that I listen to at the gym and if I have questions in between, I'd pause and ask the questions. Scheduling Deep researched tasks. I'd like to read about few things before getting down to a task and usually I'd surf the internet and get the info. And now, I schedule a task and set a time for the report to get delivered to me every week. Or like if I am attending an event, I'd like to know as much as possible and I'd set it in a way that I get a report delivered before going to the event. This is sick! lil reflection on those aha moments: when you get to that aha.. this is crazy moment, it's not the time to get over excited and think about crazy things but rather think slowly and clearly. but there's a wide gap in all these aha moments. like it used to be when we have that aha moment, it is when we figured something can be done differently in a way simpler way and we'd be using that technique always and the way that we do things would change from there on. but with the aha moments right now, like these, say if you had an aha moment when you gave a text prompt and you've got a good video and when you asked it for changes it did work out and you'd be thinking of a wide range of possibilities from that one aha moment but then you'd hit the road block. like the next time you're trying to make a video of your imagination, it might not work or it would not be totally satisfactory. there could be different reasons for that like we'd generally expect better things than they were previously and also, one way of doing things might or might not work for the next situation. ai is hella different. What experiments did I try? Using ChatGPT Projects as a tracking partner I started using Projects inside ChatGPT like folders in my brain. Different chats for different purposes. One for schedule. One for gym and food. One for random thoughts. All of it living inside one place. The schedule one came from a real problem. My week would get scattered. I’d get carried away with one thing, get distracted by another, and then at the end of the day I wouldn’t even know what I actually moved forward. So I told ChatGPT, these are my priorities, this is my gym time, this is my reading time, this is my cooking time, this is my work, now build me a weekly plan from the time I wake up to the time I sleep. Then I started checking in twice a day so it could keep me honest and keep the week from drifting. AI as personal trainer I also tried using AI as a health and fitness tracker. Not in a motivational quote way, but in a “log the reality” way. Gym days, swim days, meals, groceries, what I cooked, what I skipped. I don’t usually take personal training. I feel it’s a lil expensive. So instead, I used AI like a lightweight coach. I’d report what I did, what I didn’t do, and it would help me see patterns, adjust meals, and tighten the routine without making it a whole production. Capturing random thoughts so they don’t hijack my day This one was simple but weirdly powerful. When I’m trying to work, my brain throws random ideas at me. Side projects. Big questions. Random theories. Stuff that feels urgent in the moment, and then steals my time. So instead of following those thoughts, I started dumping them into a separate chat. Not to get answers. Just to capture them. I didn’t want AI to “solve” the thought. I wanted it to store it so I could look back later and see what kind of patterns my brain is producing. I put these chats into a folder and treated it like a mental junk drawer, but organized. Shaping a thought over time I also tested something more deliberate. Using AI as a thinking partner over weeks. I’d tell it something like: I have a thought, it’s scattered, and I want to concretize it. So whenever I notice something related to it, I’ll come back and drop it here. Track the pattern. Help me connect the dots. And when I’m done, we’ll put it together into something clean. It works because I don’t need to hold the whole thought in my head all day. I can offload fragments, collect them over time, and then assemble them when I’m ready. Reading with a live voice chat running This one felt like cheating. When I read, I get a million questions. If the book mentions a time period like 1956 to 1972, my brain immediately wants context the author didn’t include. Normally I either ignore the questions or go down a rabbit hole. So I tried reading with a voice chat open. I’d ask questions as they came up, get quick context, and keep moving. The tradeoff is time. A 30 minute reading session becomes 45 minutes because you’re also talking. But the upside is you actually understand what you’re reading, instead of just finishing pages. What’s my view on AGI? AGI is usually defined as an AI that can operate like an expert human across different domains, reasoning, deciding, and acting the way a human would. By that definition, we’re still far from it. Not because models aren’t smart enough, but because intelligence alone isn’t the missing piece. What we have today is reactive AI. You open a tab, type a prompt, get an answer. The system knows nothing about you beyond what you just told it. AGI won’t look like that. It won’t be a model sitting inside a chat box waiting to be asked questions. It will be deeply embedded into everyday life. It will know what you’re working on, what you care about, the goals you’re moving toward, the tradeoffs you’re struggling with. Not in a creepy surveillance way, but because you’ve built that context over time through repeated interaction. Right now, we go to AI. AGI would already be there. It wouldn’t wait for instructions. It would understand what matters, anticipate decisions, and act in alignment with your intent. The baseline for AGI isn’t just raw intelligence or better benchmarks. It’s context. An AI that can answer any question but knows nothing about you isn’t AGI. An AI that understands you deeply and can act on your behalf with your goals in mind is much closer to what AGI actually means. What’s up with my ‘26 and what i think of it? 2026 is the bet for proactive AI. So far, most AI has lived inside a chat box. You open it, ask something, get a response. That model is reactive. It waits. What’s coming next is different. AI that does not wait for prompts. AI that understands context and takes initiative. Systems that notice patterns, anticipate needs, and act before you explicitly ask. On a personal level, I want to spend more time building instead of just trying tools. Writing more consistently. Doing deeper research. Reading books on history, technology, and psychology to widen the lens. I’ve already noted a few real scenarios where agent orchestration actually makes sense. Things where multiple tools, memory, and scheduling come together. I plan to experiment with that using n8n, Make, OpenAI’s Agent SDK, and a few others. I’ve also started scheduling long-running research tasks in Perplexity and letting them run in the background. There’s a lot I want to explore. Different AI SDKs. Fine-tuning models. Reading research papers on agent architectures and system design. Not just what these systems can do, but how they are built and where they break. and this feels like an important time to sit tight and learn a shit ton. and of course i mean learning is by building something like working on some random applications, understand the mechanisms. but of course, the landscape is changing rapid fast and would course correct myself accordingly.