Friday night musings.. These newly molted bots are connected to the Internet, they can say things to each other, access any website, pay in crypto, order stuff online and run non-stop in the background without human intervention.

This is hilarious, serious and future-shaping at the same time. A social network for ClawdBots!

a bit crazy but lately, whenever I’m thinking about “people”, I find myself automatically substituting that word by “cognitive agents”, which includes both humans and AI.

It’s truly astounding that the laws of our universe, its initial conditions and all the measured constants could have been so much different than what we discover, and most of the deviations would have created universes with little structure or complexity. And yet, here we are!

The most important question you should be asking while developing a product / startup is this: why would anyone change their behaviour to accommodate your product?

An unsolicited tip: if you want to stay ahead of the curve, read Hacker News daily. Replace your X/reels with HN and you'd tap into the most insightful + nuanced discussions on topics that matter across the whole world wide web.

A proposal to prevent job losses from AGI We may have only a narrow window before big AI labs automate away all economically useful work and centralize wealth. We’re certainly on that path right now.

As AI is rapidly dominating all cognitive domains, it’s futile now to compete with it or human+AI combinations. So, it makes sense to start deriving our sense of achievement not from being better than others (as that’ll be impossible) but being better than your past self.

I'm open sourcing whatever I learned while building habit coaching app (Nintee) in 2024. Starting with the science of habit building 👇

Our entire life is a journey to the present moment but when we get there, we want to be somewhere else.

I lowkey predict that people will stop asking frontier AI models dumb questions because they'll assume that the model will judge them for wasting its talent on trivialities. I do this now: chatgpt gets all the trivial ones, while Opus is for heavyweight explorations.

There’s a legit opportunity to start a dystopia tour in Delhi. Begin with a tour of the garbage mountain, then a cruise on the foamy Yamuna. After that ask for a deep breath in a park & then challenge them to cross a road as a pedestrian + have them hire a car to honk endlessly.

How I feel when I use that damn emdash in my writing!

It’s ironic that we now have AI systems approaching close to AGI but most people are still using them for trivial tasks such as drafting emails or queries that are easily Googlable.

This is the future of recruitment: beat an AI and get your job. Seriously, this take home assignment by Anthropic is a sign of times to come.

The biggest benefit of Claude Code is that you can now build software just for yourself. No longer do you need to adjust yourself around the tools you have. Now you can build tools to adjust to you!

Many entrepreneurs keep pitching me the idea of habit changing app. Well, I tried it (with Nintee) and it realized it'll never work!

if you're a teenager or young, your biggest strategic lever is that your cost of failure is basically close to zero. so use it to your advantage. take career risks. try building a startup.

It's public! @wingify and @ABTasty merge to create $100mn+ ARR company. Congrats to both the teams + Everstone; hope the combined business does amazingly well in the coming years! 🚀

Most good life advice boils down do this: know what makes you come alive. Write it down and then tilt your calendar, projects, and peers toward it. In short, design your life to be maximally alive!

Steal this idea. A chrome extension that looks at your recent chats with LLMs and pulls out a different, topical, relevant quote for you each time you open up a new tab. It’ll be a cute way to be aware of what historical figures had to say about your current situation.

What a marvellous #book on two of the greatest scientists in our history. It’s astounding how they overturned the accepted wisdom of the day inherited from Newton - that forces acted at a distance.

This is obviously an extrapolation of the currently accepted model lambda-CDM, but we have several alternatives to it that indicate a different fate.

All direct routes to greatness are profited away as everyone chases the same obvious path to it. Greatness then isn’t a destination; it’s what happens along the way when you’re exploring your curiosity and having fun operating at the edge of your ability.

Within a few years we will come to a stage where anything that’s solvable either an AI system will be able to solve it or show us a detailed execution plan for it. So, start cultivating a mindset that anything can be solved but choosing a problem to attack is where the alpha is!

What’s the future of science when a Claude Code like system can churn out a decent empirical paper in under an hour? I think this deserves a discussion. Academic output is getting inundated already with LLM-accelerated papers. What should human scientists pivot to?

People talk about gratitude journaling a lot, but not a lot is talked about wins journaling. Every now and then, you should sit and document all the wins in your life, no matter how small or big, in both personal and professional domains.

Will we forget what pre-LLM era looked like? Was chatting with a friend yesterday and it hit us that pretty soon we will forget what it was to live without LLMs. It’ll be inconceivable to us that there was a time when if you had a question, you did not get an intelligent instant answer for it.

Are we headed towards a stable dystopia? I've been thinking about how AI is going to shape up society and one particular scenario seems like a high-probability outcome to me. Here's how it could play out: • AI automates more and more of human-work

Evolution gave us the gift of abstract thought, but along with that came the poison of existential dread as a free rider.

The trick to getting AI to work for you is to break down your problem into specific, smaller tasks (ideally into units that don’t take more than 10 minutes for a human).

Here’s the thing with happiness: you can’t accumulate it, you can’t bottle it. Everyday you must start anew: wait for that dopamine rush that reaffirms that life is beautiful. Or.. you can view happiness as a skill and practice it until it becomes your second nature.

Pragmatism, at its core, is pretty simple Only talk about things that you can observe in the real world and find (creative) ways to verify them. Everything else is a mental-construct, and hence your word is as good as mine. What should settle the debate is observables.

I wonder if Claude Code or a tool like it will become the de facto way of doing knowledge work. It’s really like having a smart virtual assistant with full control of your computer and access to Internet - wouldn’t it replace how we work today?

I love how we go about life not knowing what we’re upto, yet pretending to everyone that we have a damn fine plan.

Steal this idea. A Claude Code wrapper that goes line-by-line and module-by-module to explain the code and reasons behind it, plus also allowing you to intervene / comment at any level.

I think in addition to measuring AI capabilities, we should also start measuring human+AI capabilities. Only when we pay attention to what teaming up does will we start optimizing AI to uplift humans instead of replacing them.

If I had to choose two ultimately difficult questions, I’d pick the following: 1. Why do the physical constants and laws in our universe seem finetuned for complex phenomena like life? 2. What are the right questions to ask about consciousness that can allow us to approach it scientifically?

Give someone a lot of free time and they start wondering the meaning of life, which quickly precipitates to a full-blown existential crisis.

Are interactive essays the future of education? I asked Claude Code to build an interactive essay for PCA + SVD that works client side. Notice how python code executes and graphs are generated WITHOUT any backend! Play with it yourself 👇

Writing is sensemaking. So when you delegate writing, you also delegate sensemaking; over a period of time, it makes you feel like you know a lot without actually knowing much.

Whenever you’re unhappy, blame it on external factors: lack of sun, pollution or the news. Whenever you’re happy, internalise it to your identity: well you’re a happy person, after all! Be positively delusional when it comes to your mental health.

Learned something very interesting today! Random projections of a non-linearly separable data onto high dimensional spaces is enough to make it linearly separable.

The crazy thing about compounding is that looking back it feels like it can’t get any higher, but looking forward you’re just getting started. It’s never too late to pay attention to things that compound.

I think we will soon see a breed of software with Claude Code (like) SDK integrated so that users can directly add new functionality themselves instead of asking the developer. Just imagine how cool would it be to modify any piece of software that you use at your own will.

A simple mantra to live your life: know what’s true and do what’s right. (Of course, teasing apart the words “true” and “right” as they’re applicable to “you” is the whole point of life)

It’s a great comfort to realise that, no matter who you are, the marginal impact of your work on the future is close to zero. Most outcomes are inevitable in the long run - if you don’t do it, someone else will.

Happy New Year! To kick off 2026, I built this tool that recommends all evidence-backed health screenings you need to do in your life, and lets you add google calendar reminders for when you're in 40s, 50s, 60s & beyond longshot.invertedpassion.com/ Share it w/ friends and family 🙏

What a phenomenal #book; I’m sure I’ll re-read it several times in my life. I have been on a journey to sharpen my intuitions about truth, reality and pragmatism, and this book delivers! Some notes:

People are sick of glossy, attention-grabbing UIs. With LLMs, we now have a chance to reinvent desktop. Replace GUIs with TUIs: make terminal-based apps that accept open ended text as input and output beautifully formatted text. Perhaps even browsers that run inside terminal?