Fragments

edition 9.

Hello.

It’s been a long long time. I’ve sat down to write this edition 6 times, and 6 times I have failed to publish.

Here’s a quick recap: in the span of 6 months I visited India twice, watched 3 cousins get married, ate my way through Bangkok, suffered another stomach bug (unrelated), went to my 2nd Anirudh concert, and moved apartments.

A new home

We said goodbye to our view of the river. Our new apartment doesn’t have one, but the river is only a 5 minute walk away. Turns out I can’t be too far away from it. We have a view of the street, an intersection next to which is a large red maple tree (I hope), and in the background is a hill with colorful little houses on it, shrouded by more wonderful trees. The houses don’t look real from here, but I imagine little people in them looking through the windows. I love it all the same.

As hectic and physically taxing (not so much for me as for my husband and his amazing friends) as the actual process of moving is, I’ve come to enjoy the aftermath. The unpacking and organizing. Along with your home you get to arrange your life anew.

It’s a change that feels like how I imagine the arrival of a new year should - or maybe I’ve just moved every year for the last 3 and so I’ve been moulding them into our own new years, creating new habits and moving old ones around into routines that better fit our new home.

We’ve been able to settle into a routine I love, which is only punctured occasionally by travel and sickness.

Gathering my thoughts around AI

I’ve been experimenting with a bunch of different things and it took a while for me to get back into reading (more about everything I’ve read below) and writing. Instead, I spent a lot of time talking to AI and exploring what it can help me with. I wanted to gather my thoughts here around how I’m currently using (or not using) AI (my tool of choice is Claude). Here’s where I’m at right now.

I like using AI to:

• Teach myself very specific concepts in relation to something else

  This is most useful when I’m learning something new and I come across a foreign or forgotten concept. For example, when I’m looking into something for work and see a topic I need a refresher on, I will ask claude to explain the concept to me using a simple example and that’s usually enough to understand the relationship between the concept and the larger picture I’m working on. Since it has context on my role within the bank it will always end with how this topic will impact my specific function, liquidity risk. Bridging gaps in my understanding like this has helped me stay in flow while learning.

• Build/automate things at work

  When I needed to build a powerBI dashboard in a limited period of time, I designed how I wanted it to look and then asked claude very specific questions on how to implement my design. This turned out to be a quick way to learn how to use the tool itself.

• Build/automate things in my own life

  I despise meal planning. I find it tedious and draining. So I built a little web-app that helps me plan out our meals every week. It saves us about 30 minutes every Sunday but more importantly, I no longer dread having to do it. Plus I had so much fun designing and building it out. I used claude code for all the implementation and back-end functionality (this needs a lot of work and it’s been helpful having a background in CS). I’ve found that when you provide it screenshots and have it implement one screen at a time it’s fairly accurate, plus it nudges you to test every screen and work on fixes before moving on to the next. I wouldn’t try to one-shot the entire thing.

• Plans for itself

  AI is great at creating to-do lists and plans for itself to follow. This is something I will always do when I start a project - I’ll ramble to claude about the idea, how I want it to look, tools I want to use, what I want to accomplish first, what I want to add later on, and ask it to turn that into a plan instead of me having to type everything out neatly by hand (I use wispr flow for large, detailed prompts). Claude creates a comprehensive plan + checklist, and will update the plan as it goes. This saves me so much time and also saves all the context I had for the project when I started, preserving what I wanted to do with the idea like a snapshot of my thinking.

I do not like using it to:

• Come up with ideas from a blank slate

  I do have to admit I’m kind of cheating by putting this on the list, because the more I think about it, the more I’m convinced this is just bad practice as opposed to a bad use case. I’ve tried this more than once and it has never resulted in anything remotely interesting and I should never have expected it to. I think in order to get to a workable outcome you need to have curated a map of ideas you want to choose from and you can then begin to extract a direction. In which case, you need to do all of the thinking and then use AI to shape that into an idea. I’ve seen people do this with brilliant results. When dealing with a blank slate, I much prefer writing or taking a long walk to arrive at the thing myself.

• Write

  The whole point of writing for me is to solidify my learning, understand my thoughts, find the holes in them, and move towards some kind of clarity. It’s a struggle and almost always results in terrible first drafts but both the struggle and the drafts are mine and they will never not be strangely satisfying.

• Curate an experience or a list of any kind

  I would much rather ask people or find other curations by a human being on the internet. Curation is a matter of taste, on which the discourse is vast here on substack. I don’t feel the need to add to it. I simply do not like AI lists.

This will exist as a living document that I’ll update as both the tools and my thinking change.

---

reading:

The last thing I read in 2025 was Whoa vol2 - a lot of writers I love reading chatted about AI and creativity with Alex Dobrenko and I’m just a sucker for spiral bound handmade things.

This year so far I’ve read:

The Design of Everyday Things by Don Norman - this was new ground for me and gave me the vocabulary for design

Solaris by Stanislaw Lem - I really wanted to love this. I found it interesting but the long sections on Solaris exploration were hard to get through

Weather by Jenny Offill - the atmosphere of the book really embodied the anxieties of climate change but the characters and prose fell flat

Debt by David Graeber (yayyy) - still thinking and writing notes around this to form an opinion. I want to read more perspectives on money and the beginnings of finance

The Ministry of Time by Kaliane Bradley - this was a time-travel sci-fi thriller. It’s fast-paced, has wholesome friendships, and I thought the narrator’s back-stories gave the book texture - it was so much fun to read

The last 3 I finished in May.

I’m currently reading What We Owe Each Other by Minouche Shafik and A Horse at Night: On Writing by Amina Cain.

I really want to read more fiction this year, I read 10 books last year and none of them were fiction. I miss the feeling of sinking into a book. I miss never wanting to put a book down. I want more of that. If anyone has any recommendations please send them my way.

learning & making:

I’m working on a website for a friend’s non-profit that I hope to publish sometime in the next couple of months (if madam will give me the copy but I fear public slander will not work).

Like I said above, I’m also working on Omnom, a web app that creates a grocery list based on the meals I’ve selected for that week. I feel weirdly accomplished building something I actually use. Waitlist page coming soon :)