Chandra Ramanujan

One thing I really love about working at @wayground.bsky.social is the product philosophy around the use of AI. They’ve articulated it really well in the rebrand announcement

One trend I’m noticing is AI tools make it easier for designers to really think through & prototype at a very high fidelity for data intensive experiences. I wrote about some examples I noticed in my blog. Would love to hear your thoughts:

nchandrasekharr.github.io/vibing

Thanks @ken.cv - fun little easter egg in our student login where after a while this little guy pops up and follows the cursor around (and hides when the user is typing in the password 💜) - blog post on implementation is coming soon

I've been playing around with cursor for a while. I've put together a basic visual to help myself figure out what it's best for, from POV of designer who wants to make prototypes / quick proofs of concepts. Hope this is helpful for you folks also.

I’m surprised companies want to automate the fun creative stuff that people actually want to do instead of all the nonsense that they don’t..

Or take my camera SD card, search for all images of me & autoformat it.. tag my screenshots in various attributes so I can search it later. I can think of so many repetitive brainless tasks that get in my day.

I had to take some few dozen bills & enter details in a spreadsheet for reimbursement. I just put it all in Claude & it formatted it for me. I wish an AI-native OS would do this kind of data extraction & formatting natively. For eg I select in file browser hit right click & share what I want done

These are just Sunday evening thoughts because I have had too much coffee and my brain is just going brr. But I would love to hear what people think.

Imagine I want to design an android app so AI generates for me a figma just for android, where I can pull up android native libraries and APIs to prototype an app.. where I am designing and the output is directly being expressed in flutter or kotlin. Or HTML and CSS..

In the earlier case, I wanted to make a UI for generating truchet tiles. Then I made a UI for swapping out patterns. Then I made a UI for selecting colours. Then I made a UI for similarity. I can go on and on..

Most design tools are deterministic where output is dependent on manipulation of fixed number of different parameters of a particular entity. What if I can increase parameters and create new ones? But I don't want AI to just make things for me. I want it to help me think about new things I can do

image of a user interface. on left is truchet pattern, on right is a range of sliders of different parameters with AI suggestions for new ones

I think one thing we are not thinking about in AI tools for designers is fluidity and emergence of controls where AI doesn't just generate outputs, but helps expand the possibility space of what can be created

Weekend project - a truchet tile generator, which creates random arrangements of tiles in colours, which can then be downloaded. I also tried to make a "similarity" slider to see how clusters of similar colours emerge. Built with help of Claude.

As a PR pro, we used to tell people to get ahead of various online controversies by making statements that refute them. It turns out this is the absolute worst way to respond to criticism or correct errors of any kind. Social media has made this worse. When you respond, it kick-starts issues again.

I think it could be an interesting AI usage pattern. Instead of prompting to generate something high fidelity from scratch in one shot, you iteratively prompt your way in low fidelity to a better starting point then take over manually from there. so it feels more like partnership then replacement

Fun idea, if you are stuck on the high level layout / IA for a page, you can prompt claude / chatgpt to generate options for you in ASCII as a very rapid idea starter. and then you can layer on levels of fidelity.

here is one for B2B SaaS and a couple of explorations for a food delivery mobile app

I prompted Claude! I kind of like how it's responding on this

all art is made from spite or horniness, and a computer cannot be spiteful or horny, so a computer should never make art

Replit is his portfolio company and job as VC is being a hype man so that more people use his company’s product and he gets a return on his investment.. it’s how I am reading it

Yeah I doubt this. I think it replaces prototyping and generation of quick proof of concepts but overall design definitely not..

How he described it is - edtech startup hits X ARR / MAU and then a pvt equity like renaissance or publisher like Pearson comes and makes an offer. post acquisition the focus moves from product to sales and marketing. Over time the product withers. They buy another product and start the cycle again

I was talking to my founder about how all companies who reach a certain scale in edtech get acquired by these PE behemoths or large publishers.. and there are few cases of standalone companies that are making money and at a certain scale. I suspect such a dynamic is true for a lot of industries

I feel like there’s also a consolidation around early adopters which have network effects and distribution.. It’s not so hard to build the product (it’s still hard) but how do you go from there to getting a large user base that you can make money sustainably from…

The implication of this is that it requires something more than craft of design and hype from early adopters to create a viable company. And the circumstances in which these companies raised money or operated no longer exists, and it’s unviable to operate.

As a heavy read.cv and posts.cv user im not happy about this. But congrats to the team on a good exit. I noticed a lot of design founded companies like cron, campsite and now read.cv are selling out or shutting down. Even arc browser is sunsetted..

If you want to make progress with complex design problems, try creating hypotheses Define all aspects of a design problem to make it easier to solve

Working on complex topics can be intimidating, especially to get started or to make sense of things.
This is why you should rely on hypotheses: it’s a bite-sized, standardized way of breaking down the unknown into something that you can design around.

I just tried this and the spoofed UI showed up for me. This is so funny omg. .. what are they thinking 😭

In the future, I would expect to see tools where I've set out autolayout rules, component props, responsive behaviour etc in Figma, I can define any business/ domain logic, and then I just copy paste it into an AI and it turns it into a working front end.

This is something I expect to use every day. The cost of getting simple software working is going to zero. As a designer, being able to code at a high level should no longer be the constraining factor to ship my own tools / products. This might also mean software will get a lot more personal.

Screenshot of a user interface of a habit tracking app. The right column shows the date, and each column corresponds to a habit the user wants to track

Claude by @anthropic.com built a very simple habit tracker for me, that lives in my browser and uses local storage to keep track of entered data. I can also export data. There's a lot of work to be done in order to style it better and improve it visually, but I am very impressed by it