Indradhanush Gupta

Indradhanush Gupta

@igupta.in

Staff engineer, building Kubernetes for Private Cloud. I will absolutely automate anything and everything. I blog about tech and human behaviour at igupta.in/blog recurse.com alumni. He/Him.

1.1K followers250 following376 posts

In other travel related news, I’m pleased to inform you all that I am now bold enough to ask the passenger next to me to tuck in their elbow. Because I realised that this is not an awareness problem. It’s that they don’t care that their elbow is hanging out and is making you uncomfortable.

Guy: Sorry (as he tries to sneak past) Me: No you’re not. Guy: looks up, perhaps confused Me: I was here before you Guy: makes a dissapointed face and stops

And this extends to in general all people who are doing something wrong. They know. They always know. So something like, I was just following orders shouldn’t be an acceptable excuse ever when things come to light.

I’m in the absolute minority here, but emacs org-mode is way superior to anything markdown when it comes to plan files with LLMs. Individual TODO states for each headline. Tags when and where I want. And of course filtering based on state and tags.

View full thread

If I could crowdfund LLM tokens, this is what I would want to tackle. Build org-mode as a universal API. And campaign to replace markdown with org-mode.

When I talk about org-mode with my friends and coworkers, both former Emacs users and non-emacs users, I feel like a dinosaur stuck in my tools. And that’s a little bit sad.

I’m in the absolute minority here, but emacs org-mode is way superior to anything markdown when it comes to plan files with LLMs. Individual TODO states for each headline. Tags when and where I want. And of course filtering based on state and tags.

org-mode was a tool ahead of its time. And failed wider adoption because of what made emacs great: all of emacs is programmable. Which means building org-mode plugin for other editors over a core API meant you had to interface with elisp. No one wants that and for good reason.

View full thread

As you sit on the chair, momentarily relieved after having stood for the majority of the day. You feel a pinch. But think nothing of it. Ignoring it for the stiff back that is a more pressing concern.

You reach for the doomscroller in your pocket. And as you check your notifications, you see something about a bum. The internet is abuzz with it. You try to stand up going for a stretch. But you can’t leave the chair anymore. Your bum has been pwned.

I was considering adding a permanent disclaimer to every blog post on my site about it not being AI written. Then I relented. Because I don’t have to explain to strangers on the internet that I wrote it myself. I like writing, and even if it is imperfect at times I’m okay with that.

I can't tell who's teaching who anymore Discussion

View full thread

I'm trying to counter this by trying to read more fiction, but for every page that I read, I'm probably reading 50 pages or so of LLM output. So I think this is a battle I'm going to lose out on in the long term. And I can probably only delay the inevitable.

I was considering adding a permanent disclaimer to every blog post on my site about it not being AI written. Then I relented. Because I don’t have to explain to strangers on the internet that I wrote it myself. I like writing, and even if it is imperfect at times I’m okay with that.

I can't tell who's teaching who anymore Discussion

But I do catch myself editing my own words multiples times a day. Because I'm less sure if that sentence structure is truly mine or a side effect of me conversing with LLMs so much.

At the airport, queued up to get on the elevator with bags. Almost no one wants to queue up at all which is surprising. Guy comes in last minute and tries to get past just when I think I can get onto this elevator with my trolley full of bags. I’m proud of myself for what transpired next:

Guy: Sorry (as he tries to sneak past) Me: No you’re not. Guy: looks up, perhaps confused Me: I was here before you Guy: makes a dissapointed face and stops

This flight today has wifi. But I intend to not use it. Flights have remained one of the last ways to force you to be off the grid, and this is one feature I’m not too excited about. Let’s see if I can resist the evil machine. ~10.5 hour flight. Here we go.

Rule of cables: you’ll need it when you don’t have it. I always fly with a USB A to lightning cable. Off late flights also have the usb c port. Right now I do not have the USB A version. And of course this flight only has USB A. No power outlets either.

Flying SFO-LHR today on Virgin Atlantic and props to them for playing some really groovy tunes while we board. All new ones I haven’t heard before. Thanks to Shazam I now have them in my playlist.

Indradhanush Gupta reposted

NEW from us: India's rapid clean power expansion has kept power sector emissions flat for two years, for the first time in over 50 years, and sharply reduced fossil fuel imports. Oil and gas consumption have also fallen for two years in a row, helping alleviate the Hormuz shock.

Carbon Brief@carbonbrief.org

NEW – Analysis: India’s power-sector emissions flat for two years due to clean-energy surge ✍️ Written by @laurimyllyvirta.bsky.social and Anubha Aggarwal @creacleanair.bsky.social Read here: buff.ly/JIwsTKV

Increasingly feels like that some things just take a lot longer with AI now. I find myself fighting the machine and trying to get it do do what I want. And it takes so many round trips that it tires me out. And my work at $DAYJOB is critical infrastructure layer so just “trust it” doesn’t cut it.

Is this a side effect of perhaps me inherently expecting faster turnaround time now that I’m so used to the speed? The more complicated the work, the worse the output gets. So I can’t really go hands free yet.

This whole thread is one of best I’ve read that captures the nuance very well in the whole AI vs no AI debate.

Avdi Grimm@avdi.codes

My Bluesky feed is currently full of outrage about some game called "Outlier" using LLMs in development and - as someone who is genuinely extremely critical of GenAI in art, writing, anything creative: I have bad news for y'all who are using literally any software anywhere, including this site.

Indradhanush Gupta reposted

What if our ears were like our eyes, and only had three (types of) receptors? Interestingly speech is still more or less intelligible, but not much else to me this shows how much more discriminating of frequencies our hearing is than our vision. Of course our vision has much more spatial resolution

Indradhanush Gupta reposted

my car is throwing an engine trouble code, and I can just directly plug an intelligence into it and it will talk to the car computer and just figure it out the present is unbelievably cyberpunk

Penned this down as a blog post. A slightly expanded version of the original thought. igupta.in/blog/why-i-a...

igupta.inWhy I am no longer letting Claude Code add itself as Co-author in my commitsMy thoughts on why I do not want claude code as co-author to my git commits

Early on I believed we should be adding LLMs as co-authors to commits we make. My stance on this has now flipped. I think LLMs as co-authors makes us shirk responsibility for our changes. You wouldn’t have added Google or Stackoverflow as co-author previously. So you shouldn’t do it now.

Question to all the writers who still write themeselves and not use AI: Are you adding a big disclaimer on the top of your posts about not using AI for the writing? I'm conflicted and can't seem to make up my mind. How is everyone going about this?

The responses here on the why are quite intriguing.

Fibonacci Sequins@brentwaggoner.bsky.social

I wonder if Harry Potter still being so much in the pop culture zeitgeist has mostly to do with it being one of the last truly universal cultural touchstones before everything was so segmented by the internet.

So I was not the only one struggling with this non-sense. Thanks to Reddit, there's indeed a way to disable this (and they made it the default behaviour?!) Settings -> Advanced settings -> Uncheck `Shorten external ilnks` Thank you reddit: www.reddit.com/r/Slack/comm...

The new UX of hyperlinks in Slack is a nightmare. They took what looked like a hyperlink but was plain text and converted it into some weird object. You can no longer edit it inline after you pasted it. I want to understand the internal reasoning of whoever thought this was a good idea.

I've noticed a small uptick in the frequency of errors in downloading external dependencies in GitHub CI in the last few weeks. And they all seem to be pointing towards some kind of a connection reset. I hope someone at GitHub is seeing the same in their internal dashboards.

Early on I believed we should be adding LLMs as co-authors to commits we make. My stance on this has now flipped. I think LLMs as co-authors makes us shirk responsibility for our changes. You wouldn’t have added Google or Stackoverflow as co-author previously. So you shouldn’t do it now.

If you’re responsible for the commit being shipped, you’re the author (and any other humans as co-author). If an agent is fully responsible for it then the agent is the author. And you’re the reviewer. I think creating this distinct signal on who is the author for this change is really important.

In the AI vs no AI discourse I see a lot of nuance that’s missing. For a lot of things you can actually not read the code so much or care about it and you’ll be absolutely fine. But there’s a whole bunch of use cases where you should really know how it all actually works or else you’ll pay for it

Yeah I’m beginning to get this feeling myself. When I don’t use AI at all I retain things better.

Benjamin Riley@benjaminjriley.bsky.social

So we know AI helped a negligible number of students. But how many did it harm? I've crunched the raw data provided by the lead researcher, and what I've found is that ~9,500 students using AI studied less than even the *least* studious, non-AI using student. This should shock the conscience.