Suyash Srijan
LightNews LightNews
suyashsrijan.com
Suyash Srijan
@suyashsrijan.com
13 followers 28 following 4 posts
29 • 🇬🇧 • iOS Engineer • I play guitar, piano, synthesizers & bass • Radiohead / The Smile / Beach House fanboy.

suyashsrijan.com
Posts Replies Media Videos
Reposted by Suyash Srijan
edzitron.com
I covered this in my premium last week, it’s laughable!

www.wheresyoured.at/ai-mythbuste...
"You Need Al Skills Today To Be Ready For The Future"/"The Future Belongs To Those Who Know AI" So, you've probably seen a lot of stuff — maybe even in your own workplace! — about how you're gonna need Al skills in the future, because, uh. Well, you're gonna need them, and you're gonna get left behind if you don't have them. If anybody ever says this to you, please ask them to be specific, because at this point I'm not even confident I can make up a task that Al can do for me. In fact, here's a simple one: please name me somebody who has gotten significantly better at their job using Al. One person you know. One. Just one. Try it. I am, in theory, exactly the person who should be getting the most out of Al. I write a lot. I research a lot. I constantly have to keep an eye on stuff. I pull stuff out of large piles of data. I think I make like 500 searches a week, I'm a complete animal. I talk back and forth with my editor Matt Hughes constantly to brainstorm ideas. So why don't I use Al? Surely it would be perfect, right? Nope! Al is a terrible researcher. I've used everything from Al mode on Google to the most expensive generative search engines available for large hedge funds, and in 99.9% of cases I've found nothing new or interesting or novel. That edge case was one time when a search engine pulled up a single quote from an earnings statement. Once. One time. So, what do people mean by "Al skills"? Prompt engineering. They mean "how can you get this product that's sold and hyped as a panacea for every informational problem to do the thing that it's meant to do," and the answer is usually "it can't do it, but the way to get it close is to find nuanced ways to ask it to do stuff." In a very real sense, "Al skills training" is actually Al training you - making you fill in the gaps, making you do the work that this alleged "intelligence" does.
December 23, 2025 at 4:59 PM
Reposted by Suyash Srijan
lauren.rotatingsandwiches.com
Sonar Guy is always the most fun guy on the submarine in a submarine movie. no one knows why but it's always true
December 22, 2025 at 12:58 AM
Reposted by Suyash Srijan
etbeeegood.bsky.social
butter has the calcium I need for strong buttery bones
December 19, 2025 at 2:09 AM
Reposted by Suyash Srijan
leviadraconia.bsky.social
That seems to be the general thought.
It always sounds like companies are afraid they might be missing out on huge benefits in the future if they don't lean into AI in one way or another. But I'm not sure how much the pressure from competitors plays into this.
December 17, 2025 at 12:13 AM
Reposted by Suyash Srijan
internethippo.bsky.social
They're selling you the same pasta in different shapes and you're falling for it
December 14, 2025 at 12:09 AM
Reposted by Suyash Srijan
octopodeeznuts.bsky.social
pretty much by definition, what LLMs generate is "legacy code", therefore:

read theory
book cover Working Effectively with Legacy Code by Michael C. Feathers
December 3, 2025 at 3:26 AM
Reposted by Suyash Srijan
shannonvallor.bsky.social
Master Shake from Aqua Teen Hunger Force cartoon saying ‘you look at this tweet and tell me there is a god’
December 2, 2025 at 3:42 PM
Reposted by Suyash Srijan
stevenbailey.bsky.social
When I arrive at the party
A photo of a sign, white text on red background: JESUS... WHO IS HE? WHY DID HE COME?
November 19, 2025 at 5:53 AM
Reposted by Suyash Srijan
mindflakes.bsky.social
Eating a sundae with a long spoon is the height of decadence. The longer the spoon the more decadent it is. Ideally you shouldn't even be in the same room as the sundae
November 11, 2025 at 12:18 PM
Reposted by Suyash Srijan
atherton.bsky.social
If you make and show me a Sora 2 video I want you to know I think of you as infinitely less important than a worm, whose movements and excrement nuture the soil, whereas a Sora 2 user is a creature that asks robots to despoil the world out of boredom.
October 1, 2025 at 2:29 AM
Reposted by Suyash Srijan
polotek.bsky.social
I feel like one of the many effects of LLMs on our industry is going to be making it really clear that the coding was never the hard part.
September 5, 2025 at 8:53 PM
Reposted by Suyash Srijan
radiohead.wastehq.social
Radiohead will play some shows across Europe in November and December 2025.

Tickets will only be accessible by registering on radiohead.com.

Registration opens at radiohead.com on Friday 5th Sept at 10am BST / 11am CEST and runs through until Sunday 7th Sept at 10pm BST/11pm CEST.
September 3, 2025 at 2:00 PM
Reposted by Suyash Srijan
edzitron.com
The greatest irony of the AI booster is they're oftentimes not actually fans of AI. Boosterism is a kind of religion, where technology is a means of proving intellectual and philosophical superiority by telling half-truths about software sold by billionaires.
www.wheresyoured.at/how-to-argue...
So, an AI booster is not, in many cases, an actual fan of artificial intelligence. People like Simon Willison or Max Woolf who actually work with LLMs on a daily basis don’t see the need to repeatedly harass everybody, or talk down to them about their unwillingness to pledge allegiance to the graveyard smash of generative AI. In fact, the closer I’ve found somebody to actually building things with LLMs, the less likely they are to emphatically argue that I’m missing out by not doing so myself. No, the AI booster is symbolically aligned with generative AI. They are fans in the same way that somebody is a fan of a sports team, their houses emblazoned with every possible piece of tat they can find, their Sundays living and dying by the success of the team, except even fans of the Dallas Cowboys have a tighter grasp on reality. Kevin Roose and Casey Newton are two of the most notable boosters, and — as I’ll get into later in this piece — neither of them have a consistent or comprehensive knowledge of AI. Nevertheless, they will insist that “everybody is using AI for everything” — a statement that even a booster should realize is incorrect based on the actual abilities of the models. But that’s because it isn’t about what’s actually happening, it’s about allegiance. AI symbolizes something to the AI booster — a way that they’re better than other people, that makes them superior because they (unlike “cynics” and “skeptics”) are able to see the incredible potential in the future of AI, but also how great it is today, though they never seem to be able to explain why outside of “it replaced search for me!” and “I use it to draw connections between articles I write,” which is something I do without AI using my fucking brain. Boosterism is a kind of religion, interested in finding symbolic “proof” that things are getting “better” in some indeterminate way, and that anyone that chooses to believe otherwise is ignorant.
August 25, 2025 at 4:16 PM
Reposted by Suyash Srijan
michaelhobbes.bsky.social
None of this AI shit would be happening if you all had posted this on Facebook like we told you to
I hereby declare that i do not give my permission to use any of my personal data or photos
August 16, 2025 at 3:51 PM
Reposted by Suyash Srijan
matthewhughes.bsky.social
My latest newsletter is about how big tech and big business has guaranteed that the next generation to enter the workforce will be poorer, less secure, and lonelier.

And why every generation that follows will be a lost generation.

whatwelost.substack.com/p/how-silico...
How Silicon Valley and Big Business Created The Next Lost Generation
The kids aren't alright.
whatwelost.substack.com
August 8, 2025 at 11:01 PM
Reposted by Suyash Srijan
danmentos.com
laugh all you want, but I just bit into a blueberry and there were three bees inside
August 8, 2025 at 3:45 PM
suyashsrijan.com
I see Denzel Washington, I buy ticket
letterboxd.social Letterboxd @letterboxd.social · Aug 4
Watch the official trailer for Spike Lee’s Highest 2 Lowest starring Denzel Washington 👀

The film will be released in theaters on August 15 via
A24 and streaming on Apple TV+ September 5.
August 4, 2025 at 7:42 PM
Reposted by Suyash Srijan
danidonovan.com
children who are constantly criticized

grow into teens who learn to hate themselves for failing to live up to others’ expectations

who grow into adults that reflexively over-apologize because they’ve been conditioned to believe they’re always disappointing everyone
July 28, 2025 at 6:55 PM
Reposted by Suyash Srijan
distantprovince.bsky.social
I am an AI optimist, but sometimes I feel like I am at war with AI. Or maybe with other people. Ugh, listen, you just need to imagine you are a scrambler. distantprovince.by/posts/its-ru...
It's rude to show AI output to people | Alex Martsinovich
Feeding slop is an act of war
distantprovince.by
July 4, 2025 at 10:31 PM
Reposted by Suyash Srijan
ianfay.bsky.social
Very satisfying reading over on X where people are talking about how some Japanese art jobs are increasingly expecting applicants to draw on paper during the interview to prove they're not just prompt sloppers or heavily reliant on it
August 2, 2025 at 11:37 AM
Reposted by Suyash Srijan
salsantamaura.bsky.social
Fully outsource? With LLMs available to the general public, which "learn" from dreck floating around the Internet, most of which is at best wrong and at worst malevolent, one ought not outsource *any* thinking, much less critical thinking, to "Artificial so-called Intelligence."
July 29, 2025 at 5:23 PM
Reposted by Suyash Srijan
davidcrespo.bsky.social
from terry tao, nice way of putting why I tend to use LLMs more for review and analysis than for writing code. gemini 2.5 pro spots problems in PRs that I missed *all the time*. definitely some false positives too, a bit like a noisy linter, but they're never gibberish
mathstodon.xyz/@tao/1149156...
Terence Tao @tao@mathstodon.xyz Many of the proposed use cases for AI tools try to place such tools in the "blue team" category, such as creating code, text, images, or mathematical arguments in some semi-automated or automated fashion, that is intended for use for some external application. However, in view of the unreliability and opacity of such tools, it may be better to put them to work on the "red team", critiquing the output of blue team human experts but not directly replacing that output; "blue team" AI use should only be permitted up to the capability of one's "red team" to catch and correct any errors generated. This approach not only plays to current AI strengths, such as breadth of exposure and fast feedback, but also mitigates the risks of deploying unverified AI output in high-stakes settings. In my own personal experiments with AI, for instance, I have found it to be useful for providing additional feedback on some proposed text, argument, code, or slides that I have generated (including this current text). I might only agree with a fraction of the suggestions generated by the AI tool; but I find that there are still several useful comments made that I do agree with, and incorporate into my own output. This is a significantly less glamorous or intuitive use case for AI than the more commonly promoted "blue team" one of directly automating one's own output, but one that I find adds much more reliable value. (3/3)
July 28, 2025 at 3:16 PM
Reposted by Suyash Srijan
pixelatedboat.bsky.social
I would rather die than create an account to continue
July 30, 2025 at 1:29 PM
Reposted by Suyash Srijan
matthewhughes.bsky.social
After the $400m hack, Clorox and Cognizant traded words and legal filings, each trying to blame the other.

An LLM doesn't point fingers back.

And if a vibe-coded app built by a non-coder does something truly catastrophic, who is truly to blame?

whatwelost.substack.com/p/mediocrity...
July 29, 2025 at 10:23 PM
Reposted by Suyash Srijan
anthonyhyphen.bsky.social
This paragraph by @edzitron.com is everything.
www.wheresyoured.at/the-haters-g...
July 23, 2025 at 12:10 AM