Arjun
@arjunkomath.bsky.social
Founder (Techulus), building https://changes.page | https://push.techulus.com | https://capture.techulus.in | https://managee.xyz
Just bought a license for @git_fork
Amazing product, and thrilled to support a fellow indie dev!
Amazing product, and thrilled to support a fellow indie dev!
November 10, 2025 at 9:35 PM
Just bought a license for @git_fork
Amazing product, and thrilled to support a fellow indie dev!
Amazing product, and thrilled to support a fellow indie dev!
For some reason, requests in EU servers are failing for Capture. It works fine when I remove those servers. This was supposed to be a chill weekend, not a P1 weekend. 😔
November 9, 2025 at 1:57 AM
For some reason, requests in EU servers are failing for Capture. It works fine when I remove those servers. This was supposed to be a chill weekend, not a P1 weekend. 😔
why? its just polluting my logs
November 8, 2025 at 5:01 AM
why? its just polluting my logs
The funniest thing with Claude is that if you ask it to fix type errors in TypeScript, most of the time it will use "any" or "as unknown as". 😁
November 7, 2025 at 7:04 PM
The funniest thing with Claude is that if you ask it to fix type errors in TypeScript, most of the time it will use "any" or "as unknown as". 😁
This is interesting.
> Vibe coding from scratch produces terrible results.
> But, if you vibe on existing decent code, it recognizes the patterns and results are much better.
I've run several experiments to test this, and the results have been consistent.
> Vibe coding from scratch produces terrible results.
> But, if you vibe on existing decent code, it recognizes the patterns and results are much better.
I've run several experiments to test this, and the results have been consistent.
November 6, 2025 at 6:00 PM
This is interesting.
> Vibe coding from scratch produces terrible results.
> But, if you vibe on existing decent code, it recognizes the patterns and results are much better.
I've run several experiments to test this, and the results have been consistent.
> Vibe coding from scratch produces terrible results.
> But, if you vibe on existing decent code, it recognizes the patterns and results are much better.
I've run several experiments to test this, and the results have been consistent.
I tried vibe coding a simple Nextjs app using @claudeai code and holy crap, the code was absolutely terrible.
I don’t think there was a single thing it got right. It didn’t follow any good practices, the whole landing page was client side, unbelievable.
I don’t think there was a single thing it got right. It didn’t follow any good practices, the whole landing page was client side, unbelievable.
November 6, 2025 at 10:04 AM
I tried vibe coding a simple Nextjs app using @claudeai code and holy crap, the code was absolutely terrible.
I don’t think there was a single thing it got right. It didn’t follow any good practices, the whole landing page was client side, unbelievable.
I don’t think there was a single thing it got right. It didn’t follow any good practices, the whole landing page was client side, unbelievable.
Which BI tool would you recommend that runs queries locally on my machinel?
November 5, 2025 at 2:00 PM
Which BI tool would you recommend that runs queries locally on my machinel?
I want two things from @Apple:
- a folding phone or whatever you’re going to call it, it has to be a phone and tablet in one device
- an upgrade to the display lineup with ProMotion and TantomOLED
- a folding phone or whatever you’re going to call it, it has to be a phone and tablet in one device
- an upgrade to the display lineup with ProMotion and TantomOLED
November 5, 2025 at 5:00 AM
I want two things from @Apple:
- a folding phone or whatever you’re going to call it, it has to be a phone and tablet in one device
- an upgrade to the display lineup with ProMotion and TantomOLED
- a folding phone or whatever you’re going to call it, it has to be a phone and tablet in one device
- an upgrade to the display lineup with ProMotion and TantomOLED
Do @ProductHunt votes or rankings mean anything these days? I've been hearing only negative experiences recently, with PH plagued by fake votes.
November 3, 2025 at 6:01 PM
Do @ProductHunt votes or rankings mean anything these days? I've been hearing only negative experiences recently, with PH plagued by fake votes.
After 7 years, I've updated my personal page
November 2, 2025 at 6:02 PM
After 7 years, I've updated my personal page
I decided to try @cursor_ai again after seeing their 2.0 announcement, my thoughts:
> I find tab completions VERY annoying. They’re really fast now and it’s constantly interrupting my flow with random suggestions, most of them don’t make any sense. This was the thing I loved about cursor.
> I find tab completions VERY annoying. They’re really fast now and it’s constantly interrupting my flow with random suggestions, most of them don’t make any sense. This was the thing I loved about cursor.
November 1, 2025 at 11:14 PM
I decided to try @cursor_ai again after seeing their 2.0 announcement, my thoughts:
> I find tab completions VERY annoying. They’re really fast now and it’s constantly interrupting my flow with random suggestions, most of them don’t make any sense. This was the thing I loved about cursor.
> I find tab completions VERY annoying. They’re really fast now and it’s constantly interrupting my flow with random suggestions, most of them don’t make any sense. This was the thing I loved about cursor.
Web view inside app is hard.
November 1, 2025 at 4:01 PM
Web view inside app is hard.
Time to vent: what I really dislike about landing pages are pop-ups and dropdowns that open only on hover! Please let user click.
October 31, 2025 at 7:01 PM
Time to vent: what I really dislike about landing pages are pop-ups and dropdowns that open only on hover! Please let user click.
To all companies that use AI for cold calls, please don’t. Thank you.
October 31, 2025 at 2:59 AM
To all companies that use AI for cold calls, please don’t. Thank you.
Why is every product rebranding to AI native? Is that the keyword VCs search for?
October 30, 2025 at 5:58 PM
Why is every product rebranding to AI native? Is that the keyword VCs search for?
Improving Capture's success rate from ~95% to over 99% was a long-time goal. After running thousands of tests over the last month, I can confirm it's finally achieved! This is a huge milestone for me 🎉 LFG
If you’re wondering why is hard, it’s all about scaling and running a large browser cluster.
If you’re wondering why is hard, it’s all about scaling and running a large browser cluster.
October 30, 2025 at 9:56 AM
Improving Capture's success rate from ~95% to over 99% was a long-time goal. After running thousands of tests over the last month, I can confirm it's finally achieved! This is a huge milestone for me 🎉 LFG
If you’re wondering why is hard, it’s all about scaling and running a large browser cluster.
If you’re wondering why is hard, it’s all about scaling and running a large browser cluster.
This will be great for LLMs, @Web_Capture can now convert any web page to markdown in addition to plain text and HTML. I’m planning to release a free tool next that can proxy websites and return markdown content when requested by an LLM.
changelog.capture.page/post/a193a1...
changelog.capture.page/post/a193a1...
We’ve just enhanced the Capture Content API!
You can now retrieve your captured content as Markdown, in addition to the existing text and HTML formats.
What’s new
Added markdown field in the Co...
capture.changes.page
October 28, 2025 at 3:56 AM
This will be great for LLMs, @Web_Capture can now convert any web page to markdown in addition to plain text and HTML. I’m planning to release a free tool next that can proxy websites and return markdown content when requested by an LLM.
changelog.capture.page/post/a193a1...
changelog.capture.page/post/a193a1...
I asked ChatGPT what JS framework I should use:
React / Next.js — for full-stack apps with complex, dynamic UIs
TanStack Start — for type-safe full-stack React or Solid apps with SSR & routing
Vue / Nuxt — for structured apps that stay simple to build
React / Next.js — for full-stack apps with complex, dynamic UIs
TanStack Start — for type-safe full-stack React or Solid apps with SSR & routing
Vue / Nuxt — for structured apps that stay simple to build
October 26, 2025 at 5:01 AM
I asked ChatGPT what JS framework I should use:
React / Next.js — for full-stack apps with complex, dynamic UIs
TanStack Start — for type-safe full-stack React or Solid apps with SSR & routing
Vue / Nuxt — for structured apps that stay simple to build
React / Next.js — for full-stack apps with complex, dynamic UIs
TanStack Start — for type-safe full-stack React or Solid apps with SSR & routing
Vue / Nuxt — for structured apps that stay simple to build
Time to read a paper
github.com/apple/pico-...
github.com/apple/pico-...
GitHub - apple/pico-banana-400k
Contribute to apple/pico-banana-400k development by creating an account on GitHub.
github.com
October 26, 2025 at 2:01 AM
Time to read a paper
github.com/apple/pico-...
github.com/apple/pico-...
Is it true that PHP devs own the most lambos?
That should be an actual metrics we track, lambos/dev in each language.
That should be an actual metrics we track, lambos/dev in each language.
October 25, 2025 at 4:03 PM
Is it true that PHP devs own the most lambos?
That should be an actual metrics we track, lambos/dev in each language.
That should be an actual metrics we track, lambos/dev in each language.
Workflow by @vercel will be widely adopted. Everyone might not agree on its semantics, but it’s going to win because:
> It takes minutes to set up and deploy (on Vercel), the “moat”
> It will be an end-to-end solution: framework, infra, and observability, all nicely packaged.
> It takes minutes to set up and deploy (on Vercel), the “moat”
> It will be an end-to-end solution: framework, infra, and observability, all nicely packaged.
October 25, 2025 at 1:47 PM
Workflow by @vercel will be widely adopted. Everyone might not agree on its semantics, but it’s going to win because:
> It takes minutes to set up and deploy (on Vercel), the “moat”
> It will be an end-to-end solution: framework, infra, and observability, all nicely packaged.
> It takes minutes to set up and deploy (on Vercel), the “moat”
> It will be an end-to-end solution: framework, infra, and observability, all nicely packaged.
wow, look at new shiny build machine on @vercel
October 25, 2025 at 11:00 AM
wow, look at new shiny build machine on @vercel
No one should be surprised by this. ChatGPT is a B2C product, they make money by showing ads or charging for a Pro subscription.
October 25, 2025 at 7:16 AM
No one should be surprised by this. ChatGPT is a B2C product, they make money by showing ads or charging for a Pro subscription.
my commits while adding more Partial Prerendering 🚀
October 25, 2025 at 4:57 AM
my commits while adding more Partial Prerendering 🚀
Is Vercel a JS monopoly considering that it maintains Next, Svelte, and Nuxt? 🤔
October 24, 2025 at 2:56 AM
Is Vercel a JS monopoly considering that it maintains Next, Svelte, and Nuxt? 🤔