HNHacker News
TopNewBestAskShowJobs

ashwinsundar

333 karma · joined February 28, 2025

ashwinsundar dot com

ash,i,un,dar,at,g,mail,dot,com

submissionscomments
ashwinsundar··on TI to invest $60B to manufacture foundational semiconductors in the U.S.
What are some things TI promised on, but failed to deliver?
ashwinsundar··on MCP Specification – version 2025-06-18 changes
Why does MCP need to support this explicitly? Is it hard to write a small wrapper than handles async callbacks? (Serious question)
ashwinsundar··on Framework Laptop 12 review
Uninstall Firefox and stop developing junk in Next.js or any other vendor-as-a-service frameworks

Install htop/btop and be more conscious about what your machine is actually doing. Needing more than 32GB RAM to develop a website is absurd

ashwinsundar··on Show HN: Workout.cool – Open-source fitness coaching platform
Wow I really do not want this…cameras to track you in a gym? What kind of dystopia is that?
ashwinsundar··on Liquid Glass – WWDC25 [video]
Just yours
ashwinsundar··on Liquid Glass – WWDC25 [video]
A/B testing doesn’t work if the actual answer is C

I don’t want my computer (which I use for work) to be a fashion statement. I stare at the UI all day, every day. My opinion matters just as much as any designer’s.

ashwinsundar··on GCP Outage
Interesting that all Digital Ocean services are fine...
ashwinsundar··on Ignore files in Git without adding them to .gitignore
Yes, .git/info/exclude is a nice trick. No, we don’t need another AI-generated article about it.

Also, that suggestion about how to avoid committing temporary changes is not great. Just run a git restore, instead of applying strange settings to your entire repo and then wasting time trying to debug whatever went wrong after forgetting that you made that change to git behavior for a single file.

ashwinsundar··on Go is a good fit for agents
Is that what uv sync does under the hood, just curl’s over all dependencies and the python version defined in .python-version?
ashwinsundar··on Tell HN: Help restore the tax deduction for software dev in the US (Section 174)
I don't know how you're supposed to value software. I just reread your original post - picking 10% out of thin air doesn't make much sense either.

Software is more like a blueprint for a building, it's not the building itself. How much is a blueprint worth? If 100 architects spent a year on it, does that mean the blueprint is worth 100 x salaries? It might actually be worth nothing, if the blueprint asks the construction team to do something impossible.

Software is even worse though, because at least with construction, there are known physical models and real-world constraints (like physics) that decide whether a design can or cannot be implemented. A piece of software written today might be entirely unimplementable and worth nothing, but a breakthrough elsewhere in 5 years might make it extremely valuable at that time

ashwinsundar··on Tell HN: Help restore the tax deduction for software dev in the US (Section 174)
I think you are conflating "software engineers" with "software". A business that pays a software engineer doesn't automatically receive working software in return, especially not in the first year. It doesn't seem fair to assume that paying a dev $200k means that the business received an asset (some code) worth $200k in return, and thus can be taxed on it as if it were an asset producing $200k in profits a year.
ashwinsundar··on Congress.gov MCP Server
I should have done a better job explaining what this repo is for. This is an MCP server that interfaces with the (rather comprehensive) Congress.gov API. I made this because I was fed up with reading highly partisan news reports on what our US Congress is actually doing. I don't think there is any easy way to actually find out, for example, what my congresspeople from Colorado (my home state) did today. The Congress.gov API exposes that information, in near real-time. I am honestly surprised more news organizations don't seem to rely on this for, well, news.

There is still work to be done to finish out the remaining endpoints. MCP Servers are the next evolution in interacting with LLM agents in my opinion, and they are dead simple to write in Python, for example. The installation is super straightforward too.

Happy to accept any contributions or ideas. This repo is available via the MIT License.

ashwinsundar··on What “working” means in the era of AI apps
Were computers not intelligent before the latest AI wave? What is “intelligence”, with regards to computers?
ashwinsundar··on Show HN: Most users won't report bugs unless you make it stupidly easy
Great idea. How do I place the bug icon in the location I want (e.g. in a custom header), or use a different icon than the one provided?
ashwinsundar··on Scientific Publishing: Enough Is Enough
I guess I'm still not understanding the point you're making about this person being an administrator. Does science change so rapidly that all of the observations that the author made, as a scientist in the last 10 years presumably, are completely invalid? Or maybe that they have a completely perpendicular motive from scientists, because they are now an administrator?

On your last point about the system cannot be fixed, and it could be a lot worse - everything else you're citing is certainly a problem as well. Maybe even bigger in some cases. I still don't see how those are responses to this article's points. It's kinda like a version of "whataboutism" that you're proposing.

I just don't really get the casualness towards a problem that is widely regarded, among scientists, as something fairly serious - "publish or perish". Why do a handful of publishing houses get to have so much control over the careers of people who dedicate their lives to a field, often at great personal and financial expense (in the form of lost wages and consistently lower pay)?

I thought science was supposed to be about the discovery, and dissemination of information. Why is the process of doing capital-S Science so gatekept behind university or big corp doors? Why is the only output of most scientific research a dozen pages of virtually indecipherable material? Why can't we know about the process that was taken to arrive at those conclusions, especially when they are so incredibly consequential to all of our lives? Science is responsible for virtually all modern advances - why doesn't the public have the right to understand it fully?

That's my takeaway from the article, and I'm curious how you arrived at such a wildly different conclusion.

ashwinsundar··on Scientific Publishing: Enough Is Enough
The second paragraph literally says "I'm a scientist". Scientists do things other than pipette liquids and stain slides all day. They also get to think about the meta details of their field and try to solve them, which it sounds like what this author is after.

I think a lot of proof is given in the article about the claims that you suggest are "false generalizations". What "niche" field do you think the author works in, where these issues with publishing only apply there and no where else? More importantly, in what scientific fields is the publishing system actually working? Those would be effective counter-examples, and I'd genuinely be interested in learning why it works for those fields.

The whole "easy path" argument you're making is totally incomprehisible. Just say whatever it is that you're trying to say. Unless you're saying that scientists should just "care more", and their problems will be solved.

ashwinsundar··on Typing 118 WPM broke my brain in the right ways
This is a really well-made personal site as well. Nice job! I like your indie-hacker dashboard, that's a cool idea.
ashwinsundar··on Typing 118 WPM broke my brain in the right ways
Ha it was the worst when I was still learning the ortholinear layout...I lost the muscle memory to type properly on a membrane keyboard, but still was slow/inaccurate on the ortholinear layout...
ashwinsundar··on Typing 118 WPM broke my brain in the right ways
I started using an ortholinear split keyboard last year, and that was a huge adjustment. I went from typing 100+wpm on a simple membrane keyboard, to less than half of that. I had to basically relearn typing just to accomodate the ortholinear layout, not to mention that the split layout meant I could no longer "cheat" and use my right hand to type keys on the left side of the keyboard when I was feeling lazy.

I did learn the "right" way to type through all this, and my speedhas stabilized at around 100 wpm. This is more than enough for pretty much any activity I do on the computer.

More importantly, however, my wrists no longer hurt from typing continously for 30 minutes. The small sacrifice in speed is definitely worth it in my opinion.

ashwinsundar··on Ask HN: Could you share your personal blog?
https://www.ashwinsundar.com
ashwinsundar··on College Majors with the Lowest Unemployment Rates
This is pretty meaningless data without additional context. Are the graduates employed in the field that they studied? Is a 3.2% unemployment rate that much better than 6.1%, if the employed philosophy majors are actually working low-wage out-of-band jobs that they don't use their degree for?

I don't think this is a strong argument for the narrative (apparently being sold by Microsoft) that AI is automating away all the jobs for computer science grads

Also I can't help but laugh at this quote from the linked HBR article:

    "If you want to pursue a career in engineering, you should focus on learning philosophy in addition to traditional engineering coursework. Why? Because it will improve your code." (https://hbr.org/2024/04/why-engineers-should-study-philosophy)
It reminds me of that vibe in the 2010s where they wanted everyone to learn how to code. Michael Bloomberg was coding. What is the point of that? Does it make him a better politician if he becomes a crack Python developer? Either answer, yes or no, is equally idiotic.
ashwinsundar··on Hacker News now runs on top of Common Lisp
Is there data or otherwise that HN is growing stale? Or more of a general vibe
ashwinsundar··on Hacker News now runs on top of Common Lisp
This is what cmd +/- is for
ashwinsundar··on Ask HN: What are you working on? (May 2025)
Is there a way to contribute? I have some old National Geographics I bought for 10 cents each a number of years ago. The old ads are one of my favorite things in every magazine.
ashwinsundar··on Zod 4
Hard to please everyone
ashwinsundar··on Engineers develop wearable heart attack detection technology
I hope this is true for ECG sampling as well! It does beg the question as to what is the bottleneck today then, if it's not power.
ashwinsundar··on Ask HN: How to Make Friendster Great?
I disagree with this definition of "addictive"...not all behaviors that are self-reinforcing are simply 'addictive'. Is drinking water addictive by your definition?
ashwinsundar··on Animated Factorization (2012)
I believe it's called prime factorization. Each number is placed in a group of numbers (or group of groups, etc...)

E.g. 24 -> 2 * 3 * 4 = Two groups of (three groups of (four items))

Also try this for the archived version of that explanation -> https://web.archive.org/web/20130206023100/http://mathlesstr...

ashwinsundar··on Zod 4
Or the code is done and doesn’t need to be iterated on continuously?
ashwinsundar··on Ask HN: How do you store the knowledge gained in a day?
For things online, I just bookmark web pages in a folder called "Other". For written materials/books, I write notes in the margins or write free-hand notes on a piece of paper and insert it in the book somewhere. Then when I'm bored, instead of infinite-scrolling some stupid app, I try to pick out of the "Other" folder or read a book off the shelf. At least, that's what I aspire to do most of the time...

I also decided that if something's worth remembering, I'll remember it. If it's not worth remembering, then I'm also not gonna both re-reading notes about it, no matter what note-taking tool or method I use...

← PreviousPage 5 of 6Next →