shubhamjain

> For myself, in the last 10 years, my work of writing code has largely defined what I do with my working time. Now I experience large swaths of that work being created and done by AI (sometimes amazingly well, sometimes poorly), and I find myself thinking of the photographer above.

Which is ironic, because smartphone photography is astonishingly average to someone who actually understands photography. Phones try to handle every scenario for the user. It has been optimized for bright outdoors, low light conditions, even specific stuff like sunsets. So most people never realize how difficult it really is to shoot in bright daylight with the light source behind the subject. Only when you pick up a traditional camera do you realize that’s one of the worst conditions to take a photo in.

Maybe coding is headed toward the same place. AI coding will smooth out complexity for the majority of programmers, but at the same time, we’ll also see a lot of programmers building things that don’t actually need to be built, or not realizing the limits of their solutions, or just hung in the narrow space of what AI can do for them.

Having done photography with Sony-a6000, I’ve found that it made me more skilled with a smartphone camera. So, I am still optimistic that knowledge gained through deliberate effort has value, even in a world that increasingly prizes “effortlessness” in everything. But only time will tell how much that belief really holds up.

It's difficult even being someone not from United States. Even more concerning is how most of the social media has been hijacked to echo views that should have been unthinkable a few years ago. For some, it's great news of course. It's free speech—they think. But, I am not sure future generation needs to across opinions like feminism is bad, or women belong in the kitchen.

I do believe that this is a temporary phase, and things would return to a bit of normalcy in a few years. It's my hypothesis that political pendulum swings between two extremes. 2016-21 was a period of extreme leftist opinions and it was problematic too in many ways. The other extreme is a hundred times worse and most people have lost taste for that, too. Secondly, the support for these movements mostly stems from an aging demographic, a lot of which are just passing their last few years.

But I would say social media needs to be clamped down/regulated, free speech be damned. It has caused enough evil already. Can't let a few billionaires decide the direction of world politics.

In my late 20s, I felt disillusioned with this kind of wisdom. Found it too simplistic and not enough to cope with the fact that existence is absurd. I am 32 now, and I can't help think that simplicity is all that there is: curate a happy state of mind, meaningful relationships, active lifestyle, and maybe some audacious goal to keep yourself busy. Thinking that there's some higher state of mind (via spirituality, for eg) is delusion at best.

I feel everything follows the Midwit meme progression [1]: at first you use crude, obvious methods because you don’t know better. Later, complexity is alluring, you drown yourself in optimisations and finding the bestest tools and methods. In the end you come back to the same conclusion: simplicity was the most reliable tool the whole time.

[1]: https://medium.com/@obandoandrew8/bell-curve-meme-avoiding-t...

No one needs to make an investment. The playbook—as all countries are figuring it out—to navigate the crisis is simple: agree in principle to the demands, let the narcissistic toddler self-pat, brag about how fantastic of a deal maker he is. Stall, delay to make things official. Let it all die down slowly. Need more time? How about blatant corruption like approving Trump Towers, or gifting a corporate jet? Or make a token investment, just enough to let him brag again.

What every country is doing is buying time. Just enough till the next election or (hopefully) sooner.

> The primary counterargument can be framed in terms of Rich Sutton's famous essay, "The Bitter Lesson," which argues that the entire history of AI has taught us that attempts to build in human-like cognitive structures (like embodiment) are always eventually outperformed by general methods that just leverage massive-scale computation

This reminds me Douglas Hofstadter, of the Gödel, Escher, Bach fame. He rejected all of this statistical approaches towards creating intelligence and dug deep into the workings of human mind [1]. Often, in the most eccentric ways possible.

> ... he has bookshelves full of these notebooks. He pulls one down—it’s from the late 1950s. It’s full of speech errors. Ever since he was a teenager, he has captured some 10,000 examples of swapped syllables (“hypodeemic nerdle”), malapropisms (“runs the gambit”), “malaphors” (“easy-go-lucky”), and so on, about half of them committed by Hofstadter himself.

>

> For Hofstadter, they’re clues. “Nobody is a very reliable guide concerning activities in their mind that are, by definition, subconscious,” he once wrote. “This is what makes vast collections of errors so important. In an isolated error, the mechanisms involved yield only slight traces of themselves; however, in a large collection, vast numbers of such slight traces exist, collectively adding up to strong evidence for (and against) particular mechanisms.”

I don't know when, where, and how the next leap in AGI will come through, but it's just very likely, it will be through brute-force computation (unfortunately). So much for fifty years of observing Freudian slips.

[1]: https://www.theatlantic.com/magazine/archive/2013/11/the-man...

"This tool 10x the productivity of software engineers"

"GREAT! That means we can fire the people who do the actual work, and replace them with MBA robots, who neither understand nor care about making a good product"

Pardon my pessimism, but in my whole career, I have never met a PM who actual did the work of driving the product vision. Most were just middlemen shuttling information between management, marketing, design, and engineering. Thinking that hiring more PMs would increase the output in the age of AI is such a childish fantasy.

Geez! When it comes to answering questions, GPT-5 almost always starts with glazing about what a great question it is, where as GPT-4 directly addresses the answer without the fluff. In a blind test, I would probably pick GPT-4 as a superior model, so I am not surprised why people feel so let down with GPT-5.

Even if AI progress all stops and we are stuck with these models, it would at least take a few years to make use them to the full potential. What gets lost in this AGI debate is how impressive are these models without even reaching it. And the places where it has been applied are just tip of the iceberg.

It's just like dot-com bubble, everyone was pumped that Internet is going to take over the world. And even though, the bubble popped, the Internet did eventually take over the world.

Disappointing to see all the snarky remarks here. Let me tell you, it's not easy to ignore the AI wave. The fear of being left behind is very, very real. AI-based tools are quadrupling their revenues in a matter of months, even on a very high base (Cursor famously going from $100M to $300M in 3 months). Reading "it's all over", "graphic design is dead" all over social media doesn't help either.

I am an author of a Desktop App [1] myself. More often than I like to admit, I find myself wondering how I can integrate AI, because that's all the rage these days and gets views, clicks, and potentially revenue.

Taking a firm stance like this and standing up for creativity is not an easy stance. Kudos to Procreate team for taking it!

[1]: https://textquery.app/

I wasn't expecting much, and I personally didn't get a lot from the article, but if I was in early 20s, I would've been hugely inspired. Jessica surely has a gift for clear and motivating writing.

It should a case study how every tech company eventually ruin the experience of their platforms, and how Apple never does. You can fault Apple for not innovating enough, but what they never do is 'shoving in the face', which is endemic to companies like Microsoft. I had to use Windows to test my app's Windows build [1], it's crazy how many random things are there, with no obvious way to remove them.

There's donkey right in the corner of the bottom search bar reminding me that today is "World Donkey Day". On the other corner is some random info clicking, which I get breaking news, weather, and stock-related info. I just begin using the system, when there's popup about co-pilot chat or something. Search is almost useless as it seems more interested in returning bing-related results vs what's actually on my computer.

Everything seems to be designed to maximize 'user engagement' of products that are hot right now, and what upper management seems to be interested in. The news is no surprise as it seems everyone in the company is rallying behind AI efforts without paying heed to user experience.

[1]: https://textquery.app/

This is a pretty bad take, imo and confuses cause and effect. Why does US press hype their startups? Because they do have track record of creating revolutionary companies, something that can be traced back to 70s (Atari, Apple, Microsoft).

It’s much easier for an established player to replicate the success elsewhere, despite of over-regulation. Just see how successful are US firms in India, but it will be mistake to not fault Indian bureaucracy and regulation for the lack of its own set of high-tech companies. So I would take Airbnb's success in EU with a grain of salt.

In US, you can start a LLC, get a bank account, a tax ID, all within a few days. The process is completely online. Taxes? A single form for the whole year when you’re starting out. Hiring and laying off employees is relatively easy. How many European countries can claim the same?

TextQuery actually uses DuckDB under the hood :) DuckDB is no doubt an excellent piece of software already. The differentiator is that GUI client is just a good lever to make many things easier.

In context of TextQuery: you can use tabs you can work on multiple queries. With Table editor you can edit multiple field values at once. During import, you can have better control over what the final table would look like (select/deselect columns, define data/time format etc.)

Again, it's a personal preference. Some people swear by psql, and some can't live without TablePlus/Postico.

Thanks for the feedback! It is something on my mind. Will think about the best way to implement it.

I am planning to add it soon. It shouldn’t be hard. Just have to figure out the UI part.

IIRC, SQLlite requires you to define a schema first, which can be a bit tedious. DuckDB is actually a better tool for this in every regard, since it can parse CSV files quite well. The latest version includes support for XLSX as well.

My personal take is, yes, it's possible in multiple other ways, but I just like having a well-designed GUI app do it for me. Just how TablePlus does for Database Management.

Author here. It uses DuckDB under the hood, which is a fantastic piece of software.

Yes, even DuckDB has UI now, but I didn't find it to my taste. There's no table view, filters are a bit weird, and not exactly keyboard accessible. TextQuery also has Visualize feature to quickly create charts and tabs to run things in parallel. Again, it's personal preference, but there are advantages in being a desktop-first app.

It seems you're the one who's plenty misinformed. Most countries outsource when labour costs in their own country inch higher and it makes sense to use cheaper external labour supply. Same is not the case with India, where labour costs are still extremely cheap. Even with sky-high import duties, we're not competing with China.

I never said we don't manufacture. There are few notable industries that we are decent at and export it as well: two wheelers, pharma. But we manufacture only teeny tiny bit of mass-manufactured goods, especially electronics.

Self-reliance makes no sense when you're trying to one component out of hundreds involved. Sure, it won't be entirely useless, it will decouple some of the chip imports for defense and auto, but it will do nothing to bootstrap the electronics industry.

> Zoho was in talks with the Department of Electronics & IT under India’s Semiconductor Mission, and reports reveal that they had set aside some $200 million for a pilot 28 nm wafer line in Tamil Nadu.

My guess is 28nm is already an aging node, and by the time any fab based on it becomes operational, the technology will be even more obsolete. It’s hard to predict what the market will look like then. You can't expect a private company to pour billions into a dicey market.

This is a case where free-market capitalism doesn’t work. You really want a government pitching tens of billions of dollars to get to the cutting edge fast, not expecting companies to foot the bill and just dangle a few incentives.

Frankly, for India, the entire semiconductor ambition feels like putting the cart before the horse. It won't create jobs, most value will be captured by fab owners. The smarter approach would be top-down: start with component assembly, then build up CNC machining, injection molding, and other foundational industries that can fill in the gap for components. Only after that should we talk about high-value semiconductor manufacturing.

India lacks even the most basic manufacturing. We import stuff like lighters, toys, and basic electronics. Why do we want to make the giant leap to semiconductors?

Unfortunately, Indian policymakers are more interested in optics than in long-term industrial planning.

I don't get it. Are JSON and XML files more friendly to import vs CSV files? I always assumed CSVs were the standard. Any reasons to prefer structured formats?

Shameless plug: I am working on a similar problem of Excel not being a great tool for large datasets. My desktop app[1] lets you import raw data files and query them using SQL. (The website needs to be updated, the app looks much better than the current screenshots).

[1]: https://textquery.app

I've been working on https://textquery.app -- query raw data from your desktop.

I love exploring data, but it always felt clunky to juggle multiple tools, write code/commads, just to import and query a dataset. While there are multiple GUI tools for databases, none are focused on raw data.

TextQuery is the tool I built to solve that. You can import CSV, JSON, and XLSX files and start querying them instantly using SQL. Want to create a chart? Just hop over to the Visualize tab.

I'm also rolling out a major update this week — adding tabs, filters, a redesigned UI, and keyboard shortcuts.

Looking ahead, I’m planning to expand support for more formats (like Parquet and ORC) and data sources (like Postgres and BigQuery), so you can import data from anywhere, and query it right from your desktop. Something like a local data warehouse. With Apple Silicon, the capability of a desktop can make it very cost-efficient compared to something like BigQuery, Snowflake, or Athena.

Happy to hear any feedback!

The Ghibli trend completely missed the real breakthrough — and it’s this. The ability to closely follow text, understand the input image, and maintain context of what’s already there is a massive leap in image generation. While Midjourney delivered visually stunning results, I constantly struggled to get anything specific out of it, making it pretty much useless for actual workflows.

4o is the first image generation model that feels genuinely useful not just for pretty things. It can produce comics, app designs, UI mockups, storyboards, marketing assets, and so on. I saw someone make a multi-panel comic with it with consistent characters. Obviously, it's not perfect. But just getting there 90% is a game changer.

I’m not sure if it’s Covid or just entering my 30s, but both I and many of my friends have noticed a similar shift. We used to love the idea of traveling the world and meeting new people, and while we still do that to an extent, the excitement of backpacking around the globe and going to off-beat places isn't there anymore.

However, it doesn’t bother me. The idea of constant travel can feel a bit forced, because Western society emphasises it so much as a way to feel fulfilled. But, happiness and fulfillment exists in a thousand other things. Isaac Asimov, for example, spent most of his life in a New York apartment, writing articles, books, and letters. He loved it so much that the first thing he did after coming home from heart surgery was to rush towards his typewriter. He disliked being distracted from writing so much that he wasn't even willing to travel to Hollywood to get his novels adapted.

So, while travelling occasionally is enriching and helps me mentally, I am okay with the idea of just sitting at home and working on my projects, being excited about programming, writing, and learning.

I’m a bit confused. I thought the main selling point of Tauri was that it uses the OS’s native WebView to create lighter apps compared to something like Electron. So why would I want to bundle a Servo or Verso? Doesn't that defeat the purpose of Tauri?

With the asset sizes they are talking about (hundreds of terabytes), how does it make it feasible to do this over the wire? Even with 1Gbps connection, it will take ~10 days to upload a single 100TB original camera file. And there could be several.

I had the same thoughts on conversational interfaces [1]. Humane AI failed not only because of terrible execution, the whole assumption of voice being a superior interface (and trying to invent something beyond smartphones) was flawed.

> Theoretically, saying, “order an Uber to airport” seems like the easiest way to accomplish the task. But is it? What kind of Uber? UberXL, UberGo? There’s a 1.5x surge pricing. Acceptable? Is the pickup point correct? What would be easier, resolving each of those queries through a computer asking questions, or taking a quick look yourself on the app?

> Another example is food ordering. What would you prefer, going through the menu from tens of restaurants yourself or constantly nudging the AI for the desired option? Technological improvement can only help so much here since users themselves don’t clearly know what they want.

[1]: https://shubhamjain.co/2024/04/16/voice-is-bad-ui/

AI, LLM basics. A hundred hours won't even scratch the surface of recent advanced, but I’d like to understand how LLMs work and the main approaches (diffusion, RLHF), their pros and cons.

Theory of relativity. I’ve never quite been able to wrap my head around it. Given some free time, I would really want to understand how two simple postulates can lead to such far-reaching conclusions.

The more apt framing of this article would be something along the lines, of "The Agency Trap." The author grew his marketing agency to $1M quickly. That might sound great for HNers here, but there are reasons why such businesses don't attract handsome valuations. You always have to keep grinding to keep the inflow of customers, you're never free, and growth is directly tied to the number of employees. Additionally, such business can have severe ups and downs depending upon market conditions.

Building a product, on the other hand, is slow at the start but after a point, the rewards start showing up. The inflow of cash is consistent and keeps growing while the costs remain the same.

There's nothing wrong with the agency business, but it'd be a smart bet to use at least some of the proceeds into building something long-term. Some agencies, notably 37signals, have been quite successful doing that. Most, though, never do it.

Quite a lot of people have told Jared (Bun’s Author) the same thing, but his opinion is that Bun should have everything a basic project might need. Keeping it in core he can make it more optimized than it would be as third-party library.

It’s a misguided approach according to me. And I feel Jared has become way too ambitious. But what can I say, it’s his passion project.

None of those ways are sustainable. Not only because there are good reasons to use those apps, but also because there are times when forcing yourself to work isn't going to work. I mean, if I am sick, tired, and just not feeling like working, I would go out of my way to beat the system I installed.

What has worked for me is: one-sec extension [1]. The extension asks you take a deep breath and confirms if I still want to open the app. What I have realized is I don't want to completely do away with time-sink websites, I only want to moderate my behavior of pressing Cmd-T and opening reddit/youtube/twitter in the middle of work. I have increased the length of the pause to 30 second and I am actively forcing myself to actually take the deep breath. Such a pause is enough to knock enough sense into me and return back to work. I think such kind of gentle nudging is better than being overly harsh on yourself.

[1]: https://chromewebstore.google.com/detail/one-sec-website-blo...