Ask HN: What to Build in the AI Era?

I think the answer to this question, and broadly to AI in general, is to focus on building yourself.

By that I mean focus less on creating some external thing, and focus more on “sculpting” your self.

That can mean anything from learning how to repair cars, how to draw, a foreign language, getting in amazing physical shape, etc.

The point is that as people rely more and more on some external tools to create things, you’ll be less dependent on them. Such a person will stand out more and more as time goes by.

Someone that speaks five languages and can draw well is immensely more impressive and capable that someone using Google Translate and an AI image generator.

If I have not skill or any experience at something. Then I have noting to build even in AI era. Cuz I am not useful that nearly provide zero value to other. So wanna build something needs skills or sort of things can provide value to someone. But still not enough , I think whether or not in AI Era, deciding to build something is likely fight information war: you own something other do not have and also valuable for some group people. Then fight with people who might also have(that is competitive) or own something that other people do not have (this is refer to both competitors and customer). The key is own something unique, and useful to others, regardless a group of people or competitors. If you own something that your competitors don't have and customer you server don't have it too, The best signal. and useful might means that things that you have can help someone to get something , like money, times, happy... I don't think AI alters the fundamental rules of entrepreneurship; it simply acts as an accelerator, speeding up your learning process and the journey from writing the first line of code to delivering the final product.Cuz when you collect information with real people, find the hidden value you have and others do not . it helps noting

I'm working on data diodes... I've got the parts, now I've got to gather enough spoons[1] to do the work to make them useful.

Details: Building a physically irreversible way to transmit data in only one direction should be easy, using a 6N137 optoisolator. The hard part is the proxy servers that you put on both sides of it. Old timey Raspberry Pi B+ boards are my current choice for hardware, as they have NO wireless capabilities, and do have reasonable ethernet ports.

I want to keep the retail parts cost under $100, and open source it all.

[1] https://en.wikipedia.org/wiki/Spoon_theory

Personally, I'm using AI to help me learn lower level stuff, embedded programming, and new languages and frameworks that I haven't had the time to dive into on my own. I'm hoping that once the smoke clears, this whole thing will end up being like WYSIWYG editors, which didn't end up eliminating the need to code for the web.

Basically, I'm trying to balance using Opencode to do some of my work that can be tedious to do manually, and stay somewhat relevant (not that I ever really was), and keep my skills sharp. I'm not feeling confident it's going to do me any good, but who knows.

Some day, you'll get old. Then you'll discover that real needs of older folks (and you can throw in kids, and some other groups) are almost entirely unmet.

So, DJT gets an exoskeleton, intensive medical care, a chauffeur, prepared meals (not just Big Macs and KFC) and millions struggle first, to pay the rent and utilities, but also struggle with 2-point type food labels, walking around, lifting heavy things, driving long distances for medical care (because a conglomerate bought and destroyed the local medical center) and other goods, repairs, travel and even simple nearby fishing and photo excursions. I've been a techie all my life, and know what's going on, but many seniors are phioshed, and scammed mercilessly and don't know the score. They end up with annuities, with astronomical overhead and fees (and very generous sales commissions!) or end up becoming Amway franchises, or buying time-shares. Information is scattered all over the web (ask me how I know) and generally confusing, contradictory and hard to find. Search is COMPLETELY broken. I use DDG and get brutally useless results most of the time. Pity the non-techie. LLM'S instead? They have led to suicides with their obsequious answers drawn from every byte in the universe rather than a bottom up and local-first data store. Social media? Advice from trusted friends? Don't make me laugh. It's bot-ridden misinformation galore.

So, if you want a compelling project, know that search is broken, LLM's are dangerous and designed not to be objective, and that accurate and useful information is harder than ever to find.

Deshittify the universe, but one bit at a time.

You must be someone that needs something, then you will learn why chatgpt is not always the solution, despite it can do things, but there is so much friction in many tasks, because you need to take time to navigate around threads, type, etc. and in real world these actions can be awkward when the situation comes.

I feel like most consulting companies are hired because they are a brand that does a service and has some reputation. I think there will be a whole class of tech businesses that fill that niche. They might need a lot more human interaction than classic SaaS has been used to though. So I guess: build a niche and a reputation?

Assuming you are thinking to build something that would make money - whatever you build someone will clone it with AI shortly thereafter. without a serious moat (which is insanely hard to get) I don’t think there is much to build anymore. “Good” thing is that nowadays you can build things fast and fail/succeed fast unlike “before” where building something took serious time

Solve real-world problems offline using local models. Set yourself constraints: no internet, standard hardware, not a lot of RAM. Research how productive terminal applications have been in the history of computing or why grayscale or even 1-bit user interfaces worked.

That's what I'm doing.

Nothing has changed. You can build anything in the AI era, just much faster. In fact, it might be even better now to build things that are already making money for someone else. And as always, marketing comes first.

LLMs don't solve all the world's problems. There is still so much left to create. Look closely at your own needs and the needs of those around you, and you will find ideas there.

Might need a domain expert to help to be honest. I found that people often over-look marketing and who the audience/customers are.

Chat us a horrible UI for most things. It seems like there is an opportunity for those who realize this and do something about it.

Make this but open source: https://getcoldturkey.com/

What's the most ambitious thing you figure you could build?

We need to know what you're capable of.

in your own life, ask yourself, what do I need? what could make my life, or the lives of those I love, better?

Your humanity.

Hardware and specialized Android appliances (Android TV / Google TV / Digital Signage).

A lot of focus is currently on browser-based LLM wrappers, but there is huge untouched ground in headless device management, local orchestration, and edge utility tools for living-room/kiosk hardware.

For instance, managing TV boxes, local ADB orchestration, clean custom launcher provisioning, and lightweight local automation on constrained living-room devices without relying on bloatware or invasive accessibility hacks. Building tools that solve practical, physical-device problems will always have solid utility.