Hacker News

Article URL: https://www.nature.com/articles/d41586-025-00916-4

Comments URL: https://news.ycombinator.com/item?id=43630084

Points: 1

# Comments: 0

Article URL: https://aws.amazon.com/blogs/devops/amazon-q-developer-global-capabilities/

Comments URL: https://news.ycombinator.com/item?id=43630075

Points: 1

# Comments: 0

Dear HN,

I wonder if anyone has a good idea on how to turn cool technical solutions (in search of a problem) into an actual product.

By now, I have collected quite a list of cool things where I'm sure they will be useful for someone, but I have no idea for how to productize them. And I'm sure there will be many others in here who have cool tech in their basement that they built for themselves, but that might be of value to others, too. So I wonder if just like how there are freelancing job boards, are there any job boards for product ideas? I'm happy to give the ideas and product descriptions away for free because I'm pretty sure the execution is what counts and there I expect that I'd have a strong moat, given that I already built these.

Here's some examples:

1. Headphones with integrated 3D audio. The Apple stuff only works with Apple hardware and it greatly suffers from Bluetooth latency because all the processing is done inside the host device. I have my own headphones with DSPs directly attached to your head (they are tiny and lightweight) and then you plug the thing into your PC / Console / TV as a USB dongle. To the device playing back audio, it looks like any other 192kHz x 8 channel x 32 bit audio interface. But to you, it'll do real-time virtual surround at a quality level that (in my opinion) exceeds Apple's. The only competitor is the Smyth Realizer at $4,695 and, accordingly, far outside of the consumer market.

2. A fully offline voice assistant. I like the idea of Alexa, but I would never agree to all of my data being siphoned off. So I built my own 3-stage pipeline. The first stage converts the audio into an intermediary phoneme-like token representation. The 2nd stage runs an LLM on this pseudo-language. The 3rd stage converts the resulting phoneme tokens back into audible speech. Deployment works on 2x 3090 (one for audio IO, one for LLM) and since the LLM is based on Llama 3, it has a similar intelligence level. That makes it smart enough for most tasks. This would probably be great as a combined hardware+software appliance. Because setting up a Linux to host the AI models is quite tricky and messy. I wrote a paper on a variant of the first stage (TEVR Token Entropy Variance Reduction) and it works quite well, but didn't get much interest outside of academia.

2b. Somewhat related, I also have a 100% offline LLM coding assistant as a JetBrains plugin.

3. A 3D camera that can see multiple depths per pixel and generate deep EXR files. That makes it so much easier to do compositing of CGI objects into real footage, especially because this 3D camera can handle windows, transparent materials, reflections, etc. Looking at how Skydio and GoPro drones regularly crashed into windows and mirrors before they got discontinued, I'm sure this is useful to someone. The trick to make it work was that I built my own hardware from the ground up so that the AI that does the post-processing has 100% accurate information about optics, inter-camera transformations, noise levels, and physical nit to digital value mappings. That means my moat is that I have schematics, PCBs, firmware, datasheets, etc.

4. Two synchronized robot arms with real-time inverse kinematics (custom HW again) and a 3D scanner which can track the shape of viscous fluids as a voxel volume and then use its arms to modify it. An AI model consumes the current voxel state and the desired target shape and then predicts the arm movements that are most likely to bring the current shape closer to the desired shape. For example, this could be used to make pretty amazing chocolate statues.

Would you try to advance such projects to become proper products?

And if yes, what do you think are good next steps?

Are there any job boards for product ideas?

Comments URL: https://news.ycombinator.com/item?id=43630067

Points: 1

# Comments: 0

Article URL: https://infeely.com

Comments URL: https://news.ycombinator.com/item?id=43630038

Points: 1

# Comments: 0

Article URL: https://clearmvp.com/

Comments URL: https://news.ycombinator.com/item?id=43630023

Points: 1

# Comments: 0

Article URL: https://www.evidentlyai.com/ml-system-design

Comments URL: https://news.ycombinator.com/item?id=43629918

Points: 1

# Comments: 0

Article URL: https://www.viblo.se/posts/code-review-death/

Comments URL: https://news.ycombinator.com/item?id=43629905

Points: 1

# Comments: 0

Article URL: https://forum.vcfed.org/index.php?threads/pdp-11-hack-de-luxe.1252532/

Comments URL: https://news.ycombinator.com/item?id=43629899

Points: 1

# Comments: 0

Article URL: https://www.centauri-dreams.org/2025/04/08/on-sun-like-stars/

Comments URL: https://news.ycombinator.com/item?id=43629887

Points: 1

# Comments: 0

Article URL: https://mailchi.mp/shearwater/tern_terntx_product_bulletin-250808

Comments URL: https://news.ycombinator.com/item?id=43629886

Points: 1

# Comments: 0

Hi HN,

I built Webtor — an open-source engine that lets you stream .torrent and magnet links directly in the browser.

- No swarm exposure — all torrent traffic runs server-side

- In-browser playback via HLS

- Progressive download with resume support

- Lightweight JS SDK and public API

- Fully self-hostable via Docker

Demo: https://webtor.io

GitHub: https://github.com/webtor-io

SDK: https://github.com/webtor-io/embed-sdk-js

Self-hosted: https://github.com/webtor-io/self-hosted

API: https://rapidapi.com/paveltatarsky-Dx4aX7s_XBt/api/webtor

Would love feedback!

Comments URL: https://news.ycombinator.com/item?id=43629869

Points: 2

# Comments: 0

Article URL: https://www.kellton.com/kellton-tech-blog/react-design-patterns-dominating-react-development-lifecycle

Comments URL: https://news.ycombinator.com/item?id=43629866

Points: 1

# Comments: 0

Article URL: https://www.taiwannews.com.tw/news/6080396

Comments URL: https://news.ycombinator.com/item?id=43629838

Points: 2

# Comments: 0

Article URL: https://www.semafor.com/article/04/06/2025/big-tech-squares-off-with-eu-regulators

Comments URL: https://news.ycombinator.com/item?id=43629832

Points: 2

# Comments: 2

Light and terminal version of datagrip or dbeaver

Comments URL: https://news.ycombinator.com/item?id=43629822

Points: 1

# Comments: 0

Article URL: https://twitter.com/cremieuxrecueil/status/1909821615547838484

Comments URL: https://news.ycombinator.com/item?id=43629801

Points: 1

# Comments: 0

Article URL: https://minireference.com/blog/fixing-the-statistics-curriculum/

Comments URL: https://news.ycombinator.com/item?id=43629772

Points: 1

# Comments: 0

Article URL: https://www.infiniteconversation.com/

Comments URL: https://news.ycombinator.com/item?id=43629767

Points: 1

# Comments: 0

I did some research, but lots of offers seem to be ghost jobs and job boards look scammy

Where do you find fully remote (global) jobs in software/ML engineering nowadays?

Comments URL: https://news.ycombinator.com/item?id=43629732

Points: 9

# Comments: 2

Article URL: https://www.theverge.com/tech/638566/micro-usb-is-the-worst

Comments URL: https://news.ycombinator.com/item?id=43629723

Points: 3

# Comments: 0