Hacker Newsnew | past | comments | ask | show | jobs | submit | heltale's commentslogin

Location: Baltimore, MD, USA

Remote: Yes.

Willing to relocate: Yes, anywhere in the US.

Technologies: Languages[Python, Golang, TypeScript, Rust], Data Engineering Stack [DuckDB, Spark, Polars, FastAPI], GIS [Wails, React, Cesium], AI [HuggingFace, PydanticAI, PyTorch, vLLM, Transformers, OpenCV]

Resume: https://heltes.com/resume.pdf

Email: dat.adithya@gmail.com


The entire experiment just reminds me of Manna. We’re progressing a little too fast for comfort.

https://marshallbrain.com/manna1


Thank you for the reference, it's a fascinating read.

It would be good to highlight that this is fiction, though.


Hey Peter, I’m an F1 visa student and I’m trying to found a company with two Americans while being an Indian citizen. I plan on taking significant equity and will be part of the cofounders. Are there any significant hurdles that I should be on the lookout for here? I’d like to sign the founders agreement and found the company while I’m still pursuing my education but I’m not sure whether my visa status would throw any wrenches here.


It’s pretty much all you see nowadays on LinkedIn. Instagram is infected by AI videos that Sora generates while X has extremist views pushed up on a pedestal.


I suppose it’s only worth 256 words at a time right now. ;)

https://arxiv.org/abs/2010.11929


The CALM paper https://shaochenze.github.io/blog/2025/CALM/ says it is possible to compress 4 tokens in a single embedding, so... image = 4×256=1024 words > 1000 words. QED


2.4% relative error is not bad.


Reminds me of Babbage making allowance for meter.

"""

    ... it is said that he [Babbage] sent the following letter to Alfred, Lord Tennyson about a couplet in "The Vision of Sin":

         Every minute dies a man,
         Every minute one is born

    I need hardly point out to you that this calculation would tend to keep the sum total of the world's population in a state of perpetual equipoise, whereas it is a well-known fact that the said sum total is constantly on the increase. I would therefore take the liberty of suggesting that in the next edition of your excellent poem the erroneous calculation to which I refer should be corrected as follows:

         Every minute dies a man,
         And one and a sixteenth is born

    I may add that the exact figures are 1.167, but something must, of course, be conceded to the laws of metre.
"""

    Charles Babbage and his Calculating Engines


Shouldn't it be the other way around if the population is increasing? Every minute one is born = 1440 born/day, every minute and a sixteenth ~= 1335 dead/day for a net population increase of 105/day.


It means that in every minute, one and a sixteenth of a man is born.


Wouldn't "one and a sixth" be more accurate in both respects?


how do you decompress all those 4 words from one token?


Not from one token, from one embedding. Text contains a low amount of information: it is possible to compress a few token embeddings into a single tiken embedding.

The how is variable. The calm paper seems to have used a MLP to compress from and ND input (N embeddings of size D) into a single D embedding and other for decompress them back


The mechanism would be prediction (learnt during training), not decompression.

It's the same as LLMs being able to "decode" Base64, or work with sub-word tokens for that matter, it just learns to predict that:

<compressed representation> will be followed by (or preceded by) <decompressed representation>, or vice versa.


Hey, that's a pretty cool concept!

I work on developing technology to solve sales and real-estate problems and streamlining/populating connections with people is a big issue. Exporting these to an excel sheet/csv is something that sales folk would absolutely kill for to make things easier to manage.

However, the hurdle to have enough get the application and have it running on both parties' systems while at the event is certainly a pretty strong bottleneck.


for sure, I need to start with events where the organizers can tell everyone ahead of time to have this app installed. Or I was thinking of just getting a local bar/restaurant to let me put a big auraphone sign outfront and show how that's a draw to get more customers.


Some vim user too! oil.nvim is a pretty popular package that does this.


I obviously love oil.nvim and that's why I ported it to vscode. But I think in some ways voil is even more powerful than oil. Specifically:

- It can work across multiple vscode windows

- The top line (that shows the current directory) can be used to filter files. For example, if you add "*.{txt,md}" to the end of that line, it will only show the txt and markdown files.

- The ability to defined custom shell commands and bind keybindings to them. For example, I can create a command that zips selected files and run it with a single keybinding in voil.

- Undo functionality


I have largely the same setup but use neovim instead of obsidian.

A friendly tip: you don’t have to populate the metadata yourself and can use ‘hugo new <dir>/<post_name>.md’ to create the file with the metadata.


Nice, didn’t know that. Will try it out, thanks!


I've recently been prototyping a mobile application to track your food nutrition. The key feature lies in auto-detecting the food based on a given image, and breaking it down into it's ingredients and then into it's macros.

Existing apps such as MyFitnessPal and HealthifyMe fall into two ends of the spectrum where you either need to add ingredients one by one, or your food is logged with a standard macro count where you cannot change the ingredients used.

Weit ideally provides a seamless experience in taking a picture to retrieving ingredients to retrieving macros per ingredient. Once that's sorted, food tracking should be granular enough to build intelligence around it to improve one's diet based on their requirement.

Honestly, I used to constantly struggle with the realisation that none of my ideas are unique and whenever I see someone having built something similar, I feel like I'm wasting my time. I'm getting better at dealing with it now though.


This is exactly the app I would've wanted when I was still trying to lose weight by tracking calories, really great idea. I paid for a calorie-tracking app that couldn't do this, I think it was MyFitnessPal, and know several people currently paying for the same type of thing.


That's good to know!

Thanks for sharing, it feels nice to know that people actually have an interest in the concept.


Not sure if you are already familiar but I use Cal AI and think it does a pretty decent job. How granular do you track the macros? hard to figure out from a picture if I used some cheese or how much oil.


Indeed, I intend on having the user take over after identifying the food that we're working with.

The idea is to reduce the amount of redundant work (clicking a button to add a single ingredient over and over again), and rather push for minor modifications instead.


CalAI is a scam. It's highly inaccurate, there's no way to determine calories from a picture for a multitude of reasons


works fine for my use case and found it alot better than myFitnessPal, ymmv and yeah of course it's impossible to determine with certainty but it gets you 80/90% of the way there.


if you cooked a meal with a ton of butter or oil, it would not be able to detect the quantity. it could possibly undercount by hundreds of calories, making it useless. if you're not serious about the counting then it could be fine


that's what I mentioned above, read again


Not to be pedantic. But being able to be 100s of calories off per meal is not 80-90% there.

I'd be surprised if it came close to the accurate amount of calories even 50% of the time. If by there you are talking about accuracy.

How would it know how many calories my curry has from a photo? It cannot. It could be full of oil or have none.

How can it know how big the bowl of food is? It cannot.

It cannot know the actual fiber content of any baked good just by looking at it.

The promotional photo gives 7/10 nutritional score of pancakes with syrup? Why? because it has a couple of blueberries sprinkled on top? I'm sorry, but this does not seem like a useful app.

The promotional text on their site says: "It is consistently 90% accurate when predicting nutritional value of food" They might as well say 100%. They get to cherry pick the food they photo. It would be impossible to be 90% accurate if they actually included similar looking foods with different fat contents. Mash potatoes made with butter, without butter. Mash potatoes in a big bowl vs small bowl.

The site just reeks scam.


I invested in http://usenourish.com (a YC company) that recently launched this. If you want to work on this for them, happy to intro.


Just went through the company's website, and I like how well the software ties into the company's motive/direction.

I'd appreciate an intro, thanks! I could send an email, or you can find mine on my bio :D


I thought about this too, if you need any help let me know, email in bio.


Thanks! I'll send you an email sometime if I make any progress on it! :D


this idea has been rinsed and repeated many times already. unless you're going to include manual input in addition to the image recognition, it will be highly inaccurate


Oh yeah, manual input is going to be a big part of it. The system should ideally try to map the image to a standard recipe with the ingredients and then hand over control to the user to make any modifications they like.

Computing the nutrients from the ingredients should be reasonably accurate post this because that has a direct mapping in the USDA FoodCentral dataset.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: