HNHacker News
TopNewBestAskShowJobs

mromanuk

2,214 karma · joined January 26, 2012

Currently building https://LunaraAI.app a meditation app. If you want to reach me, email me: martin at romanuk

website: https://romanuk.org

submissionscomments
mromanuk··on Show HN: I'm a doctor and built a responsive breathing app for anxiety and sleep
I'm working on a meditation app, using an llm as a guide. It tracks your heart rate using the main camera of the phone, later will add breathing. Soon to be released.
mromanuk··on The last six months in LLMs, illustrated by pelicans on bicycles
The last animation is hilarious, represents very well the AI Hype cycle vs reality.
mromanuk··on Getting Past Procrastination
I procrastinate all the time, listen to much to your mind or chase the fun stuff only, would not get you traction, probably it’s a distraction, because the mind is lazy. Most of our systems wants to conserve energy or expend at little as possible. Going to the gym in a cold morning is not something that the mind or body is seeking, so listen to the idea of not going would be bad for you. Muscles are lazy too, they just want to chill. But if you make them do a little work, they like it and ask for more. We are weird and we need to force us to do stuff. That’s your job, you command your body
mromanuk··on LLMs are mirrors of operator skill
Every time I ask an LLM to write some UI and model for SwiftUI I have to specify to use @Observable macro (is the new way), which they normally do, after asking for it.

The LLM tells me that they prefere the "older way" because it's more broadly compatible, that's ok if you are aiming for that. But If the programmer doesn't know about that they will be stuck with the LLM calling the shots for them.

mromanuk··on Typing 118 WPM broke my brain in the right ways
I didn’t take typing lessons, but I’ve been typing since the 1980s, probably since 1987. At some point, I discovered that people typed without looking, decided that using 10 fingers and typing without looking at the keyboard was better, so I started optimizing for it, and it worked.
mromanuk··on A faster way to copy SQLite databases between computers
I was surprised that he didn't try to use on the flight compression, provided by rsync:

  -z, --compress              compress file data during the transfer
      --compress-level=NUM    explicitly set compression level
Probably it's faster to compress to gzip and later transfer. But it's nice to have the possibility to improve the transfer with a a flag.
mromanuk··on Qwen3: Think deeper, act faster
I was expecting a different outcome, that you tell us that Qwen3 nailed at first.
mromanuk··on C. Elegans: The Worm That No Computer Scientist Can Crack
isn't free will, more about inner control, in opposition to external control?
mromanuk··on The Model Is the Product
> This is also an uncomfortable direction. All investors have been betting on the application layer. In the next stage of AI evolution, the application layer is likely to be the first to be automated and disrupted.

I'm not convinced, we tend to think in terms of problem-products(solutions), for example editing an image => photoshop, writing some document => word. I doubt that we are going to move to a "Any problem => model". That's what ChatGPT is experimenting with the "calendaring/notification". It breaks the concept that one brand solves one problem. The App store is a good example, there are millions of apps. I find it really hard that the "apps" can get inside the "model" and expect that the model will "generate an app tailored" for that problem at that moment, many new problems will emerge.

mromanuk··on NotaGen: Advancing Musicality in Symbolic Music Generation with LLM Training
There is a demo, with several examples: https://electricalexis.github.io/notagen-demo/
mromanuk··on I Blog with Raw HTML
I used wintersmith.io. At some point it was abandoned and eventually the node version needed was old and other stuff associated with it. It became a hassle.
mromanuk··on Starlink in the Falkland Islands – A national emergency situation?
Self-determination does not apply here because the current population is not indigenous but was implanted after the UK forcibly took the islands in 1833, displacing the original Argentine population. The UN considers this a decolonization issue, not a matter of self-determination.

The argument that the islanders “chose” to remain under UK control ignores that democracy cannot legitimize colonial occupation. A referendum by settlers cannot override international law, just as a vote by European colonists in Africa in the 19th century wouldn’t justify occupation. The UK has consistently refused to negotiate, violating UN resolutions and international norms.

mromanuk··on Starlink in the Falkland Islands – A national emergency situation?
The 2013 referendum is not a valid argument in sovereignty disputes. The UN has repeatedly recognized the Islas Malvinas as a disputed territory and has urged the UK to negotiate with Argentina (Resolution 2065 and subsequent resolutions).
mromanuk··on Starlink in the Falkland Islands – A national emergency situation?
The use of the name Falklands is problematic because the Islas Malvinas are a disputed territory between Argentina and the UK. The UN has recognized this dispute and has urged the UK to negotiate their return, something the UK has ignored so far. Using the name imposed by the occupying power contributes to erasing Argentina’s legitimate claim and disregards the international legal stance on the issue. The proper approach would be to refer to them as Islas Malvinas or at least acknowledge the dispute in any discussion about them.
mromanuk··on OpenAI says it has evidence DeepSeek used its model to train competitor
At Microsoft’s size, they don’t care; they just buy out others.
mromanuk··on I've been advocating for RSS support, and you should too
I used Feedly, tried several RSS clients, I was into RSS but content providers started to only give you a portion of the content, subverting the very idea of RSS (at least what I wanted), even some clients had (or have) the feature to download the content and present it to you with a clean RSS style. So I quit it, because it was a new tech arms race. Not sure how is all that now.
mromanuk··on Ask HN: Am I the only one here who can't stand HN's AI obsession?
I’m AI-obsessed and have been since forever (mid-1990s). I’m longing for AGI, so I tend to consume every piece of content that looks like an advance or is related to AI business.

AI-related content is seasonal. This new LLM trend (which is now shifting towards “agentic AI”) will eventually fade and be replaced by another shiny new term. Basically, there’s a lot of noise, but every now and then, some gems pop up. Finding those articles is the fun part and the reason why I visit HN every day.

mromanuk··on How Britain got its first internet connection (2015)
Yes, please! I love these comments on HN, this is blog material.
mromanuk··on Coconut by Meta AI – Better LLM Reasoning with Chain of Continuous Thought?
Because group estimation is superior to individual estimations: The phenomenon is called wisdom of the crowds. When a group of people independently estimate something, individual errors tend to cancel each other out, leading to a surprisingly accurate collective result. This works because of:

Diversity of opinions: Different perspectives bring a range of estimates. Independence: Errors aren't systematically biased as long as individuals estimate without external influence. Error averaging: Overestimation and underestimations balance out when averaged. Law of large numbers: More participants increase accuracy by minimizing random errors. It was demonstrated by Francis Galton in 1906, where a crowd's average guess of a bull's weight was almost spot-on. (estimates must be independent and reasonably informed for this to work.)

mromanuk··on 'Conan the Bacterium' is radiation-resistant for a surprising reason
> When dried and frozen, Deinococcus radiodurans could survive 140,000 grays, or units of X-and gamma-ray radiation, which is 28,000 times greater than the amount of radiation that could kill a person.

I would love also if they called it Hulk the Bacterium

mromanuk··on Scientist treated her own cancer with viruses she grew in the lab
Interesting point, though I think we should also consider that our ability to understand diseases like cancer has historically been limited by the observational and analytical technology available. For example, it was recently suggested that Alzheimer’s might be linked to Candida albicans, a fungus that naturally inhabits our bodies but could play a role in the disease. Just as it took time to discover that iodine deficiency causes goiter or that H. pylori produces ulcers, the complexity of cancer might partly stem from the fact that we don’t yet have the necessary technology to closely observe the cells and underlying mechanisms. It’s not necessarily the inherent complexity of the problem but our technological limitations that delay understanding—and possibly a cure.
mromanuk··on Scientist treated her own cancer with viruses she grew in the lab
BTW, saying that someone "is a bit of a quack" is an ad hominem fallacy. The article linked talks about the "keto" part of the protocol, it does not discuss the glutamine portion of the treatment. I'm not sure who is right, but I would love to see, someone debunking it with data. Protocol: https://www.nature.com/articles/s42003-019-0455-x
mromanuk··on Scientist treated her own cancer with viruses she grew in the lab
According to Thomas Seyfried, cancer is indeed one disease, fundamentally a metabolic disease [0]. Seyfried's theory suggests that cancer stems from mitochondrial dysfunction, which disrupts cellular metabolism and leads to abnormal cell growth. He argues that the root cause of cancer is not genetic mutations, as commonly believed, but rather metabolic disturbances that alter how cells process energy. Basically returning the damaged cells to "old pathways" of energy generation, without oxygen: fermentation. This process, known as the Warburg Effect (named after Otto Warburg who first described it in the 1920s), shows that cancer cells primarily rely on fermentation for energy production even in the presence of oxygen - a phenomenon called "aerobic glycolysis." However, glucose fermentation is only part of the story. Cancer cells also heavily depend on glutamine, an amino acid that serves as another crucial fuel source. Through a process called glutaminolysis, cancer cells convert glutamine into both energy and building blocks for rapid cell division. This dual dependency on glucose and glutamine makes cancer cells metabolically distinct from normal cells.

This metabolic theory challenges the traditional somatic mutation theory, which views cancer as a result of DNA mutations accumulating in cells. Seyfried proposes that targeting the metabolism of cancer cells—primarily through dietary interventions like ketogenic diets or therapies that restrict glucose—could effectively "starve" cancer cells while leaving healthy cells less affected. His approach implies that a general strategy for treating cancer could involve targeting this metabolic vulnerability shared across many cancer types. Furthermore, this theory suggests that combination approaches targeting both glucose and glutamine metabolism might be particularly effective, as they would address both major fuel sources that cancer cells rely on. This could include strategies such as ketogenic diets (to restrict glucose), glutamine inhibitors, and other metabolic therapies that work together to compromise cancer cell energy production while preserving normal cell function.

0: https://nutritionandmetabolism.biomedcentral.com/articles/10...

mromanuk··on Chain-of-thought can hurt performance on tasks where thinking makes humans worse
chatGPT, from smaller to largest: Blueberry Orange Grapefruit
mromanuk··on Using an 8K TV as a Monitor
I normally work with a 40", I'm using a a hammerspoon to divide the screen, but normally I end using one main window, with some smaller window at the side and cmd-tabbing between info. How do you manage the distraction of so many information at the same time? Do you switch between apps? use the mouse? don't you loose track of where the focused window is?
mromanuk··on Show HN: Ambulate – Detailed Trip Planning
a variation of this problem is carrying your shopping cart around. A shopping list that you can buy in place A or B. Also splitting shopping wisely in both.
mromanuk··on Using AI Generated Code Will Make You a Bad Programmer
Any tool that improves any work will and shall be used.
mromanuk··on Hetzner introduces GPU server for AI training
I use runpod or vast for training my (small) models (a few million parameters) mostly using RTX4090 up to 4 GPUs. Training is a sporadic task. Is not worth it for me to book it monthly (at these prices)
mromanuk··on SwiftUI for Mac 2024
> Swift UI makes simple UIs easier, moderately complex UIs harder, and very complex UIs impossible. The moment your UI is more complex than a stack of stacks, the elegance and comprehensibly falls apart compared to autolayout. It starts to feel like CSS all over again, using trial and error to try out combinations of elements and properties with unclear interactions.

Sadly, this has been my experience. I built a moderately complex app, only to bump into more and more problems, and I find myself fighting the framework, building hacks, and implementing workarounds. When you venture into more advanced UIs, the leaky abstraction comes back to haunt you.

mromanuk··on NanoGPT: The simplest, fastest repository for training medium-sized GPTs
Yes, fine tuning is fast, cheap and "easy", you can do it in a no so expensive GPU without issues.
← PreviousPage 2 of 21Next →