vlad_ungureanu
Joined 9 years ago
Karma
28
Submissions
12
Member since
May 9
Recent submissions
Comment · 1 month ago
Comment · 1 year ago
Great insight into the real applications of LLMs. Also interesting to see that you find the Opus model being more successful for app development. In my experience, a combination of ChatGPT with Anthropic models yields the best results in coding.
Out of curiosity, how useful do you find the thinking mode in Claude Code?
Good luck with the PDF app!
Comment · 1 year ago
Unfortunately the article is behind paywall :(
Comment · 2 years ago
This is amazing, thanks for sharing!
Comment · 2 years ago
All the tips given in the article are very useful, but for me it was also useful learning how to make different types of jam or cakes helps you understand the amount of sugar and butter used. Replacing store-bought sweets with homemade options gives you a sense of appreciation and a 'ritualistic' aspect when enjoying something sweet, which may reduce how often you indulge in sugary treats.
Comment · 2 years ago
I found the documentation for networkx much better than the one from igraph[1] (at least the Python version). However, for community detection algorithms graph-tool[2] is better (it also uses a different class of models than the standard in literature)
Comment · 3 years ago
Interesting, I always thought that research and startups are very similar. Where you have something (product/research-idea) which you think is novel and try to sell it (journals/customers).
The management skills which you potentiated differentiated the success of the two firms. I can see how the lack of this might be wildly spread out in academia.
Comment · 3 years ago
How the World Thinks: A Global History of Philosophy - Julian Baggini [1]
The author does an amazing job presenting the different views about the world and some important differences between cultures. If you are like me and haven't interacted that much with other schools of thoughts outside of the Western world, but are interested in learning more about them, you'll enjoy the introductions to Indian thinking as well as Chinese, Islam and some African philosophies. Also, he does a good job of highlighting some of the limitations of each, including Western thought.
Comment · 4 years ago
If you have knowledge both in biology as a physician and can code, you can also become a bioinformatician, where you can still apply ML. Maybe the difference will be that it will be more on the application side of the ML rather than the theoretical part.
Also, I will be cautious about "... landing a PhD position at a good US school", from my experience and others, I've found out that it's more important to look after a supervisor with whom you can resonate rather than chasing a reputable institution.
(Disclaimer, I'm a PhD student in the UK)
Comment · 6 years ago
My grandparents are in their early 80s or late 70s and aren't very open to learn new technologies.
For the past 7 years they used and iPad (one of the early models 2nd or 3rd generations) and it served them well. I've introduced the the family contacts, put FaceTime and Photo library on the first screen and it seems ok for them. I've also annotated what volume button do and teach them that if they have a problem with it to press the big button (home button).
The only problem with the iPad is that recently there are some connectivity issues (it seems it doesn't always stay connected to Wi-Fi and they don't receive my FaceTime calls). I think this may be due to the old iOS it's using (v10 or v9). So, considering to buy them a newer version.
I started it as a quick proof of concept while my wife was in labour with our first child, partly to cope with the stress, partly I wanted to create a personal gift - tracking our journey together. So I needed, a simple way to caption the important moments of our life together with the when & where.
The goal was simple and fast, with everything on the device: no account, no tracking, no uploads. You can edit the captions and adjust the crop, then export or share. The first 5 photos are free, then a one-time $1.99 unlocks unlimited exports.
App Store link: https://apps.apple.com/app/id6753585464
Side note: This side project was also a second experiment: how far can LLM coding agents get on their own? I wrote the first version in Swift, then had Claude Code port it to Flutter. It did most of the work (rendering, geocoding, the in-app purchase flow, even driving the simulator to check its own screenshots). I think Opus 4.8 was the point when the agent become truly autonomous and my input became less needed.