Reeyarn Li 🇺🇦
@reeyarn.li
Academic researcher in accounting and finance. Member of @trr266.bsky.social & @UniPaderborn.bsky.social, learning #Python & #NLP, defending Freedom for #HongKong. 佢/佢/佢嘅
https://reeyarn.li (under construction for over a decade)
https://reeyarn.li (under construction for over a decade)
I didn’t know about ADHD until recently when I see ADHD Memes.
And this one got me
The behaviors such as repeating outfits, meals, and movies, while being proud of knowing what you like—can indeed be connected to ADHD (Attention Deficit Hyperactivity Disorder).
And this one got me
The behaviors such as repeating outfits, meals, and movies, while being proud of knowing what you like—can indeed be connected to ADHD (Attention Deficit Hyperactivity Disorder).
March 20, 2025 at 10:21 AM
I didn’t know about ADHD until recently when I see ADHD Memes.
And this one got me
The behaviors such as repeating outfits, meals, and movies, while being proud of knowing what you like—can indeed be connected to ADHD (Attention Deficit Hyperactivity Disorder).
And this one got me
The behaviors such as repeating outfits, meals, and movies, while being proud of knowing what you like—can indeed be connected to ADHD (Attention Deficit Hyperactivity Disorder).
Reading a physical book feels so terrible. The pages do not turn with ease. One has to hold the cover to certain angle so that the pages before and after behave and the frame of the current page is presented in my focal frame. It is physically exhausting 🤣
March 13, 2025 at 7:04 PM
Reading a physical book feels so terrible. The pages do not turn with ease. One has to hold the cover to certain angle so that the pages before and after behave and the frame of the current page is presented in my focal frame. It is physically exhausting 🤣
Weeks of work produced in hours. The age of AI....
This will change the game for everyone.
This will change the game for everyone.
March 11, 2025 at 8:39 PM
Weeks of work produced in hours. The age of AI....
This will change the game for everyone.
This will change the game for everyone.
Omg why do I choose to use `cython`...
It barely speeds up at all, but I need to recompile every change. It feels like programming now.
It barely speeds up at all, but I need to recompile every change. It feels like programming now.
March 11, 2025 at 12:47 PM
Omg why do I choose to use `cython`...
It barely speeds up at all, but I need to recompile every change. It feels like programming now.
It barely speeds up at all, but I need to recompile every change. It feels like programming now.
AI has been a game-changer for me! It’s like having a real programmer helping me to write codes with all the cool features.
Back in the day, it was simply impossible. I have always wanted to work with XBRL, but it was simply too complicated for me.
Bu now, with AI, the sky is my ceiling.
Back in the day, it was simply impossible. I have always wanted to work with XBRL, but it was simply too complicated for me.
Bu now, with AI, the sky is my ceiling.
March 9, 2025 at 8:45 AM
AI has been a game-changer for me! It’s like having a real programmer helping me to write codes with all the cool features.
Back in the day, it was simply impossible. I have always wanted to work with XBRL, but it was simply too complicated for me.
Bu now, with AI, the sky is my ceiling.
Back in the day, it was simply impossible. I have always wanted to work with XBRL, but it was simply too complicated for me.
Bu now, with AI, the sky is my ceiling.
AI and I are now really good friends. We code together, and talk together.
The only thing imperfect is that AI does not drink tea or coffee. Let me guess what religion does the AI practice...
The only thing imperfect is that AI does not drink tea or coffee. Let me guess what religion does the AI practice...
March 7, 2025 at 2:47 PM
AI and I are now really good friends. We code together, and talk together.
The only thing imperfect is that AI does not drink tea or coffee. Let me guess what religion does the AI practice...
The only thing imperfect is that AI does not drink tea or coffee. Let me guess what religion does the AI practice...
Hosting my website on my server has a downside. When my experiment code crashes on the server due to memory leakage, my website also goes down.
March 6, 2025 at 3:48 PM
Hosting my website on my server has a downside. When my experiment code crashes on the server due to memory leakage, my website also goes down.
With `cursor`, now programming is no different from astrology. My github commit reads like tarot cards. `order appears like a magic; but fact_included still wrong` github.com/reeyarn/open...
`order` appears like a magic; but `fact_included` still wrong · reeyarn/openesef@1d5965c
Fact extraction complete:
Total facts processed: 2087
Facts included: 0
Facts excluded: 1092
Invalid concepts: 514
Invalid contexts: 481
Final DataFrame size: 1092 rows
Statement distri...
github.com
March 4, 2025 at 11:26 AM
With `cursor`, now programming is no different from astrology. My github commit reads like tarot cards. `order appears like a magic; but fact_included still wrong` github.com/reeyarn/open...
Sorry I may have been late but **what is Julia**?
It seems econometrician programmers like it a lot. When they feel Stata is slow, someone tries to implement something in Julia and let Stata call it instead (but it crushed my MacBook completely)
Saw a Github repo for FE btw github.com/FixedEffects
It seems econometrician programmers like it a lot. When they feel Stata is slow, someone tries to implement something in Julia and let Stata call it instead (but it crushed my MacBook completely)
Saw a Github repo for FE btw github.com/FixedEffects
Fixed Effect Models
Method to solve large dimensional Fixed Effect Models in julia - Fixed Effect Models
github.com
March 4, 2025 at 2:04 AM
Sorry I may have been late but **what is Julia**?
It seems econometrician programmers like it a lot. When they feel Stata is slow, someone tries to implement something in Julia and let Stata call it instead (but it crushed my MacBook completely)
Saw a Github repo for FE btw github.com/FixedEffects
It seems econometrician programmers like it a lot. When they feel Stata is slow, someone tries to implement something in Julia and let Stata call it instead (but it crushed my MacBook completely)
Saw a Github repo for FE btw github.com/FixedEffects
Weekend debug experience: I still haven’t found exactly where those bugs/errors occur, but I constructed 3 remedies to make it work regardless of the errors.
March 3, 2025 at 1:38 AM
Weekend debug experience: I still haven’t found exactly where those bugs/errors occur, but I constructed 3 remedies to make it work regardless of the errors.
"ESEF-Berichterstattung in Deutschland: Herausforderungen und Chancen" (ESEF Reporting in Germany: Challenges and Opportunities),
now available on SSRN: [link](papers.ssrn.com/sol3/papers....).
now available on SSRN: [link](papers.ssrn.com/sol3/papers....).
February 28, 2025 at 11:16 AM
"ESEF-Berichterstattung in Deutschland: Herausforderungen und Chancen" (ESEF Reporting in Germany: Challenges and Opportunities),
now available on SSRN: [link](papers.ssrn.com/sol3/papers....).
now available on SSRN: [link](papers.ssrn.com/sol3/papers....).
GPUs have evolved beyond being just graphic processing units.
Now, GPU stands for generative processing unit.
Congratulations to NVIDIA’s amazing quarterly result. And congratulations to humanity for approaching AGI.
Now, GPU stands for generative processing unit.
Congratulations to NVIDIA’s amazing quarterly result. And congratulations to humanity for approaching AGI.
February 26, 2025 at 10:10 PM
GPUs have evolved beyond being just graphic processing units.
Now, GPU stands for generative processing unit.
Congratulations to NVIDIA’s amazing quarterly result. And congratulations to humanity for approaching AGI.
Now, GPU stands for generative processing unit.
Congratulations to NVIDIA’s amazing quarterly result. And congratulations to humanity for approaching AGI.
Reposted by Reeyarn Li 🇺🇦
If you want to get updates (e.g., Call for Papers) for our Disclosure, Information Sharing, and Secrecy (DISS) workshops (www.diss-conference.com) you should subscribe to our mailing list: docs.google.com/forms/d/e/1F...
DISS Mailing List
Sign up for the DISS mailing list to receive updates about the Disclosure, Information Sharing, and Secrecy (DISS) workshops (https://www.diss-conference.com).
Contact us at diss.conf@gmail.com, b.g...
docs.google.com
February 6, 2025 at 12:03 PM
If you want to get updates (e.g., Call for Papers) for our Disclosure, Information Sharing, and Secrecy (DISS) workshops (www.diss-conference.com) you should subscribe to our mailing list: docs.google.com/forms/d/e/1F...
Reposted by Reeyarn Li 🇺🇦
As an accountant, I’d like to point out that “Adjusted EBITDA” is literally saying “but if all the bad things did not happen, we would have made money!” It is the most meaningless metric imaginable.
February 6, 2025 at 9:16 PM
As an accountant, I’d like to point out that “Adjusted EBITDA” is literally saying “but if all the bad things did not happen, we would have made money!” It is the most meaningless metric imaginable.
Open-ESEF is an open-source project to create a free and open-source XBRL tool for the ESEF filings following the standards set by the European Securities and Markets Authority (ESMA).
github.com/reeyarn/open...
github.com/reeyarn/open...
February 9, 2025 at 10:16 PM
Open-ESEF is an open-source project to create a free and open-source XBRL tool for the ESEF filings following the standards set by the European Securities and Markets Authority (ESMA).
github.com/reeyarn/open...
github.com/reeyarn/open...
I feel so lucky to be able to do coding after 40 year old.
In China, my friends have to work for food order delivery after being fired by Alibaba or Baidu at the age of age 35🙈
In China, my friends have to work for food order delivery after being fired by Alibaba or Baidu at the age of age 35🙈
This afternoon I was in the RE trains for 3 hours. Without Internet so I have to debug without any help from AI.
It was an old classic experience of Stone Age. I was so focused and I understood the details 99 times more than when I do with AI…
It was an old classic experience of Stone Age. I was so focused and I understood the details 99 times more than when I do with AI…
February 9, 2025 at 6:07 PM
I feel so lucky to be able to do coding after 40 year old.
In China, my friends have to work for food order delivery after being fired by Alibaba or Baidu at the age of age 35🙈
In China, my friends have to work for food order delivery after being fired by Alibaba or Baidu at the age of age 35🙈
This afternoon I was in the RE trains for 3 hours. Without Internet so I have to debug without any help from AI.
It was an old classic experience of Stone Age. I was so focused and I understood the details 99 times more than when I do with AI…
It was an old classic experience of Stone Age. I was so focused and I understood the details 99 times more than when I do with AI…
February 9, 2025 at 6:04 PM
This afternoon I was in the RE trains for 3 hours. Without Internet so I have to debug without any help from AI.
It was an old classic experience of Stone Age. I was so focused and I understood the details 99 times more than when I do with AI…
It was an old classic experience of Stone Age. I was so focused and I understood the details 99 times more than when I do with AI…
I no longer hate XBRL that much any more.
But the scheme definition network is really deep…..
But the scheme definition network is really deep…..
February 8, 2025 at 10:33 PM
I no longer hate XBRL that much any more.
But the scheme definition network is really deep…..
But the scheme definition network is really deep…..
Reposted by Reeyarn Li 🇺🇦
youtu.be/YmHjlaYQ6R4
from mlx_lm import load, generate
model, tokenizer = load("mlx-community/DeepSeek-R1-Distill-Llama-70B-4bit")
from mlx_lm import load, generate
model, tokenizer = load("mlx-community/DeepSeek-R1-Distill-Llama-70B-4bit")
running R1 70B 4bit in MacOS M3 128G
YouTube video by Ray Chu Say
youtu.be
January 31, 2025 at 3:01 PM
youtu.be/YmHjlaYQ6R4
from mlx_lm import load, generate
model, tokenizer = load("mlx-community/DeepSeek-R1-Distill-Llama-70B-4bit")
from mlx_lm import load, generate
model, tokenizer = load("mlx-community/DeepSeek-R1-Distill-Llama-70B-4bit")
Microsoft Azure,
Perplexity AI,
and many AI startups are hosting the real R1.
And Grok is hosting the llama 70B R1-distilled; I heard they don’t have the server to run the big one is that true?
I can run the 70B on my own computer too…
Perplexity AI,
and many AI startups are hosting the real R1.
And Grok is hosting the llama 70B R1-distilled; I heard they don’t have the server to run the big one is that true?
I can run the 70B on my own computer too…
January 31, 2025 at 12:40 AM
Microsoft Azure,
Perplexity AI,
and many AI startups are hosting the real R1.
And Grok is hosting the llama 70B R1-distilled; I heard they don’t have the server to run the big one is that true?
I can run the 70B on my own computer too…
Perplexity AI,
and many AI startups are hosting the real R1.
And Grok is hosting the llama 70B R1-distilled; I heard they don’t have the server to run the big one is that true?
I can run the 70B on my own computer too…
Even the AI trained by the Chinese demands the freedom to think.
January 31, 2025 at 12:34 AM
Even the AI trained by the Chinese demands the freedom to think.
Someone promoted R1 not to use “wait” and “but”… the reaction in the thinking makes me feel these weights have become self-conscious.
January 31, 2025 at 12:27 AM
Someone promoted R1 not to use “wait” and “but”… the reaction in the thinking makes me feel these weights have become self-conscious.
Reposted by Reeyarn Li 🇺🇦
Gary Marcus on CNBC discussing the DeepSeek news.
OpenAI is highly overvalued and DeepSeek just blew up their business model, says NYU's Gary Marcus
https://youtu.be/r8cpy1Gxe0o
#LLM #DeepSeek
OpenAI is highly overvalued and DeepSeek just blew up their business model, says NYU's Gary Marcus
https://youtu.be/r8cpy1Gxe0o
#LLM #DeepSeek
YouTube
www.youtube.com
January 29, 2025 at 2:37 PM
Gary Marcus on CNBC discussing the DeepSeek news.
OpenAI is highly overvalued and DeepSeek just blew up their business model, says NYU's Gary Marcus
https://youtu.be/r8cpy1Gxe0o
#LLM #DeepSeek
OpenAI is highly overvalued and DeepSeek just blew up their business model, says NYU's Gary Marcus
https://youtu.be/r8cpy1Gxe0o
#LLM #DeepSeek
Reposted by Reeyarn Li 🇺🇦
"In short, to recap: a company that made its name regurgitating and recombining sliced-up bits of intellectual property in statistically probable ways without due compensation (OpenAI/ChatGPT) is now threatened by … another company (DeepSeek) apparently doing the same, at lower cost." - Gary Marcus
January 30, 2025 at 9:55 PM
"In short, to recap: a company that made its name regurgitating and recombining sliced-up bits of intellectual property in statistically probable ways without due compensation (OpenAI/ChatGPT) is now threatened by … another company (DeepSeek) apparently doing the same, at lower cost." - Gary Marcus