Anatoly Tsyplenkov

305 posts

Anatoly Tsyplenkov banner
Anatoly Tsyplenkov

Anatoly Tsyplenkov

@atsyplen

Holding a PhD in hydrology, interested in riverine sediment transport issues and all things #geospatial. Love mountains, mystery stories and #rstats 🤎

Palmerston North City เข้าร่วม Temmuz 2018
457 กำลังติดตาม259 ผู้ติดตาม
ทวีตที่ปักหมุด
Anatoly Tsyplenkov
Anatoly Tsyplenkov@atsyplen·
Huge update to the {pastum} VS Code extension — you can now control the decimal separator (point or comma). Paste as default dataframe is now sensitive to the editor language, so you can easily paste in your favorite format with just one click. See some showcases below.
English
0
5
34
2.9K
Jainam Parmar
Jainam Parmar@aiwithjainam·
BREAKING: Claude can now research like a Stanford PhD student. Here are 9 insane Claude prompts that turn 40+ research papers into structured literature reviews, knowledge maps, and research gaps in minutes (Save this)
Jainam Parmar tweet media
English
101
1.4K
10.5K
3.9M
Anatoly Tsyplenkov
Anatoly Tsyplenkov@atsyplen·
@lino_locurcio @EmmanuelIgunza @aiwithjainam From my experience the way LLM converts PDF to md matters most. By default it probably uses very quick parsers, which can lead to information loss. Here is an example of using the marker lib to convert the Elseveir's paper. Left is quick, right is slow with postprocessing
Anatoly Tsyplenkov tweet mediaAnatoly Tsyplenkov tweet media
English
1
0
1
19
Lino L Locurcio
Lino L Locurcio@lino_locurcio·
It may be anecdotal, but it's difficult to extrapolate precise numbers from medical PDFs with free-tier LLMs. I have experience using the pro version of ChatGPT, and yes, it's mostly accurate but not always. It's certainly exciting news, but we are still far from the precision we may need.
English
1
0
0
773
Anatoly Tsyplenkov
Anatoly Tsyplenkov@atsyplen·
4/5 Since this is a bit of a hack (relying on browser automation), I am really curious about how the rest of the community is solving this. How else are you getting your agents to handle literature searches? Are you relying on Semantic Scholar/OpenAlex APIs? #academicchatter
English
1
0
0
20
Anatoly Tsyplenkov
Anatoly Tsyplenkov@atsyplen·
I still write my own papers, but finding that one missing citation during major revisions is a massive flow-breaker. To fix this, I'm proposing `scholar-labs-search` a #Codex/#Claude Code skill that uses Google Scholar Labs to find the best-fit papers for any topic. 🧵👇
Anatoly Tsyplenkov tweet media
English
1
0
0
76
Anatoly Tsyplenkov รีทวีตแล้ว
Ryzerth 🐲
Ryzerth 🐲@ryzerth·
Someone wants me to rename the Microkernel I wrote when I was a kid... Brother, I don't give a shit AIs see my decade old project more than your slop, you should have checked that before naming your shit. I'm not changing history because of your naming incompetence.
Ryzerth 🐲 tweet media
English
303
734
25.4K
1.2M
Anatoly Tsyplenkov รีทวีตแล้ว
Anatoly Tsyplenkov
Anatoly Tsyplenkov@atsyplen·
@AJThurston I am using it and it works just fine! Try to delete the zotero-library-1.json and trigger regeneration
English
1
0
1
37
AJ Thurston, PhD
AJ Thurston, PhD@AJThurston·
Anyone use Zotero references in Obsidian? Specifically, anyone using obsidian-pandoc-reference-list? Seems Zotero 8 broke it. How are you all managing references in Obsidian?
English
1
0
0
164
Luong NGUYEN
Luong NGUYEN@luongnv89·
Still feel unreal for me. 5 minutes to have Claude Code + kimi-k2.5, totally FREE. - install claude code, ollama Open terminal: ollama signin # login via google/github ollama pull kimi-k2.5:cloud ollama launch claude --model kimi-k2.5:cloud thanks @ollama @Kimi_Moonshot
Luong NGUYEN tweet media
English
1
0
0
239
Mwavu
Mwavu@kennedymwavu·
library(tidyverse) iris %>% select(Petal.Length, Sepal.Width)
Mwavu tweet media
HT
7
43
503
23.9K
Bruno Rodrigues
Bruno Rodrigues@brodriguesco·
in Germany it's "Die slowly", not "Die Hard"
Bruno Rodrigues tweet media
English
1
0
3
924
Antoine Fabri
Antoine Fabri@antoine_fabri·
This freaked me out: typeof(elmer::type_boolean("foo")) #> [1] "object" It seems that #rstats added an object type named "object" in July 2023 and I didn't know about it #L206" target="_blank" rel="nofollow noopener">github.com/wch/r-source/b…
English
2
0
4
482
Anatoly Tsyplenkov
Anatoly Tsyplenkov@atsyplen·
Ditching code autocompletion feels so freaking good 🍹. It was an illusion to think it actually speeds up #rstats writing. For the past month, my workflow has been opening the same project in Positron and Cursor. If I get stuck, I start chatting with an LLM in @cursor_ai.
English
3
0
3
511
Anatoly Tsyplenkov
Anatoly Tsyplenkov@atsyplen·
@backtobourbon @cursor_ai 1. Claude 3.5 Sonnet, which I found to be better in solving R tasks. 2. Indexing documents and referring to them in chat as regular files. 3. Adding cursor rules. While one may add the Composer feature, I don’t use it frequently.
English
2
0
2
84