youtu.be/ShusuVq32hc?...
I think this was a good and interesting presentation. Note: I’m not against AI or LLM usage in any way. I’m using it a lot myself. But it’s interesting reflect about “reasoning” and “thinking” being used in the context of LLM models. Isn’t that misleading in a way?
~5 months later: OpenAI has now announced starting internal testing of ads for Free and Go adult users in the US. Will be interesting to see the experience of this (2nd link shows examples of what it could look like).
help.openai.com/en/articles/...
openai.com/index/our-ap...
I wonder when we’re going to start to see ads in chatbot interfaces (ChatGPT, Claude etc). It’s not a question about _if_, rather _when_, or what would do you think?
Today I learned about Git submodules, i.e. keep a Git repository as a subdirectory of another Git repository. I guess it was time to learn about that after using git for over a decade. 😅
I used it to add my forked kickstart.nvim repo to my dotfiles repo: github.com/mansn/dotfil...
Especially the fact that Agent mode in Copilot Chat uses 1 premium request per user prompt, i.e. same as a chat message when set to "Ask". I would assume that the token usage differs quite a bit on those two modes, but maybe I'm wrong on that
Do you still feel the same? =)
I think it's going to be very interesting to see how the pricing plans and limits evolve over time. I wonder if this premium request multiplier pricing model feasible in the long term..
Friendly people also want to be helpful and fix what seems to be a typo, but then realize that it was actually intentionally added: github.com/npm/cli/pull...
=)
I just learned that there are quite a few "typo friendly(?)" aliases for npm CLI commands, such as `npm udpate` which is an alias for `npm update`😃
That specific alias was actually added in mid 2017 (8 years ago! time flies): github.com/npm/npm/pull...
Seems like it's common in other countries as well
Random post: About a week ago I learned that it's common for french people spell their last name in uppercase letters. I've seen it a few times, but never really thought about wh. =)
One reason is to avoid mixing up the firstname and lastname.
Started tinkering a bit with Neovim with kickstart. =) I didn't think I would be excited about configuring things and exploring it, but I keep going back to it. I use VSCode (with the Vim extension) on my job, so it's a bit slow to build a habit with neovim on my free time.
I also use the Vim extension (vscodevim), and I had to go into insert mode to be able to accept a next edit, but that wasn't the case with Cursor as far as I know (I'm not paying for Cursor atm). It makes sense that it requires insert mode, but I think it would be fine not requiring insert mode.
I've used it for roughly a week now, and initially it felt pretty slow compared to Cursor. I kind of like the UI when the "suggestion" pops up, but it also feels a bit clunky. Hard to explain. Cursor felt less fancy (no/less animations), but faster.
Styled components has officially moved into maintenance-only mode: https://opencollective.com/styled-components/updates/thank-you
That's the bell I think, the era of runtime css-in-js libraries is now definitively over. So long and thanks for all the divs.
I haven't used the new "next edit" functionality that VS Code should have now. I hope it's as smooth and fast as Cursor, if not even faster/better!
www.youtube.com/watch?v=duty...
I've been running Cursor a bit on my free time, and been impressed by the speed of the tab autocomplete. But I wonder if VS Code will win me back with the latest updates. I'm not a big user of Agent Mode, but I like autocomplete/tabbing to speed up my output =)
"The latest image does not contain any text. If you see any unintended text, let me know, and I can refine it further! 🎈🎂🎉"
😃
Tonight I used ChatGPT to generate a party poster for my son's birthday party. It had a hard time to understand what I meant with not having any text in the poster 😂
Check 1,2,3
