Have you tried VS Code? It's a text editor for coding. You can soup it up to be an IDE but it doesn't start that way. And you'll have a better time coding on Windows.
Not really on topic but since you bring it up: Bridgebase.com, a big online site with some expert and even world-champ players, has some games with very good AI players, which can bid and play well enough to give you quite a good challenge. Most of their games are free. Try it out!
The problem for the moment is that the tools are one size fits all. That is, when you start working with one, it's not feasible yet for you to "generate a fork colleague" where all the history, discussion, and attempts to clarify are not just another shot at re-prompting, but in fact part the of personalized, even project-specific fork-shaping reinforcement learning process, the way a new human assistant colleague would gradually change by learning over time how to give you in particular what you want based on how you phrase your instructions, and tailored to the needs of the particular endeavor. Humans are kind of like contingent forks of particular potentialities, developing per iterations and experiences and adapting to improve in the particular niche or context in which they are employed. You don't need "Claude 4.0", you need Claude 4.0-Kling- seminar-project fork. Maybe that's a thing already, but I suspect it's currently really expensive. Then again, human forks are also really expensive.
While this line came from Claude, which gave me quite a laught to read, it could easily have come from any of a thousand students I have worked with over the years:
“I’ve forgotten everything that we did the last several hours. Can we start all over?”
Like all of us, I have way more to read each day than I have time for, yet I make time for every one of your posts. I've enjoyed following along as you challenge yourself with this project and I find it interesting, inspiring, and today I find it gave me a good laugh.
Might this be a signal for the $200 GPT plan? My understanding is that the context window is much much larger, Ben Thompson seems to be a big proponent of it.
I’m also a Claude fan, but it has the smallest context window of the big three right now. I have been using the paid Gemini a lot more often lately and am surprised at how good it has gotten. I would recommend giving it a quick try just to see if it meets your needs, since I believe its context window is something like one million tokens.
I am similarly frustrated with Claude, as I would prefer to be working with it right now on a bigger writing related project but it also always runs out of context window before I can get much done. If it would match Gemini’s context size, Claude would be my runaway favorite.
You might also try getting Gemini to write the code and have Claude define the characters or some other way to use them both at once to reserve how much capacity Claude uses up. I can make that work sometimes, in certain circumstances.
If you *don't* know Vim and need to be sold on it, a few years ago I put up this site: https://vimforeveryone.com -- it actually opens with a quote from you on it.
I haven't really tested my plugin on Windows and there's a learning curve to Vim (though I also wrote a book on how to learn it -- https://techtoolsbook.com/) so probably a long shot in your specific case but if you/the readers ever do want to go down a rabbit hole it's a pretty powerful tool.
Have you tried VS Code? It's a text editor for coding. You can soup it up to be an IDE but it doesn't start that way. And you'll have a better time coding on Windows.
Not really on topic but since you bring it up: Bridgebase.com, a big online site with some expert and even world-champ players, has some games with very good AI players, which can bid and play well enough to give you quite a good challenge. Most of their games are free. Try it out!
The problem for the moment is that the tools are one size fits all. That is, when you start working with one, it's not feasible yet for you to "generate a fork colleague" where all the history, discussion, and attempts to clarify are not just another shot at re-prompting, but in fact part the of personalized, even project-specific fork-shaping reinforcement learning process, the way a new human assistant colleague would gradually change by learning over time how to give you in particular what you want based on how you phrase your instructions, and tailored to the needs of the particular endeavor. Humans are kind of like contingent forks of particular potentialities, developing per iterations and experiences and adapting to improve in the particular niche or context in which they are employed. You don't need "Claude 4.0", you need Claude 4.0-Kling- seminar-project fork. Maybe that's a thing already, but I suspect it's currently really expensive. Then again, human forks are also really expensive.
While this line came from Claude, which gave me quite a laught to read, it could easily have come from any of a thousand students I have worked with over the years:
“I’ve forgotten everything that we did the last several hours. Can we start all over?”
Like all of us, I have way more to read each day than I have time for, yet I make time for every one of your posts. I've enjoyed following along as you challenge yourself with this project and I find it interesting, inspiring, and today I find it gave me a good laugh.
Might this be a signal for the $200 GPT plan? My understanding is that the context window is much much larger, Ben Thompson seems to be a big proponent of it.
Yes. All of this. https://x.com/anecdotal/status/1930595579446579678
Only remembers stuff its owners can sell (or insider trade on). Like search engines 😊
I’m also a Claude fan, but it has the smallest context window of the big three right now. I have been using the paid Gemini a lot more often lately and am surprised at how good it has gotten. I would recommend giving it a quick try just to see if it meets your needs, since I believe its context window is something like one million tokens.
I am similarly frustrated with Claude, as I would prefer to be working with it right now on a bigger writing related project but it also always runs out of context window before I can get much done. If it would match Gemini’s context size, Claude would be my runaway favorite.
You might also try getting Gemini to write the code and have Claude define the characters or some other way to use them both at once to reserve how much capacity Claude uses up. I can make that work sometimes, in certain circumstances.
I don't suppose you know Vi/Vim/Neovim -- if so I (and Claude) wrote a plugin that helps deal with this problem:
https://github.com/nathanbraun/nvim-ai
If you *don't* know Vim and need to be sold on it, a few years ago I put up this site: https://vimforeveryone.com -- it actually opens with a quote from you on it.
I haven't really tested my plugin on Windows and there's a learning curve to Vim (though I also wrote a book on how to learn it -- https://techtoolsbook.com/) so probably a long shot in your specific case but if you/the readers ever do want to go down a rabbit hole it's a pretty powerful tool.