Did OpenAI Just Kill Cursor? Let's Check Out ChatGPT Canvas for Coding
Вставка
- Опубліковано 13 жов 2024
- Learn to code with Scrimba v2.scrimba.com...
Check out topmate.io/dor... if you're interesting in chatting with me about anything!
My Resume & Cover Letter Bundle (sowl.co/s/Rfqsd)
Join my Discord ( / discord )
My Amazon store: www.amazon.com...
Did not know johnny sins was a coder
🤣
Had to google Johnny Sins. Think he looks familiar but didn't know the name, but I could probably name a lot of his "co-stars"
You're surprised? I mean he already is a Doctor, Personal Trainer, Message Therapist, Pilot, and more.
Haha lolita
@10:55 it went line by line because it was updating references to said refs. You can see this most easily on Line 9 when it passes by.
V-model="prompt" || V-model="state.prompt"
You have to specify "local editing" or "don't rewire the code", when you want it to make changes. Then it will only change the necessary code.
E.g
"Can you fix that bug?
Local edit only"
Muchos gracias gonna try this when I get home
I love Cursor! Though I do see your point that it looks more enjoyable to just chat with an ai about it on a separate chat. The only thing is that cursor can switch between models and you can use llama 3.2 offline so that’s nice if you’re on a plane or something.
18:58 best part lol.
I believe this isn't a Cursor killer but moreso a response to Artifacts. Although I may be wrong. Probably a response to both. OpenAI is selling the shovels so they probably will have their own code editor built in eventually that is far better than this. But I like it. I still use Cursor. They all have their pros and cons. There isn't just one good model anymore.
the big head startled me LOL. I had you on full screen!
hahaha 😂
Also there is File diff in cursor (which isn't present in openai Canva) to notice when the LLM is adding / changing variables names across a file for example ( which breaks a full app.) - or pieces of code it was not intended to.
I am enjoying cursor. It works well, and bugs I experience one day are fixed the next. If it was more agentic, then I would treat it as a team of colleagues. For example, if it could execute tests, and check test failures, and fix them until it got stuck.
Just ask it on the left side to add comments based on your preference and don't use the icon then.
This is not really a "cursor competitor". It's more of a Claude Artifacts competitor. It's not an IDE with extensions, etc and it doesn't run your code. To develop with more control you need to use an IDE still.
something to be said about doing one thing, and doing it really well.
i like cursor because certain models are better at things than others, i can switch, i will always want that as long as competition keeps up with eachother. openai can never launch a editor that brings in other models, and they can never have it as a priority, alongside a billion other projects they have not to mention the whole AGI narrative. Cursor can, its their main product.
also OpenAI seems to like to try and compete with companies, using their gpt4 models, that have sufficient traction by building a proprietary openai version, doesn't really signal much confidence to the market if they want people to build big products on their APIs. you build an ecosystem by cutting your arm off and letting the market make some money with it with ultimately feeds back to them in form of PR and well money. if they rebuild anything that gets traction...
not to mention claude's models are far superior (saying that as a heavy user of most models in non trivial software development) for coding
this is all moot if AGI is achieved
Thank you. Was informative :)
You can ask the model to decrease the amount of comments.
kudos Dory!
Cursor AI is SO full of bugs, that I had to stop using it. Deleting existing code; saying it updated my code but no actual changes made; stray curly braces that break the whole page;and and many other issues I've run in to. To add insult to injury, it has a team behind it that seems severally understaffed as their email response is basically nonexistent, yet boasting on their blog they had a 60 million investment. You have 60M but can't hire more staff? Suffice to say that I canceled my subscription.
Totally. It sucks. It was a pain to remove from my system as well and interfered with regular VSCode. Good riddance.
@@jumpstar9000 In all fairness, after testing ChatGPT 4o with canvas I have to say it also rewrites my code and removes entire functions. I was working on a 330 lines JavaScript file and after about 20 back and forth messages it reduced it to a mere 200 lines, removing existing functionality. And this happened after I let it explain to me how it was different:
"Here’s how it differs from traditional ChatGPT:
Full Function Rewrite: In most cases, I rewrite the entire function to include both existing and new logic. This ensures all existing behaviors are maintained while incorporating any necessary updates.
More Contextual Understanding: The canvas feature retains the complete document context, allowing me to verify all dependent functions and their roles. This helps prevent any breaking changes."
...which turned out to be false. The model STILL forgets, and does so incredibly fast. Or so it seems, but according to itself, it needs to memorize each message as well as the full code state in that particular moment, so, understandably it adds up quickly to where it no longer has sufficient memory and removes the earliest context in order to store the latest context.
Sorry for my stupid question. I am currently learning how to code and have been for almost two months, this AI stuff scares me in the way that i feel like if i wouldn't learn fast enough to get a job fast enough i will never be able to get a job. I mean, us, junior and entry level developers wouldn't stand a chance in the way of AI. And only experienced dev would be able to get jobs. Could someone explain to me if it is still worth it and it will be worth it to learn how to code. Thank you in advance, and sorry if my concern is unfounded, but i genuinely am scared
I don’t know about this cursor got a better experience
How to make him finish the code, it just is ending where he want later, especially for codes longer than 300 lines...
Also I'm giving him the code and he is rewriting what i suggest, then he just finish somewhere, not on the end, that is a no go now, worse than earlier in chat mode...
15:59
When you said "port it over to html"
I literally thought "that sounds like the birth of another JavaScript framework 😂" like you write the code in your preferred framework and then have AI write out the pure html
At the to right there is also a history function to step back …
I really hope all those existing coding assistants go away. It really bugs me there are so many of them and they don't really work with my collaborative workflow. Also, forking VSCode really is a bad idea, and now loads of people are forking it, which is nuts. Cool video, thanks!
I wouldn't say kill, but I would say, gave a warning
You can ask for an empty canvas and paste the Vue code there…
good review
Canvas can't handle a large script so it is limited right now.
AI is a calculator for mathematicians that's getting out of control xd
the biggest W is PHP being treated as a first class language. Nice one.
How can you be late this within a few hrs. This whole thing is moving too fast.
Collaboration is great. But can they handle the dependence on the relying on logged in users. They always have reliability issues
Wonky as F... LoL, U should test to run that vue.js.. I am sure it killed the thing on its first prompt already xD
I find it very funny how little the model gives a damn about what U ask him haha
Just tell it to write fewer comments
Exactly! Common sense. Better prompts, better output; it's not hard.
I imagine you could simply tell it to be more sparing with the comments. It's not all going to be buttons.
Suggestion for 30+ coders.
Vs code + claudedev could be the best thing ever if it wasnt for the limit for some reason cursor claude 3.5 can't fix code properly for shi
Awseome
If your asking an LLM to add comments to your code, you need to write better code.
Ask it for less comments then
Johnny sins.
Pretty terrible if you're comparing to cursor
I think cursor is better
Clickbait… cuz no
Replace what, an AI-powered code demolisher? Because hays what Cursor AI is at this stage, nothing more.
@@BlackByDopeDemand Weird. I’ve used it to create some pretty advanced projects.
we dont need to see your head man lol
brother i like your video your video on dont waste your 20s is life changing i love your content peace out🕊