All these tools are straight garbage. I’ve been using windsurf for 3 weeks now and it constantly makes components that are already established and integrated into the framework of the project instead of identifying that the components exist already it just starts making new ones with completely generic names and adds and deletes files without permission and changes design language etc etc etc and the list goes on. Even if you define rules for it to not do what it just did, it will a few steps later do exactly what you just instructed it not too. And as far as all the other tools, yeah it might be helpful to visualize a basic data structure for the project or give non functioning components or cards to use within a project But FORGET ABOUT having it build you anything that will work. These are simply reference tools at best and even then most of the time just completely cannibalize code that you aren’t even focused on. Like you’re working on one specific container within a page of the stack and it will go off and edit code on other pages in your stack deleting headers components or whatever else it does without any explanation or approval and then breaks the project leading to hours of trouble shooting and tracking down what happened and eventually Just having to hard reset your last working commit and start all over again. These assistants just aren’t there yet whatsoever. Better just to learn how to code and do it yourself. These tools seem to be subscription based harvesting profit tools by whomever is releasing them and marketing them. Honestly. It’s insane how much time I’ve wasted using these products
So just write them all off and give up? Or acknowledge this first step, fumble around in it for a bit, and pay close attention to what comes next? Isn’t forward progress inevitable with these tools?
@@CoachKenWins well since we’re off to the races in the ai IDE world, I assume these companies will keep shilling half baked products since each company seems to want to one up each other back and forth without improving the fundamentals of their core concepts to truly make an environment and service that is reliable enough to even trust with your code base. I’m not saying they will NEVER get there.. but it seems to me that a lot of these products are truly a money grab. Sure if you want to toy around with them to dive into some basics, the one nice thing is that you can have the assistant write markdown documentation for you that you’re able to reference and possibly learn from so it does have potential in helping somebody with 0 experience and knowledge learn about many key concepts and practices .. but the models need to be much more refined, run locally to actually be context aware in order to maintain flow state with the developer. I’d be ecstatic if these tools worked 50% as good as how they’re portrayed and advertised. And obviously you can tell the companies are more focused on brand marketing than making the products work, instead of paying content creators to shill non stop and adding attractive sounding features to entice consumers into subscribing to their services.. only to find out that it’s not as magical as the TH-cam video made it sound. Literally. This is the scenario. Yes they’re in their infancy, but you’ll quickly find out that you’re better off running your pocket books into the ground using your own api keys to take full advantage of more advanced models. These companies Charging 20$ a month just isn’t realistic or mathematically possible to overcome the downsides within the more practical use cases for these conceptual IDEs which cannot offer a stable experiences for that price… quality > quantity.. and quality is being ignored to make a quick buck right now. That’s my opinion. I’m sure the environments will improve overtime, but definitely not in this price range. It will cost much more to deliver very robust end to end environments that work as advertised. You’re better off using private apis with Claude/GPT .., or using 7b, Qwen or similar open source models directly and implement examples of the code you generate than to trust some generic chatbot that cannot have consistent or manageable behavior and has extreme limitations in regards to quality output and reliability in order for a developer to actually trust using these current tools in real world use cases and larger code bases. If paying money to endure frustration is your thing.. hey .. by all means you do you and enjoy your time tinkering. But if you spent all that time and energy… and tried to understand these tools are broken tutors at best and to only use them from time to time as a point of reference or maybe get a few basic ideas generated for you, you’d be better off at the moment. They can be useful in some capacity but you have to know how to get funky with the machines and not let them get funky with you.
Matt, You should create (with the community) and host the Yearly Prompt Awards on World convention for Ai & Prompt fans & enthusiasts. Let's create and go to a world convention , every year organize in a continent where Prompt fans come together, meet, exchange, create, laugh, cry and hold local day competition . the P3 - games (instead of E33 games) . And we hand out awards for the most beautiful Prompt? Who would go? Award for best Image Award for best Sound Award for best Video Award for best Painting Award for best Code Award for best Prompt (In honor of robin Williams)
Can't happen fast enough for me! We just lost our programmer whom passed away from Liver Cancer. It was a long term relationship and we worked together with 100% complete trust. You know how hard that is to build? Starting over going thru dozens of programmer wanna-bee's isn't anything I'm looking forward to.
bolt is only good at frontend or first few prompts. after that it's starting to hallucinate. last thing you know, your 10m tokens is gone just debugging a very simple bug.
It gonna happen someday, so its normal for everyone, it need some time, but as soon as everyone knows you can create Software/Web Applocation by simply explain your problem in one or two sentence@@drumbassclassics
cursor is good for the backend, sucks for the u.i ... / frontend , bolt ,vo , etc are good for the u.i but suck for the backend.. For now .. All are somewhat limited .. Shty apps tho u can make but most already exist..
Thanks for watching!! Don't forget to grab the free guide: build your own Custom GPT 👉 clickhubspot.com/wpr
Thanks for the shoutout Matt
enterprise-ai AI fixes this (Code complete projects in PHP or Python). Testing four AI coding tools.
All these tools are straight garbage.
I’ve been using windsurf for 3 weeks now and it constantly makes components that are already established and integrated into the framework of the project instead of identifying that the components exist already it just starts making new ones with completely generic names and adds and deletes files without permission and changes design language etc etc etc and the list goes on.
Even if you define rules for it to not do what it just did, it will a few steps later do exactly what you just instructed it not too.
And as far as all the other tools, yeah it might be helpful to visualize a basic data structure for the project or give non functioning components or cards to use within a project
But FORGET ABOUT having it build you anything that will work. These are simply reference tools at best and even then most of the time just completely cannibalize code that you aren’t even focused on. Like you’re working on one specific container within a page of the stack and it will go off and edit code on other pages in your stack deleting headers components or whatever else it does without any explanation or approval and then breaks the project leading to hours of trouble shooting and tracking down what happened and eventually
Just having to hard reset your last working commit and start all over again.
These assistants just aren’t there yet whatsoever.
Better just to learn how to code and do it yourself.
These tools seem to be subscription based harvesting profit tools by whomever is releasing them and marketing them.
Honestly. It’s insane how much time I’ve wasted using these products
So just write them all off and give up? Or acknowledge this first step, fumble around in it for a bit, and pay close attention to what comes next? Isn’t forward progress inevitable with these tools?
@@CoachKenWins well since we’re off to the races in the ai IDE world, I assume these companies will keep shilling half baked products since each company seems to want to one up each other back and forth without improving the fundamentals of their core concepts to truly make an environment and service that is reliable enough to even trust with your code base. I’m not saying they will NEVER get there.. but it seems to me that a lot of these products are truly a money grab. Sure if you want to toy around with them to dive into some basics, the one nice thing is that you can have the assistant write markdown documentation for you that you’re able to reference and possibly learn from so it does have potential in helping somebody with 0 experience and knowledge learn about many key concepts and practices .. but the models need to be much more refined, run locally to actually be context aware in order to maintain flow state with the developer. I’d be ecstatic if these tools worked 50% as good as how they’re portrayed and advertised. And obviously you can tell the companies are more focused on brand marketing than making the products work, instead of paying content creators to shill non stop and adding attractive sounding features to entice consumers into subscribing to their services.. only to find out that it’s not as magical as the TH-cam video made it sound. Literally. This is the scenario.
Yes they’re in their infancy, but you’ll quickly find out that you’re better off running your pocket books into the ground using your own api keys to take full advantage of more advanced models. These companies Charging 20$ a month just isn’t realistic or mathematically possible to overcome the downsides within the more practical use cases for these conceptual IDEs which cannot offer a stable experiences for that price… quality > quantity.. and quality is being ignored to make a quick buck right now.
That’s my opinion. I’m sure the environments will improve overtime, but definitely not in this price range. It will cost much more to deliver very robust end to end environments that work as advertised. You’re better off using private apis with Claude/GPT .., or using 7b, Qwen or similar open source models directly and implement examples of the code you generate than to trust some generic chatbot that cannot have consistent or manageable behavior and has extreme limitations in regards to quality output and reliability in order for a developer to actually trust using these current tools in real world use cases and larger code bases.
If paying money to endure frustration is your thing.. hey .. by all means you do you and enjoy your time tinkering. But if you spent all that time and energy… and tried to understand these tools are broken tutors at best and to only use them from time to time as a point of reference or maybe get a few basic ideas generated for you, you’d be better off at the moment. They can be useful in some capacity but you have to know how to get funky with the machines and not let them get funky with you.
Windsurf?
I have tried n8n, Replit, etc, and with Windsurf has been the only time I build actual working apps in python
Matt, You should create (with the community) and host the Yearly Prompt Awards on World convention for Ai & Prompt fans & enthusiasts.
Let's create and go to a world convention , every year organize in a continent where Prompt fans come together, meet, exchange, create, laugh, cry and hold local day competition . the P3 - games (instead of E33 games) . And we hand out awards for the most beautiful Prompt? Who would go?
Award for best Image
Award for best Sound
Award for best Video
Award for best Painting
Award for best Code
Award for best Prompt (In honor of robin Williams)
i love this idea, imagine how many new ideas would happen on this types of events
@ thx
Good stuff, thanks for testing and sharing your findings!
So could any of these help with backend program on Solana to do swaps ie... SOL to JUP for instance?
You have to tell it what framework to use if you want a nice looking webiste
Can't happen fast enough for me! We just lost our programmer whom passed away from Liver Cancer. It was a long term relationship and we worked together with 100% complete trust. You know how hard that is to build? Starting over going thru dozens of programmer wanna-bee's isn't anything I'm looking forward to.
Bit surprised there was no comment about these apps cf ChatGTP and Claude
bolt is only good at frontend or first few prompts. after that it's starting to hallucinate. last thing you know, your 10m tokens is gone just debugging a very simple bug.
Now it's time for Devin lol
What happens to big business and capitalism when EVERYONE can create custom software just by talking to their web browser?
Most people won’t.
It gonna happen someday, so its normal for everyone, it need some time, but as soon as everyone knows you can create Software/Web Applocation by simply explain your problem in one or two sentence@@drumbassclassics
@@drumbassclassicsits gonna be just as normal as whatsapp
Right! Also building is one thing, deployment, and support is also a different beast.
What happens to restaurants when everybody has an oven, when everyone can buy prepared meals….
Why not test Codeium Windsurf ? and also Aider
❤❤
cursor is good for the backend, sucks for the u.i ... / frontend , bolt ,vo , etc are good for the u.i but suck for the backend.. For now .. All are somewhat limited .. Shty apps tho u can make but most already exist..
Wym backend, I'm tryna create a masterpiece on the fly lol
Why aren't you or anyone else talking about Kline? It's by far the best end to end IMO
Kline?
@@Mitra3050 Probably meant Cline.
Tried but not as good as cursor, windsurf @@haroldpierre1726
what about deepseek
Wow you just stole Matt Wolf's Channel.
You could just have made an easy chrome extension that download the images as jpg .. Direct conversion.. Lol dunbaf