the auto-completion in jellyfish is rather cool. Talking to Gemini: I asked it to give me an example to fetch data from Firebase storage. The first two solutions would not even compile. It is very polite in apologizing, but it seems weird, because one would assume that a Google AI bot would at least know how Firebase works? Sorry if I mistake the concept.
I like how it completely invents APIs that don't exist but would solve the problem perfectly lol. The chat feature from Gemini hallucinates a lot while ChatGpt usually does it better. The auto complete is nice, I just wish they would take the real classes into account instead of trying to predict the class and giving fake field names as suggestions.
I'm kind of surprise, I would not say each time i'm using gemini, but it happens quite often that when I type my code and ask some questions about what I wrote or to find a mistake or to change some part of the code, it can answer correctly. However, i agree on the fact that chatgpt is globally beter but indeed in the future i'm expected gemini to be much more better on kotlin topics.
Gemini give me right answer (Jul, 2024 ) Philipp Lackner is a well-known Android developer and content creator. He's recognized for his TH-cam channel, where he shares tutorials and insights on Android development, primarily using Kotlin. His content covers a wide range of topics, from beginner-friendly introductions to more advanced concepts and libraries. Many developers, especially those who are starting their Android journey, find his tutorials helpful and easy to follow.
Gemini gives good answers sometimes and with well explanations but I am keep on getting bad answers, sometimes totally wrong answers by it. Also after some questions it fails to remember things. I need to remind it again.. Better use it to improvise existing code, add comments, writing tests and to apply kotlin language style guides for now..:)
I have not yet tried the integrated Gemini in Android studio, but I use the Gemini webpage a lot to find answers to my questions about code, and it is usually pretty good with getting me started. If a Gemini answer does not work, I just say, "That is not correct" and then it often comes up with an alternative I can try. The good thing about Gemini is that there is, at least for me, often not an alternative to find a better answer.
Omg, its so true, gemini always give me the wrong answer, and its very confident. And when i tell it what i want and to just help me improve on my thinking, it keeps trying to sabotage me by telling me about other solutions taht are mostly wrong. I feel like gemini was modelled after a gaslighter...
I'm a beginner. Text based Gemini caused my M3 MacBook Air (16GB) to get warm, as if I were gaming or something, after only 3 or 4 queries (as well as not giving me any useful answers). Gemini suggestions kept giving me unhelpful suggestions. It got annoying over time as it would create text that isn't really there and it was just a bit confusing. So I disabled Gemini entirely. I'm using ChatGPT instead which isn't great for Android Development, but is many times more reliable than this Gemini integration.
Gemini needs to get way better. I use ChatGPT, and it explains so well a lot of things about Jetpack compose. For example, today I was trying to do apply modifications conditionally on a Modifier, so I tried to use "apply" function, and it didn't worked. So I asked chatGPT how could I apply a modification conditionally on a Modifier in the modification chain, and it told me to use '"let" function. I got confused on why let worked and apply didn't. So I asked GPT why, and it explained well that when we call a modification function on a Modifier, it creates another instance of Modifier with the modification and returns it, instead of returning the original with the modification, and that's why "apply" function didn't worked. "apply" takes the original, executes the lambda and returns the original in the final, whereas "let" returns the modified version. I think Gemini would never be able to do some good concrete explanations like this, or maybe it could, but I would not trust it because of so much AI hallucinations it is hard to believe in Gemini. Hope they fix it.
Definitely. If you stick with it trying to point out why its answer are wrong, it eventually loops back to the 1st wrong answer. I've compare to paired programming where other programmer has a head injury.
I've been using it since it was released on AS Jellyfish. I think Gemini is REALLY powerful when it comes to provide documentation for composable functions. For the rest I think it's still buggy and unreliable....
My first impressions are not that good since i use it but it's too easy to use because it's built inside the IDE. Code completion was better than i expected though
Hey, he said 1.5 mil subscribers because that's the minimum number of subscribers you should have, he was just reffering to the content quality 😁and you're almost there anyway, so no harm there
I tried the Gemini 1.5 pro and it actually works good, I even gave it an HTML CSS code to translate it to compose code, and it worked from the first time.
Mr Philip i have a serious question to you Iam working as a android java developer Since from 1 year So my concern is currently iam seeing that there are no jobs android java So should i continue on it or i shift to some another tech What do you recommend Iam waiting for your reply
😂😂for some reason i am the CEO of the largest tech school in Kenya 😂😂...it also crushes when you give it a statement that starts with a comment : //something...
Code is complex for human developers, how could it be easy for AI? If the humans who wrote the AI code made mistakes and there's bug in the code. I don't see AI building an huge app anytime soon, when we talk about apps that need performance optimizations with thousands of users a long arrayList of infinite scrolling pagination, release memory and resource it's just too much for an AI, make dicision is complicated for humans, we struggle to make decision, how can we teach machine to make decision in unpredictable and new context? AI sucks.
Congratulations on becoming the CEO of mindorks
Phillip is living a double life according to Gemini 😂🎉
the auto-completion in jellyfish is rather cool.
Talking to Gemini: I asked it to give me an example to fetch data from Firebase storage. The first two solutions would not even compile. It is very polite in apologizing, but it seems weird, because one would assume that a Google AI bot would at least know how Firebase works? Sorry if I mistake the concept.
Firebase changes very often. I'm not surprised that the AI doesn't know how to implement that. You're expecting too much
I like how it completely invents APIs that don't exist but would solve the problem perfectly lol. The chat feature from Gemini hallucinates a lot while ChatGpt usually does it better.
The auto complete is nice, I just wish they would take the real classes into account instead of trying to predict the class and giving fake field names as suggestions.
I'm kind of surprise, I would not say each time i'm using gemini, but it happens quite often that when I type my code and ask some questions about what I wrote or to find a mistake or to change some part of the code, it can answer correctly. However, i agree on the fact that chatgpt is globally beter but indeed in the future i'm expected gemini to be much more better on kotlin topics.
According to gemini, you are BATMAN
HE IS
Gemini give me right answer (Jul, 2024 )
Philipp Lackner is a well-known Android developer and content creator. He's recognized for his TH-cam channel, where he shares tutorials and insights on Android development, primarily using Kotlin. His content covers a wide range of topics, from beginner-friendly introductions to more advanced concepts and libraries. Many developers, especially those who are starting their Android journey, find his tutorials helpful and easy to follow.
Gemini gives good answers sometimes and with well explanations but
I am keep on getting bad answers, sometimes totally wrong answers by it.
Also after some questions it fails to remember things. I need to remind it again..
Better use it to improvise existing code, add comments, writing tests and to apply kotlin language style guides for now..:)
I have not yet tried the integrated Gemini in Android studio, but I use the Gemini webpage a lot to find answers to my questions about code, and it is usually pretty good with getting me started. If a Gemini answer does not work, I just say, "That is not correct" and then it often comes up with an alternative I can try. The good thing about Gemini is that there is, at least for me, often not an alternative to find a better answer.
I use gemini as you, but i use stackoverflow as support if i need more options for questions.
Omg, its so true, gemini always give me the wrong answer, and its very confident. And when i tell it what i want and to just help me improve on my thinking, it keeps trying to sabotage me by telling me about other solutions taht are mostly wrong. I feel like gemini was modelled after a gaslighter...
We need comparison Github Copilot with Gemini in Android Studio
Copilot is better but let's see overtime.
I'm a beginner. Text based Gemini caused my M3 MacBook Air (16GB) to get warm, as if I were gaming or something, after only 3 or 4 queries (as well as not giving me any useful answers).
Gemini suggestions kept giving me unhelpful suggestions. It got annoying over time as it would create text that isn't really there and it was just a bit confusing.
So I disabled Gemini entirely.
I'm using ChatGPT instead which isn't great for Android Development, but is many times more reliable than this Gemini integration.
Thanks for saving us the time!
what theme do you use?
Gemini needs to get way better. I use ChatGPT, and it explains so well a lot of things about Jetpack compose.
For example, today I was trying to do apply modifications conditionally on a Modifier, so I tried to use "apply" function, and it didn't worked. So I asked chatGPT how could I apply a modification conditionally on a Modifier in the modification chain, and it told me to use '"let" function. I got confused on why let worked and apply didn't. So I asked GPT why, and it explained well that when we call a modification function on a Modifier, it creates another instance of Modifier with the modification and returns it, instead of returning the original with the modification, and that's why "apply" function didn't worked. "apply" takes the original, executes the lambda and returns the original in the final, whereas "let" returns the modified version.
I think Gemini would never be able to do some good concrete explanations like this, or maybe it could, but I would not trust it because of so much AI hallucinations it is hard to believe in Gemini. Hope they fix it.
Gemini is good, but at some point it feels like talking to someone with Alzheimer's
This is so apt 😂😂
Definitely. If you stick with it trying to point out why its answer are wrong, it eventually loops back to the 1st wrong answer. I've compare to paired programming where other programmer has a head injury.
Gemini is not good compared to chart GPT 4
please make a video about the most common design patterns in android development and how to incorporate them into clean architecture.
Good stuff :)
Which theme do you use in your Android Studio?
They have clearly written preview mode even its released in stable version of AS
What am happy about is the device streaming.... I can now debug on different devices without using a more powerful machine
Can it generate white themes or only dark and brown themes?
I've been using it since it was released on AS Jellyfish.
I think Gemini is REALLY powerful when it comes to provide documentation for composable functions.
For the rest I think it's still buggy and unreliable....
in my fedora android studio cannt open browser for enable ai .when i close android studio then its triger and open browser.how can i solve this error?
My first impressions are not that good since i use it but it's too easy to use because it's built inside the IDE. Code completion was better than i expected though
"Deadline for feature is a deadline, we must realese it thats all"
Phillip found Gemini from the future.
Is it possible to use code completions without sending data to google? When developing for a company, this is often not allowed.
He explained this on a video from 6 months ago
🤔They probably made it to focus on the conventional android development as opposed to the newer languages and concepts
isn't this the same AI that's supposed to replace programmers. 😄
Given how bad Google are at API design, the auto-complete AI might actually turn out to be useful at some point.
Hey, he said 1.5 mil subscribers because that's the minimum number of subscribers you should have, he was just reffering to the content quality 😁and you're almost there anyway, so no harm there
This version is randomly crashing in my PC
Can you please review codeium?
Its infant today but will mature in future
I tried the Gemini 1.5 pro and it actually works good, I even gave it an HTML CSS code to translate it to compose code, and it worked from the first time.
Anyone thoughts on gpt4 vs claude3 opus?
Mr Philip i have a serious question to you
Iam working as a android java developer
Since from 1 year
So my concern is currently iam seeing that there are no jobs android java
So should i continue on it or i shift to some another tech
What do you recommend
Iam waiting for your reply
Bro casually called a google product "it's trash".💀😭
i once asked gemini to identify all my bugs and it told me to fix classes that didn;t exist...
Because it foresees the future code you will create! 😊
This is so cool
What's cool about it?
It's rather bad
It gives you the illusion that it has found The solution, its like approximating a solution 😂
Gemini is not available in my country Myanmar. 😂 oh my goodness
Also available in Koala
:)) so I use copilot
😂😂for some reason i am the CEO of the largest tech school in Kenya 😂😂...it also crushes when you give it a statement that starts with a comment : //something...
Nothing compared to co-pilot
Code is complex for human developers, how could it be easy for AI? If the humans who wrote the AI code made mistakes and there's bug in the code. I don't see AI building an huge app anytime soon, when we talk about apps that need performance optimizations with thousands of users a long arrayList of infinite scrolling pagination, release memory and resource it's just too much for an AI, make dicision is complicated for humans, we struggle to make decision, how can we teach machine to make decision in unpredictable and new context? AI sucks.
Easy there boy
yey first viewer!
Modern AI is the most dumbest bot ever for coding, just a waste of time.
it's really sucks