Did you already try the Google flash thinking model in cursor? Or the deepseek v3 model. Both work by setting custom models and endpoints. Really cool stuff.
This is a timely video, I was just thinking about this - some kind of syntax or schema for AI to better understand webpages. This will surely be a big thing in the future.
That's what he thought, but it makes no sense to me. Unless the model is trained on data in base64. Though most LLMs are trained on human readable data, they are able to more or less decode base64, but they hallucinate a lot when they do because it adds a lot of mental gymnastics
Great content, as always! A bit off-topic, but I wanted to ask: I have a SafePal wallet with USDT, and I have the seed phrase. (alarm fetch churn bridge exercise tape speak race clerk couch crater letter). What's the best way to send them to Binance?
web search tools usually scrape body and paragraph tags ... b64 encoding is from my perspective an unnecessary workaround (until you have images or videos directly encoded). easiest way is to include the necessary text (maybe as yaml with a specific prompt upfront to instruct that the following text is in yaml format) and b64-encoded binary informations (images, text - likewise with specific prompt instructions within the page) only in a simple body part
I wonder if you can add mischievous prompt for bots, especially ones with tools that have access to your email or something worse like your bank account, to perform some bad actions. I can see this as a deterrent for 'ai scraping' web sites or something. this topic is interesting, even when dealing with reading documents on your hard drive.
Indeed encoding to base64 makes no sense for plain text because it takes more space than simple basic text. I believe the json format would be the best option.
I have a question for you pls since your are native Norwegian, what is best and better to write, understand and translate Norwegian language? Is it gpt 4o? O1 ? Or Claude? Or others??
You will upload to a database, not publish on the web. Maybe an aggregator. Hell why not just push content to Google directly or upload to another aggregator with Blockchain id so if your content is surfaced in an llm response you get rewarded? Something is going to change. It will not be this ever again
Класное видео❤ Всех с новым годом друзья пусть он как и для меня так и вас всех будет лучше :-) и я серьезно он для меня может многое значит видь я только только начинаю свой путь на ютуб
i said it before many times: new era of web will truly begin once there are no more webpages - estimating conversion will start 2026 (2025 first experimental ones for research / some private ones)
Did you already try the Google flash thinking model in cursor?
Or the deepseek v3 model.
Both work by setting custom models and endpoints.
Really cool stuff.
I was thinking about it a couple of weeks ago! I think this will be the future of web as well
This is a timely video, I was just thinking about this - some kind of syntax or schema for AI to better understand webpages. This will surely be a big thing in the future.
So is what you are saying is, that you need to convert the web page code/ data to base 64 for AI to read it - otherwise it cannot?
That's what he thought, but it makes no sense to me. Unless the model is trained on data in base64. Though most LLMs are trained on human readable data, they are able to more or less decode base64, but they hallucinate a lot when they do because it adds a lot of mental gymnastics
Great content, as always! A bit off-topic, but I wanted to ask: I have a SafePal wallet with USDT, and I have the seed phrase. (alarm fetch churn bridge exercise tape speak race clerk couch crater letter). What's the best way to send them to Binance?
web search tools usually scrape body and paragraph tags ... b64 encoding is from my perspective an unnecessary workaround (until you have images or videos directly encoded). easiest way is to include the necessary text (maybe as yaml with a specific prompt upfront to instruct that the following text is in yaml format) and b64-encoded binary informations (images, text - likewise with specific prompt instructions within the page) only in a simple body part
Maby u need an AI agent team to mine the information, process the information into a format that's optimized for an AI to read?
I wonder if you can add mischievous prompt for bots, especially ones with tools that have access to your email or something worse like your bank account, to perform some bad actions. I can see this as a deterrent for 'ai scraping' web sites or something. this topic is interesting, even when dealing with reading documents on your hard drive.
interesting idea. there are some websites that have a lot of images or are a bit messy that could use something like this
Indeed encoding to base64 makes no sense for plain text because it takes more space than simple basic text. I believe the json format would be the best option.
Do you think ai will replace programmers?if do you recommend study computer science?
Should we, yes. An interface specifically for AI will emerge
I have a question for you pls since your are native Norwegian, what is best and better to write, understand and translate Norwegian language? Is it gpt 4o? O1 ? Or Claude? Or others??
You will upload to a database, not publish on the web. Maybe an aggregator. Hell why not just push content to Google directly or upload to another aggregator with Blockchain id so if your content is surfaced in an llm response you get rewarded? Something is going to change. It will not be this ever again
Класное видео❤
Всех с новым годом друзья пусть он как и для меня так и вас всех будет лучше :-) и я серьезно он для меня может многое значит видь я только только начинаю свой путь на ютуб
it's really crazy how nobody is talking about the book the hidden path to manifesting financial power
No way everyone is talking about it, sometimes just randomly bringing it up for no reason
Crazy that you think it's crazy
Ask Gemini it will tell you NO!
i said it before many times: new era of web will truly begin once there are no more webpages - estimating conversion will start 2026 (2025 first experimental ones for research / some private ones)
also OSs and software (in the current sense)
Web is dead. GUI is dead. Goodbye SEO and web dev
Lol
nice overengineering
lol this is hilariously useless