It seems like you might be trying to use it like a search engine, which is a common mistake people make when first trying LLMs. LLMs are not like Google.
The key is to give it context so it can help you. For example, if you want it to help you with Spark configuration, give it the Spark docs. If you want it to help you write code, give it your codebase.
Tools like cursor and the like make this process very easy. You can also set up a local MCP server so the LLM can get the context and tools it needs on its own.
Thank you very much for the ideas here, i will try the approach of giving it context. I havent got into cursor, since i use helix and intellij… i need to look into the MCP server thing
thank you, I will try this too. I feel like I didn't have to do this much work with other models like o1/o3/4o... but if it provide the return value I'm hearing from the hype around Claude I am willing to try.
This makes me think of the early/mid-2000s & https://blosxom.sourceforge.net. Blosxom had this delightful concept of file extensions as "flavours." For example, you could have a ".rss" flavour that would present that hierarchy of your site as an RSS feed if you added ".rss" to the URL. Brilliant!
This used to be fairly common. Reddit is another site. A company I worked at.aroind the same time also had .xml, .rss, .atom. .xml would serve up the raw xml our middleware generated, which was normally "rendered" via xsl (what can I say to redeem myself for that?) server side. It was great for both debugging (you could browse the site in "xml mode") and to provide an API.
I still like the url approach - being able to browse until you have the view you need, and then just copy the URL and change format in order to find the right API call can be very nice. The challenge, of course, is that you need to be very cautious about which urls you guarantee will be stable, or you'll be locked into a site structure you might regret.
The big problem was that the easy way out is that your XML ends up being changed to be "XSL-friendly", which means a ton of concessions that effectively encodes knowledge of the expected presentation no matter how much you want to keep it largely semantic.
Small presentation changes far too often result in changes to the XML to accommodate weaknesses in XSL.
I still like the idea. But not the use of XSL to achieve it. Unfortunately, we don't have any great alternatives that aren't horrible in all kinds of different ways.
I remember fondly following the Project Aardvark blog posts through the summer of 2005-- it was such an exciting time! It's still online here: https://www.projectaardvark.com
I appreciate the intent here of making levels more transparent & evenly applied, but in practice this seems to be tying levels/comp to a demographic attribute people have no control over: their age. This makes me feel uncomfortable.
Carnegie Mellon had a rule (mid-2000s) that you could only host .orgs from the static IPs you could register on your dorm Ethernet. I launched openpodcast.org & bibme.org as a result of that. I always thought it was a senseless rule.
