@nilayparikh@simplifyinAI Flat index? I need to get educated on that. I'm just using pgvector, where as you said, embeddings are stored in a single column.
I find that the trick to getting good results is the right chunking approach, which can be tricky for different data types.
One of the newest skills from Anthropic - Playground! I see so many possibilities using this skill, here are a few examples.
You can easily install this skill from the plugin marketplace, and see more examples from the official article here:
https://t.co/ERCiSK9sof
@ShaneSmitas I was hesitant of using Cursor as it added more friction, but I can see how a few extra minutes spent up front can save hours of debugging in the future!
I love 'Plan' mode from Claude. But, I have to admit it's not as impressive nowadays.
Running a separaet Codex CLI and making it review the plan from Claude, against the actual code base is such a time saver.
Tried using codex CLI via Claude agent, but the results not impressive
@_alphashark_@grok Yes, it's also "you don't know, what you don't know", until something clicks. Just need to keep trying until you find something that works best
Dumb move of the day: Spent half a day wrestling a Claude Chrome extension to analyze my X posts... when I could've just asked @grok.
Claude extension felt clunky/slow/hallucinated half the replies or missed a bunch.
Grok gave a clean, spot-on breakdown in minutes.
Thread: Why Grok wins for quick X analysis + my real workflow tweaks. 1/
@housecor It's all fun and games until it isn't anymore.
My hunch, those who push this technology "forward" will be the first to fall by the technology they try to advance.
Or maybe I watched too much Terminator and Matrix
Personally, I have a hard time with architecture visualization and marketing demos.
This new skills make it easier to do both, better!
How are you planning on using this new skill?
What if your bot could help you stay on track of your Tasks?
I'm testing an idea where an AI is able to not just organize or prioritize tasks, but be an active support partner, to remind and cheer me on to complete my curren Tasks.
@nilayparikh Yes. For some reason people feel obligated to use products as they see fit without paying or respecting the rules.
Yes, there are open source models that people can use all they want and not complain.
@dani_avila7 I've had several https://t.co/3XYS8yGWeA files for many months now. I just keep a main https://t.co/3XYS8yGWeA with references and some top level instructions, while I create detailed https://t.co/3XYS8yGWeA for each app and testing
When choosing a vector DB, don't reinvent the wheel and burn your money.
Go with simple solution unless you really have the scale or very complicated requirements.
This article shows a great example with pros and cons
https://t.co/5jHftPpGfC
@fortelabs Tbh, with the right setup and workflow maintaining is not that hard either.
But, if you don't understand the architecture from the beginning and build a house of cards, then you may as well break it and start anew.