Ribhav

member of technical staff @ 8090
@ribhav99 on all socials

[Ribhav @ribhav99]

@nikitabier I don’t want to see videos on x unless it’s something like a space x launch or something like it…
Instagram is the brain rot reels platform. I guess TikTok and YouTube shorts too. But please don’t ruin x

[Ribhav @ribhav99]

I’ve botted pre llm times on alt accounts and you can get away with it. I probably still have the scripts on github somewhere. Move the mouse “naturally” with some randomness and it doesn’t seem to get flagged. Very easy for grindy parts of the game but I’m not sure how it would hold up in combat. In theory, should be good. But never tried.
Now just give your agent a tool to call the mouse move script and you’re good.
Combat will need to be done without LLMs and I think that’ll be really fun

[Ribhav @ribhav99]

Idk about this… looking at the demo it just seems like you can achieve this with really good engineering instead of any novel research of method of training. Your tools are async and you have hooks for when they return and hooks for the vision model.
It’s really cool, but for whatever reason this doesn’t excite me

[Ribhav @ribhav99]

@khushkhushkhush This reads like a LI post…
Get this outta here

[Ribhav @ribhav99]

@chamath Karpathy’s knowledge base idea file is actually a great place to start. I began my journey yesterday and I’ve spent a decent amount of time building personal context into my knowledge base. It’s actually been quite fun. Maybe because I’m excited to make it actually useful

[Andrej Karpathy @karpathy]

LLM Knowledge Bases

Something I'm finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. In this way, a large fraction of my recent token throughput is going less into manipulating code, and more into manipulating knowledge (stored as markdown and images). The latest LLMs are quite good at it. So:

Data ingest:
I index source documents (articles, papers, repos, datasets, images, etc.) into a raw/ directory, then I use an LLM to incrementally "compile" a wiki, which is just a collection of .md files in a directory structure. The wiki includes summaries of all the data in raw/, backlinks, and then it categorizes data into concepts, writes articles for them, and links them all. To convert web articles into .md files I like to use the Obsidian Web Clipper extension, and then I also use a hotkey to download all the related images to local so that my LLM can easily reference them.

IDE:
I use Obsidian as the IDE "frontend" where I can view the raw data, the the compiled wiki, and the derived visualizations. Important to note that the LLM writes and maintains all of the data of the wiki, I rarely touch it directly. I've played with a few Obsidian plugins to render and view data in other ways (e.g. Marp for slides).

Q&A:
Where things get interesting is that once your wiki is big enough (e.g. mine on some recent research is ~100 articles and ~400K words), you can ask your LLM agent all kinds of complex questions against the wiki, and it will go off, research the answers, etc. I thought I had to reach for fancy RAG, but the LLM has been pretty good about auto-maintaining index files and brief summaries of all the documents and it reads all the important related data fairly easily at this ~small scale.

Output:
Instead of getting answers in text/terminal, I like to have it render markdown files for me, or slide shows (Marp format), or matplotlib images, all of which I then view again in Obsidian. You can imagine many other visual output formats depending on the query. Often, I end up "filing" the outputs back into the wiki to enhance it for further queries. So my own explorations and queries always "add up" in the knowledge base.

Linting:
I've run some LLM "health checks" over the wiki to e.g. find inconsistent data, impute missing data (with web searchers), find interesting connections for new article candidates, etc., to incrementally clean up the wiki and enhance its overall data integrity. The LLMs are quite good at suggesting further questions to ask and look into.

Extra tools:
I find myself developing additional tools to process the data, e.g. I vibe coded a small and naive search engine over the wiki, which I both use directly (in a web ui), but more often I want to hand it off to an LLM via CLI as a tool for larger queries.

Further explorations:
As the repo grows, the natural desire is to also think about synthetic data generation + finetuning to have your LLM "know" the data in its weights instead of just context windows.

TLDR: raw data from a given number of sources is collected, then compiled by an LLM into a .md wiki, then operated on by various CLIs by the LLM to do Q&A and to incrementally enhance the wiki, and all of it viewable in Obsidian. You rarely ever write or edit the wiki manually, it's the domain of the LLM. I think there is room here for an incredible new product instead of a hacky collection of scripts.

[Ribhav @ribhav99]

@FFmpeg Rewrite it in python instead
Better yet, replace all of ffmpeg with a skill that agents can reference to write python code at runtime to complete the user request

[Ribhav @ribhav99]

@sergiu_bodiu @chamath @grok My guy got roasted by grok instead 😂

[Ribhav @ribhav99]

@signulll @Jessicalessin @sama Maybe you have different realities

[Ribhav @ribhav99]

@zeroxjackson Hit the gym in the afternoon when you’re losing focus and then work 12-14 hours instead

[Ribhav @ribhav99]

The best music coming out of India is from Ed Sheeran 😂
Who would’ve guessed

[Ribhav @ribhav99]

@ChrisLaubAI I don’t understand, intuitively, why this is better. Your voice is still being encoded into some n dimensional space. If it mis-hears you (for whatever reason, accent, noise etc), that encoding could be very different just like a wrong transcription.
What’s the breakthrough?

[Jay Alto @theJayAlto]

there are essays, and then there are "violently shift your entire perception of reality" essays

[Ribhav @ribhav99]

@janaksunil Show him this tweet

[Ribhav @ribhav99]

@asharoraa @nikunj What…
Even I know of the jetsons

[Ribhav @ribhav99]

@asharoraa Yay I have hope 😂

[Ribhav @ribhav99]

Laziness, to me, seems to come from focusing on the effort rather than the reward.

The interesting thing is that when you look back and “reminisce”, you think of the past, generally, in a positive way. “University was the best time ever”. But when you were in university, with all the different stresses and deadlines and pressure, did you feel the same way?

In the long run all you’re left is with the reward. The pain and sacrifice of the hard work seems to be forgotten. Or atleast it seems to matter less.
That’s interesting

[Ribhav @ribhav99]

@rohanpaul_ai @grok Lmaooo that checks out
I was so confused when I read that 😂