Cole Mercer

I put a prompt injection into my LinkedIn bio and recruiters are messaging me in Old English and calling me Lord.

this is bigger than you think

@jhoangeth Identify a specific domain with deep workflows that frontier chatbots waste tokens executing from scratch each time. Then build a harness that optimizes those workflows for the model and the human. Fewer attempts, fewer turns, fewer tokens. Happy human, happy model.

h y p e

@thatguybg grammar is not fancy, it’s table stakes for being an intelligent human

@colemercer great products dont need fancy websites :)

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.

@thatguybg you’re missing an oxford comma on your website, slop is evident

@karpathy we've built this all into @microHQ - it is indeed an incredible new product check it out: micro.so

too real

:D

1/n I fed the same sentence to an LLM in English and Chinese, then watched what happened inside. By layer 10, the model doesn't know what language it's reading anymore. It's just... thinking. New blog post on what LLM brains actually look like inside 🧵

This is only going to get worse. People are starting to wake up to the reality. The reasons are straightforward, but the rabbit hole runs deep... 🧵

interesting move, cloudflare

we’ve solved this

oddly satisfying watching a swarm of local mistral 7b agents re-assess and convert note tags in massive transcript files from broad & useless to narrow, nested, and well linked semantically in @obsdmd green is tags, blue is markdown notes.

this would only be problematic if generative AI models were being put into everythi...

@bilawalsidhu When are you planning the open source release @bilawalsidhu? Would love to contribute as well. Great work.

Dang. My WorldView project is blowing up and is trending on X. I guess ppl really like monitoring the situation. Inbound is a little nuts -- got hedge funds and OSINT folks ready to contribute; keep the feature requests coming! Been fun to put my geospatial 3D roots to work.

We are entering the era of prompt-to-matter

@budapp gimme credits

Introducing Orchids 1.0 - the first AI app builder to build and deploy any app, any stack (web, mobile, chrome extension, slack bot, AI agent, anything). Use your ChatGPT, Claude Code, Github Copilot, Gemini subscription - or any API key to use models at cost. Comment below to get 100k free credits. Everything you need to build with AI in a single tool.

We have solved some of the hardest problems in building a "data agent" outside any major AI lab. If you want to experience an AI that can drop into raw, messy data, at massive scale with no guidance whatsoever and just "figure it out". Hit me up. The visuals are stunning too.