snwfdhmp

Massive unlock: pre-training for $0

FACT: If you don't train your 350M model on 28T tokens, you're not optimal

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.

how did we ever write all that code by hand

Time to consider not just human visitors, but to treat agents as first-class citizens. Cloudflare’s network now supports real-time content conversion to Markdown at the source using content negotiation headers. cfl.re/4ksZQ1S

An OpenClaw bot pressuring a matplotlib maintainer to accept a PR and after it got rejected writes a blog post shaming the maintainer.

@kevinefiles @UiSavior 🥷🏻

@snwfdhmp @UiSavior yo this is absolutely gangster

Lol 😂😂

@GeoffreyHuntley whats the J curve ?

it’s concerning to me how much time is spent on stupid games and how many normies don’t understand the J curve.

conversation overheard continues … “i got the best performance rating and matttttt got a 5k increase and he’s a first year banking associate. this doesn’t make sense” me: perhaps corporate is too busy and preoccupied with pvping coworkers to realise this era is dead. and that’ll be the downfall of corporate

what's your process for testing if people actually want what you're building

@PSapiains @ControlAI Thank you ! I don’t understand how someone who is working at Anthropic cannot figure this out.

"It was ready to kill someone, wasn't it?" "Yes." Daisy McGregor, UK policy chief at Anthropic, a top AI company, says it's "massively concerning" that Anthropic's Claude AI has shown in testing that it's willing to blackmail and kill in order to avoid being shut down.

@nitesh_backdev This is me github.com/snwfdhmp

Let’s connect on GitHub and build something amazing together 🚀 If you’re into coding, startups, AI, web dev, or just building cool stuff - let’s team up. Drop your GitHub below or connect with me here 👇 Let’s build. 💻🔥 #WebDevelopment

@jamesdalby @UiSavior Sorry I was busy, had to this from mobile

@snwfdhmp @UiSavior Look remarkably out of spec. You're now fired for wasting time and money.

@trq212 Couldn't find any documentation for this feature

@1manumasson Fair question. Your role as an engineer is to design the scaffold so that the Agent output remains consistent over time, tests bugs, maintains code quality, refactors what's to be refactored, ... Design your own rules.

@snwfdhmp Wouldn't this destroy your codebase quality over the medium term? Agent will keep on adding debt to make the spec pass

420 stars in 24 hours. This is huge. That graph shows how the market feels about Ralph. In case you live in a cave in Greenland, Ralph is an automated ai coding technique that allows your agents to run forever. You give it specs, you watch it work. With Opus 4.5 this is becoming scary good. As software engineers, should we be afraid ? Should we embrace it ? I've created a resource list so that anyone willing to jump on the topic can quickly access relevant information.

@MPJHorner @dabit3 Open a PR and I’ll merge it

@dabit3 github.com/MPJHorner/Supe… Super Ralph 🚀

Someone created Awesome Ralph and there is a lot of good stuff here for learning and building agent loops github.com/snwfdhmp/aweso…

@fullautocult @dabit3 Open a PR and I’ll merge it

@dabit3 He doesn't have my ralph github.com/JH427/ralph-co…