Aaron Bumgarner

@signulll There are a lot of people that say they are using CC but are really just using Claude chat on the browser, when I would do demos they thought I was doing magic by using env variables/api keys/skill files to connect

@signulll It seems like something that might have been necessary 6 months ago and got some traction and is now a complete waste of time. The one nice thing is for non tech users on browser that can sso/oauth into enterprise env applications without any friction

mcp was a mistake.

@karpathy I’ve been messing around with this to help map everything in my personal encyclopedia. Curious if the node/edge approach is an efficiency gain. github.com/NoahIrzinger/b…

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.

@fed_speak Memories, new feature as of a couple weeks ago

Sometimes I think Claude is just fucking with me weaving in context that I have never told it directly but it has picked up on over the course of our conversations.

@ettingermentum He has embraced full stack substack ai slop journalism hahah look at the start of this article 🤣🤣🤣

Gage Skidmore does not clock in day in and day out for you to use AI as the cover pic for this

@default_friend When that happens for me in a book it makes me think… if they got this wrong… how many other things did they get wrong that I am just not as familiar with

Random but I’m reading a Christian book and it’s amazing to me the way Wicca is misunderstood … book is otherwise fine but things like that really turn me off

I will never understand the mariners passing on the best vibes player in the league twice

Geno my fcking DOG

@Dustin206 It’s going to be 2 or 3

It’s 5 or we riot

@MikeDTheGod Link?

Me and my niggas share mad music with each other so we decided to create a shared playlist and it’s turning out to be crazy.

Uhhh

@responsiblerob Where do you watch these?

“YOU WANNA SEE THE LINDSAY LOHAN PROGRAM?! I BLADE RUNNER NOW!” he’s simply one of the funniest people of all time. been rewatching all of his stern appearances and every line is funnier than the last.

@signulll Yeah I also haven’t a lot of agents that do it

signs that you’re down really bad: - you start a gratitude journal. - you sign up for a marathon. - you join a singles dating run club. - you decide to get a master’s degree in data science. - you decide to go to law school (really bad). - you sign up for an mba program. - you decide to get bangs. - you decide to get another master’s degree. - you decide to do an executive mba on nights & weekends. - you randomly decide to go to medical school. - you decide to build a custom gaming pc. - you decide to get (another) tattoo or piercing. - you spend time browsing endlessly on dating apps. - you fill your loneliness gap by going on endless online dates. - you pick up a hobby you have zero interest in. - you decide to start an anon twitter account.

@TheStalwart Sam Altman basically compared people to livestock in that last speech, I don’t blame them

This is a good take. I haven’t heard anyone in the AI world credibly articulate why the average person should assume it will make their life better. Typically they say the opposite.

Big dick Darnold

JorHOG Polanco

@matthewstoller The type of effective altruism I can get behind

As we begin attacking the bad habits embedded in corporate America/Wall Street, those guys are going to do as much harm to us as possible and claim we're doing it to ourselves. That's just the price we have to pay for taking power back.

2023 was a pretty good year and hopefully 2024 will be even better.

Repost for a chance to win flights and tickets to see @UW_Football in the Pac-12 Championship! Terms & conditions: gohuskies.com/AlaskaFlights

@mucha_carlos @Chris_arnade Not sure state prices would be a 1 to 1 comparison, even if it remained flat for 30 years. Don’t get me wrong, I’m pro Japanese zoning codes at just about every level. Their mixed use zoning is art. But the population curve probably eases pricing pressure

@AaronBumgarner @Chris_arnade I dunno, California’s population decreased this past Census for the first time in over a century (and continues to decline). Do you really think California housing prices have decreased with its declining population? 🤔

A lot of Japan, including the rural parts, is modern concrete bleh, without any signs of the magical "Totoro" type stuff. A mishmash of commercial strips, tiny rice fields, train tracks, industrial complexes, and housing estates. Then you run across these small magical nooks.