Co-founder @IndieHackers, the world's largest online startup community. Building @UseHatchAI, the world's most productive AI assistant.
NYC
Joined December 2009
- Tweets3.3K
- Following104
- Followers21.9K
- Likes5.3K
Pinned Tweet
We built an app called Hatch
It enables Karpathy's entire LLM Knowledge Base workflow out of the box, in 2 or 3 clicks, in a single interface.
No need to stitch together Obsidian + plugins + markdown files + custom tools + etc.
Hatch is an AI workspace where files, docs, and chats all live together. Not only can the chats READ files and docs, but they can WRITE and ORGANIZE them as well.
This means you can set up an LLM Wiki in literally 3 steps, as I show in this video.
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.
A funny paradox about work I dislike is the best way to enjoy it is to go all in on it.
My instinct is the opposite. To hold it at arm's length. Do it with half-attention, procrastinate, skip steps that annoy me, etc.
But the trick is to give it my full attention, try doing it better or faster than ever before, go on little side quests to learn and master the parts I don't understand or feel familiar with. Basically, to "know" the task in the biblical sense.
Man, AI is so underutilized! Even by tech people. Problem is you have to remember to prompt it.
Instead we need the AIs prompting us. With ideas. Instructions. Warnings. Reminders.
Of course we'd have to let it monitor our lives passively and few people would go for that.
Channing Allen retweeted
hitting $35K MRR took 4+ years of struggling
AND I ALMOST GAVE UP 2 SUMMERS AGO...
but i didn't, and things worked out so far, so i did an interview about it ↓
AI tools are unique in the history of software in that if a user is confused about how to use one, the best solution is usually to ask the tool itself.
In some seasons of my life I go full try-hard and in other seasons I kick my feet back a bit but I feel this is a moment not just in my own life but in history where I need to go all out and create/capture as much value as humanly possible
My executive assistant just inked us a $7.5k ad deal without any guidance, supervision, or even awareness from me. Just needed my final sign off.
The sales cycle included 10+ email exchanges and 2 Zoom calls. She upsold the client!
AI won't replace her for 18+ months
Closing in on a quarter million email subs!
Good place to be now that distribution is the biggest moat
Software is still a moat and people should still learn to code.
If you disagree with this, at least one of the following is true:
1. Your product is very basic
2. You haven't watched non-technical people try to vibe code non-basic products
Planning a Eurotrip to Portugal and Paris for the second half of July (between 7/15–7/29).
Wanna throw an indie hackers meetup in each city!
Anyone down? Reply/DM me!
cc @levelsio @Pauline_Cx
What's wild is:
Here's the kicker:
The contrarian bit:
The interesting part:
The thing people miss:
The insane part is this:
The part everyone resists:
The thing most people skip:
Stop it. Please. I'm begging you. You can write your own thoughts. I believe in you.
Forgot how fun founder events are. Going to start doing Indie Hackers meetups in NYC again soon!
If you're in NYC today,
I'm doing a panel discussion about @IndieHackers and @UseHatchAI at Port Sa'id at 10:30am for NY TechWeek
Come say hi!
If you're in NYC today,
I'm doing a panel discussion about @IndieHackers and @UseHatchAI at Port Sa'id at 10:30am for NY TechWeek
Come say hi!
Speed used to be an advantage when nobody could build apps 5–10 years ago.
Now that everybody can build, it all comes down to quality.
Today's Luddites are in the universities, not the factories.
Booing Eric Schmidt is like booing gravity. The forces behind AI are now bigger than Schmidt and bigger than Google/OpenAI/Anthropic.
Maybe 1/10 of these students will adapt & land on their feet. As for the rest?
