James Lee

Building something new with AI

[Yohei @yoheinakajima]

you need to run agents outside of your core work where it's low risk so you can push them further than you would otherwise

and then you'll have a better sense of where you can trust them today, and can keep bringing that back into your work agents

[jack @jack]

https://t.co/jgZkBvYOPt

[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.

[John Zeratsky @jazer]

🚨 Applications for Character Labs G6 are due in 3 weeks!! The program for this group will be radically different from before, and we think it's going to be much more effective. Can't wait to run it down live with the next group of founders!

👉 If you're at inception/idea stage and interested in working with us, please take a look at https://t.co/vvU4hN0s1N and let me know if you have any questions!

Here's what's changing for G6, and why:

1. All-new Sales Sprint and Marketing Sprints.

Many founders in our portfolio are now finding and signing customers in weeks, not months — often well before they have a "finished" product ready to sell.

We reverse-engineered their techniques and designed new sprints to help every founder start selling uncomfortably fast.

2. All-new Pitch Sprint.

Founders in our portfolio have raised from a16z, First Round, Index, Obvious, Bessemer, General Catalyst, and other elite investors.

We completely revamped our Pitch Sprint to capture the most important lessons from the founders who've had the most success in fundraising.

3. Demo Day!

The biggest change EVER to Character Labs.

For the first six groups, we provided bespoke support to each team as they were ready to raise, thinking it was a better experience.

I think we were wrong on this one — by encouraging founders to raise at the same time, we can help them more AND provide a better experience to investors (who can meet 15 great founders in one place instead of being inundated by emails from us).

Caveat: Our first-ever demo day (for G5) is in a couple of weeks, so we reserve the right to be wrong again and change course :)

👇 Bottom line, we are incredibly excited about these changes and think G6 will be the best ever.

But more than that — we are just so eager to meet and work with this next group of founders.

Because Character Labs is a special program. Biased of course, but I really think that's true. We believe it's the best way to get started and build momentum in the early days of founding of a startup.

If this resonates, we'd love to meet you. Let's talk. If not for Labs, maybe we find another way to work together. If not for G6, maybe another time.

But please, check out what's new in the comments, head over to https://t.co/vvU4hN0s1N, and get in touch if you are interested in working together!

[Yohei @yoheinakajima]

hahahahah i made THE sea shanty for Claude Code lovers. play it at hackathons, blast it out your window, and sing it with me:

"Ode to Claude Code"

🏴‍☠️🎶🏴‍☠️🎵🏴‍☠️🎶🏴‍☠️🎵🏴‍☠️🎶🏴‍☠️🎵

[Yohei @yoheinakajima]

[tiny project] ai-share-widget

the easiest way to add* ai capability to a site with a front-end widget where you can customize the prompt and allow the user to pass that into their AI tool of choice

github (102 lines of code) in thread 👇

[Morten Just @mortenjust]

This is fast. Chrome running Gemini locally on my laptop. 2 lines of code.

[Julie Zhuo @joulee]

One of the stories we used to tell in the early days of Facebook was how a small, two-engineer project came to dominate the entire photo sharing landscape in the late 2000s.

Thread 👇 1/10

[Paul Graham @paulg]

Of the 20 richest self-made Americans under 60, 17 have technical backgrounds.

[Sam Altman @sama]

Move faster. Slowness anywhere justifies slowness everywhere.

2021 instead of 2022. This week instead of next week. Today instead of tomorrow.

Moving fast compounds so much more than people realize.

[Sam Altman @sama]

Setting incentives is a superpower.

5 words that changed how I think about almost everything.

[Justin Kan @justinkan]

Any other brave Robinhood employees who are willing to whistleblow should come forward anonymously to reporters. I can put you in touch.

[Steve Schlafman @schlaf]

I want to see a school that:

•Has no majors
•Focuses on independent study
•Preaches curiosity and exploration
•Awards no grades
•Optimizes for collaboration
•Provides advisors / coaches
•Encourages public sharing
•Teaches mastery
•Sells personal transformation

[Minda Brusse | First Row Partners @MindaBrusse]

So proud of what this group has accomplished together.

[Shreyas Doshi @shreyas]

How to develop PM Career Skills?

First understand the components of each PM skill area.

And consciously choose to grow in each of the components:

[Shreyas Doshi @shreyas]

PM Career Skills Map:
Analytical, Execution, Product Sense
Commercial Sense
PM Mentorship & Management
Product Vision & Strategy
Organizational Design & Development
Company Strategy & Execution

Tech, Design Sense is nice
Influential Communication & Critical Thinking is vital
⬇️

[Palak Zatakia @zatakia]

Elon Musk on how he learns so much so fast.

[Venture Out @GoBigtoSmall]

First and foremost-Black lives matter.

We firmly believe that entrepreneurship is one of the best ways to bridge the economic gap that is inherent in systemic racism.

As we continue to learn, advocate and support, here is what VO is committed to doing: https://t.co/HzIRtuORGG

[Barack Obama @BarackObama]

My statement on the death of George Floyd: