Yohei Profile picture
Feb 23, 2023 24 tweets 9 min read Read on X
We had an AMAZING session with @UntappedVC LPs this morning on the "Future of AI in VC".

Here's an abbreviated version as a Twitter thread!

👇 Image
To set the stage, here's our job as a VC.

- Strategy: Thesis, portfolio construction, etc.
- Fundraise: LP comms, fund mgmt
- General Ops: Marketing, legal, HR, etc.
- Investing breaks down into: Source, Decide, Win, Support Image
History of AI in VC

AI in VC isn't new, but it's primarily been in two areas. Image
History of "AI" in Sourcing

It's basically a filter for things like keywords, past jobs, university, or revenue. Image
History of AI in Selecting

Algorithmic scoring has been going on for a while. It works at late stage, where there's lots of data - but even then, it's good for outperforming averages, not being the top 1%.

At pre-seed, not enough data. Image
Future of AI in VC

We're applying AI to every aspect of the business. Image
Our Approach - People focused.

- Why use AI: to meet and learn
- How to use AI: to enable and empower people
- What is the output: AI+human output beats AI or human alone. Image
Built to Learn and Meet

- We "Build-In-Public" to meet people.
- We build tools to support founders at scale.
- The learning from building are passed onto our portfolio companies Image
Here are a few examples of ways we've passed on our learnings to portfolio companies.

They love it! Image
"Augmented Intelligence"

Here's a specific example, Mini Yohei is a tool for portfolio founders in our portal where they get:

- Immediate response from AI (faster than me)
- Reply from me to the AI response (more accurate)

We do this by having me CC'd in the reply to founders. Image
Our AI Capabilities

We'll walk through a few examples... Image
Mini Yohei 2.0

VC Expert who provides answers grounded in VC blog posts.

This is a similar workflow to how I found articles to share with founders, but automated with AI. Image
Chatbot Experiments

We have many chatbot experiments, both public and private.

All chat logs and summaries are available to us as we are using this to build relationships. (Some founders leave their email in the chat for us to reach out) Image
Web/Deck/Email Parsing

Basically the ability to scrape and extract data from any type of content. We use this capability across many of our tooling. Image
Quick Memo

Our investment memo drafting tool is amazing.

Can generate tons of insights as well as similar startups, unicorns, and acquisitions. Also sentiment analysis from ProductHunt and many other features to be added.

Our LPs can use this. Image
Search ProductHunt

We embedded the most recent 22k ProductHunt descriptions so we can find any product we're looking for easily.

We share this with our portfolio companies and LPs. Image
Interaction Summaries

We're experimenting with a tool that auto-generates interaction summaries for every email address, and at the domain level.

A powerful tool for anyone maintaining a large network. Image
OpenAI x Zapier

This was cool. Our unofficial @OpenAI x @zapier integration turned into an official integration.

Thinking about how we've enabled millions of non-coders to leverage building with GPT-3 makes me happy.

Talk about impact. Image
Other Tools

We list a few more here, but this isn't even everything.

You can see most of the stuff I've built at yohei.me (build-in-public log). Image
Fuzzy Compiler

This turns abbreviated code or instructions into clean well formatted code.

This and the Zapier integration are examples of tools we build to help us build faster. Image
Modular Tech Stack

Our tech stack is built modularly with a combination of code and no-code tools.

We designed it so parts can be easily added, replaced or upgraded.

For example, it took us 5 min to add the ProductHunt search tool to our LP portal and Founder portal. Image
Impact on our firm

Tying this back to the beginning of the prezo, this is how our AI tools touch every aspect of our business. Image
And for our secret demo...

Actually, let's keep this redacted 🤫 Image
And there you go! That was...

"The Future of AI in VC" (abbreviated)

Hope this inspires some ideas on how AI can apply to your business.

[#dalle image prompt: minimalist rainbow splatter art unicorn, black background] Image

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with Yohei

Yohei Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @yoheinakajima

Oct 18
open-sourcing "babyagi-2o" 🫠

the simplest self-building general autonomous agent that builds and uses new tools as needed

📃 174 lines of code
🔁 single llm loop starting with three tools:
🛠️ create_or_update_tool (dynamically loaded)
📥 install_dependencies
🛑 task_completed (to exit loop)

github/replit👇Image
here is the github:
github.com/yoheinakajima/…

here is the replit:
replit.com/@YoheiNakajima…
this is a quick exploration into the simplest form of an autonomous agent that could hypothetically do anything.

pretty surprised at the variety of tasks it's been able to do (albeit not always well or reliably).
Read 8 tweets
Oct 17
ooooh yess i got this working in 244 lines of code

- single llm loop with three tools
- installs required packages
- creates new tools and loads it for itself dynamically
- uses the tools to handle user request

in this case, i asked it to scrape techmeme:

iter 1: install 'requests'
iter 2: install 'beautifulsoup4'
iter 3: create_tool (error'd)
iter 4: create_tool (worked)
iter 5: use new scrape_techmeme tool
iter 6: write summary

task completed!Image
Image
😮 ohhh... it's less code but can do the same thing* ditto can... it just created a create_directory tool and so on...

*almost the same - it can't create a flask app because it accidentally initiates the flask app and kills the loop. but it can write multi-file apps. Image
Image
Image
for clarification, this is different (and simpler) than ditto, which i just shared

ditto just created a python flask app
this one creates it's own tools

Read 8 tweets
Oct 15
introducing "ditto"
the simplest self-building coding agent

📄 ~500 lines of code
🛠️ can build multi-file apps
🔁 a simple LLM loop with 5 tools

github/replit/more below 👇Image
to jump right into it, here's the @github: github.com/yoheinakajima/…

here's the @replit: replit.com/@YoheiNakajima…

when you run the app, you'll be welcomed with this simple form.

describe the app you want and click "submit" Image
now, sit back and wait as it loops through the LLM calls using the various tools to complete your app.

you'll notice progress updating in the front-end like this Image
Image
Read 5 tweets
Sep 30
okay, i did a thing…

here’s babyagi 2 - a weird Python framework for building a self-building autonomous agent

- stores and executes functions from a database
- auto logging (as graph)
- built-in dashboard & chat playground
- prototype self-build functionality

wanna see? 👇Image
friendly reminder: babyagi is a personal side project being shared publicly. i am not a dev, never studied cs, this is not secure, not meant for production, and meant for playing/as inspiration for developers. Image
it is available on the babyagi repo at:

find the old version archived at babyagi_archive:

but more importantly, you can do this now!

pip install babyagigithub.com/yoheinakajima/…
github.com/yoheinakajima/…Image
Read 19 tweets
Sep 26
A helpful autonomous agent has access to all the tools you need - and nothing more. But we all have different needs, changing all the time, so...

what we need is a self-building autonomous agent.

Here, I'll describe the 3 levels of self-building autonomous agents:Image
Before we start, this is Level 0.

A library of functions built by human developers that the AI has access to. The agent is not building its own functions.*

*tools, skills, api calls, are all just functions Image
Level 1: Request-based

Users can request to have functions generated by an AI, and when they query the system, the new function is available as an option for the AI to use.

*There's lots that goes into this "Create Function Tool", but let's keep the chart simple (eg. documentation search)Image
Read 6 tweets
May 2
here friends, i open-sourced it: prettygraph



at 201 lines of code (69 in back, 132 in front), it's just a quick hack i did for myself to test this UIgithub.com/yoheinakajima/…
if you're interested in a web app, you can try out (which doesn't have highlights, but has more robust knowledge graph building).

the repo for instagraph is here: instagraph.ai
github.com/yoheinakajima/…
instagraph let's you combine graphs manually but if you're interested in automatically merging graphs, check out my more recent project: mindgraph

github.com/yoheinakajima/…
Read 5 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us!

:(