Share this page!

Yohei Profile picture
Twitter logo
Jan 29 • 7 tweets • 3 min read
Okay, it's getting weird now...

Introducing "fuzzy.js" 🧶

An experimental approach to using #gpt3 as a back-end, communicated via HTML attributes.

Automatically fills in data.
You can get multiple items by repeating the same fuzzy_atr.

Just make sure the number of each match. (ie. if you have 3 "names", make sure have 3 "addresses")
Attributes are used for describing the data field you want.

Describe the overall data you want when you call the fuzzyapi().
Similar to the Dummy API from last night (included here), it's limited by the capabilities of LLM.

- Can hallucinate
- Slow if you ask for lots of data
- Token limits

But if you can imagine a future where LLMs are fast and reliable, it's possible to imagine using it as a database, maybe?

This is purely an experiment at this point.

It's rough, but dropped it on Github (open-source) if you want to play around: github.com/yoheinakajima/…
Here’s a Typescript version:
And a svelte version (GitHub included in thread):

• • •

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

Yohei Profile picture
Jan 28
Inspired by this idea, and a comment I can't find...

Built a "Dummy API" which can provide dummy data for any front-end with a simple API call.

Basically, feed the API:

- description of app
- keys requested
- number of results

This provides JSON array of dummy data.
Here's all 7 lines of code that make up the "API".

It's for front-end devs who don't want to spin up a back-end to test their build.
Woah, you can get real data this way. 🤯

In this example, I'm asking for VC firms on Sandhill Road and got real data.

We're not quite there, but it's possible to imagine a future where you don't need a back-end.

H/t @DYtweetshere for the original idea
Read 7 tweets
Yohei Profile picture
Jan 25
Wha...what's happening...⁉️

Just pushed Mini Yohei 2.0 to our founders portal!

🧠A VC expert who answers questions using knowledge from 100+ VC blogs.

With citation! ✍️
(Not using embeddings)

See thread for example answers + how it was built. 👇
Here's an example answer to the question:

"What should go in a fundraise pitch deck?"

cites @msuster @onstartups and @avc
Quick background, I launched askanything.vc a search engine that only searches VC blogs with @DotDotJames a few years back.

I often use it for research when founders ask me a question.

Basically, I gave Mini Yohei the ability to do this.
Read 9 tweets
Yohei Profile picture
Jan 24
Oh wow!🥳

#gpt3 can build #nocode workflows with #ai.

Fed it two of my workflows as examples, and it successfully generated a third workflow based on query.

Imported to @zapier and it works! ⚡️

Right image: output. Left image: Imported flow. ImageImage
I should throw all my workflows as JSON in an airtable, and embed to find the most similar flows (cc @gpt_index), and use two of those as examples and this should get pretty robust.
Some ppl who I know will like this: @bentossell @mikeknoop @bencmejla
Read 4 tweets
Yohei Profile picture
Jan 18
It's been a blast building on #gpt3 since August. 😆

I've built and shared a few dozen apps and prototypes.

I often get asked this, so...

Here's a list of some favorites! 👇
1) ⚡️Official @OpenAI x @zapier integration

This is the most popular by far.

Originally an unofficial integration I built for myself, it is now powering the next generation of #AI x #NoCode tools.

2) 🔀Fuzzy Compiler

Turns "fuzzy/sloppy" code into clean code.

It speeds up my build a ton. It's mobile optimizes so I can generate code and add to Replit mobile app while I'm bouncing our baby down in the Ergo.

Read 23 tweets
Yohei Profile picture
Dec 17, 2022
📑 URL-to-Report

Combined a web scraper with our diligence capabilities! Mini Yohei can now produce reports from a URL.
So this is really cool! Added Semantic search.

Auto-identifies:
- most similar Unicorns
- most similar startups in our CRM. ImageImageImage
Added a relevant recent acquisitions section Image
Read 4 tweets
Yohei Profile picture
Dec 11, 2022
Don’t write code often?
Been a while?

Worry no more!
Introducing the “Fuzzy Compiler” 🧮

Just make up your own code, no worrying about syntax. Translates to working code w comments.

See example 1:
As you can see, I build a Flask app simply with this:

Flask app

/
index.html

/summarize
template = “summarize the following:”
prompt = template + data
openai completion text-davinci-003
return result

run app
Works with front-end as well, see example #2 here.

It’s not quite production ready, but this could be improved on with tuning.
Read 10 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 on Twitter!

Send Email!

:(