DueDilly v1.0 - "Reputation analytics with AI" 👍👎

AI due dilligence agent for crypto

Think: "Moody's - credit ratings, research, and data for global capital markets." But For crypto using AI. AAA+++ or DDD---?

Crypto DD is very laborious, but you do the same steps everytime.
Due dil in crypto is hard. always moving. social proof can be faked. Dune.xyz is too pro, takes too much time.
Data is avialable but sometimes faked. How do you know whats what, you need a trusted source?

DueDilly.xyz uses the same proven expert DD tactics on every run
DueDilly.xyz Uses state of the art reasoning AI models to make sense of data. Scoring each param from 0-10. 24/7
DueDilly.xyz is an AI agent that can analyze any project on demand. Has a leader-board with live DD on breakout projects

Input field for urls to DD on. (5 DD calls are free. 20$ a month unlimited calls)
Get notified when projects DD score drops. So you can get ahead of the market (Free service for popular projects)
Everyone has access to emerging leaderboard data. top 10 (market-movers / buzzing / breakout / rising etc)
Twitter account where DD is published daily on new projects. This creates buzz and new customers.

Microsoft playwright cloud: 60min24h$0.01 = $14.4 per 24h use. (assuming PW cloud uses non blocked IP's, research this)
Unit economics: Avarage DD. Will take 10min on avarage. 1440min in 24h. 144 DDs = $14.4 (lower cost to run these on own HW)
Backend testing can be done in Github actions. as it supports running playwright and can receive POST CURL from a website.
Local testing can be done in CLI. With python + playwright lib
OpenAI LLM API calls for rasoning. Over all goal insruction. Path straegy reasoning. And screen step reasoning.
For each screen step. Send json of UI hierarchy from playwright to OpenAI. Mix in overall goal. previous steps completed. template strategy for DD type.
AI interprets next "intent step" based on accumulated context. Recursive recovery strategy, when dead-ends occure. timeout if all strategies fail. return nil. From these available intents. which intent should be picked in current context to reach overall goal xyz. Determine if intent was success or failure. Read new state. Make new move. etc etc

Social DD: Collect social media account links on landingpage. Check these accounts with socialblade like services for authenticity and reach
Chain DD: use dune.xyz or smth to verify some basic block stats. Life. Cash Flow. Wallet count etc.
Investor DD: Collect investors from landing page. Use crunchbase to check investor stats. Assets on hand. Invest count. Life etc
Discord DD: Log into Discord. Check activity last 24h. Check user count logged in.
Check google news. Cointelegrap etc for any breaking news. Sentiment bull or bear?
twitter / social news. Any famous accounts making buzz?
Founder DD. Collect linkedin / github / twitter / crucnhbase. Count followers. notable achivements etc.

Prototype:

Get button w/ title / id
Scroll page down / up while assert == false
get text w/ title / id
get inputfield w/ id
get paragraph w/ text / id
get element w/ id / text
go to url
OpenAIFindElement("find social links", page_hierercach.json) -> [element1, element2]
OpenAISummary("summaries due diligence, key points, pros and cons", collected_material)
press button with id / text (find btn, press btn)
get string for element with id / text (find text, copy content of text)
go back to previous page (keyboard combo?)
start over (go back while assert == true)

Social DD:
1. OpenAIFindElement("find social links")
2. Filter [Twitter, LI, GH]
- twitter. socialblade.com/user/twitter-user-name -> score
- Github. github.api/github-user-name -> followers, total stars
- LI: go to website via PW. OpenAIFindElement("find follower element, and whats the count?", playwright.json)
1. compare results to arbitrary benchmark. get avarage score. from 0-10
For the other DD steps. Use api calls and playwright when there is no API
Use AWS Lambda or EC2 to run python workflows. (since python can run and easily be developed lcoally) (price is near free)
OverAll summary is also returned in the Due dil.

Static website. Hosted on Github pages. (secure and free, and supports semi-large traffic spikes)
SVG-Logo + basic css + html
One url text input field ("Enter url to perform Due dil with AI"). [Enter Button]
top 10 DD leaderboard with scores. (Newcomers top 10 list with scores)
Social account link (twitter)
Enter email form. DD process will ping once complete. 10-20min.
Detail-page. DD type scores. And summary.

Private beta (we cover API cost, send only to trusted friends, send secret url)
API access to social blade, cointelegraf, dune.xyz etc etc.
Making playwright recovery loops. Or give up after x amount of tries etc.
Adding nodejs + passport.js later or smth for auth. or supabase for signup + auth etc. something cheap
Add demo account. 5 DD calls on the house. Test with friends of friends. Post in SoMe. max 1000 users.
Adding stripe payment integration if we see PMF from testing with friends and industry peers.
Testing PMF. We can slide up into twitter crypto accounts comment fields and roast their ratings w/ url. = traffic

Hack together a solution super fast. 24h hackathon. Make things work in local CLI. then add python scrpts to GA/AWS
Launch to peers and friends after 24h. Collect feedback and tweak things in the next 24h.
Decide to add singup + demo 5 calls on the house.
Decide to add stripe subscription for full access. Soft public launch on social media.
If we see traction. People paying. We know we have PMF. Just make sure infra scales. and we can go for full launch on producthunt etc. Sky is the limit.