from Hacker News

Show HN: Lean customer development methodology simulated by AutoGPT

by jchrisa on 4/19/23, 1:02 PM with 4 comments

Provide an elevator pitch and target customer, and GPT generates personas to interview, with catchy names like Serverless Sarah. Then simulate user interviews, summarize the outputs and provide follow-ups. It's inspired by the legendary startup book Four Steps to the Epiphany by Steve Blank.

I am building a new database (I was one of the creators of CouchDB / PouchDB) and wanted an example app that more fun than just database crud operations. Fireproof is designed to be the easiest way to add dynamic data to any app or page. It uses IPFS multiformat blocks for content addressing (and high-performance traditional infra like Cloudflare and WebRTC). What sets it apart from other browser-based database is that all the operations are cryptographically verifiable as Merkle proofs.

You can read more about Fireproof's database architecture here: https://fireproof.storage/documentation/how-the-database-eng...

  • by jchanimal on 4/19/23, 2:06 PM

    Of course serious prompts are more interesting, but you can have a lot of fun by asking it to analyze somomething silly like "candy shaped like Platonic solids for math professors." It'll even tell you when the customer would have no interest in the product.

    Also, there is a video if you want to see it in action: https://twitter.com/FireproofStorge/status/16485419508432568...

    React devs, to try Fireproof as your app's database, start here: https://fireproof.storage/documentation/usefireproof-hook-fo...

  • by jchanimal on 4/19/23, 7:14 PM

    The getting started docs to write an app basically look like: npm install @fireproof/core and import it to your page (or the React hook -- which I skipped for this app but use in other examples). Once connected you open a database and put and get documents. You can write CouchDB-style map indexes and all storage is content addressed and encrypted. Encrypted databases can be replicated to untrusted storage over HTTP, UCAN and IPFS, or synced with trusted collaborators via WebRTC or Socket Supply. The whole thing runs in the browser or server and relies on Merkle CRDT proofs for data verifiabiliy and lightwight diffs.

  • by carsonfarmer on 4/19/23, 4:54 PM

    Super fun example of the combination of AI and DB. You want to get back exactly what you put in, use a DB! You want something extrapolated from what you put in, AI is pretty great! Also Fireproof is :fire:

  • by ZaphodB17 on 4/19/23, 1:18 PM

    This is awesome! 10/10! Fireproof is a great use of IPFS’s verifiability