Home | Search

2026-10-04

My guesses for how to build social media forum running on ethereum blobdata

Disclaimer

• Quick Note

Main

• See also: Review of censorship-resistant storage soluions
• See also: Script: I publish daily archives of my website to ethereum blobdata
• Ethereum blobdata is the most censorship-resistant storage on the planet, while also being affordable for daily use. It can store both leaked classified information and child porn, and therefore also everything else.
• Getting censorship-resistance and affordability at the same time is a huge deal. If you manage to solve discoverability as well, I do see this as somewhat of a holy grail for decentralised social media forums.
• Why indexing is a problem
  • The above repo lets you post arbitrary tarballs to blobdata, but the entire tarball must be published from only from one ethereum address.
  • Ideally, you want to use ethereum blobdata for personal writings, 1-to-1 chats and social media forums.
  • Ideally, you want to use ethereum blobdata as the native place of storage, not a place to crosspost a discussion that already happened elsewhere. Archives can be crossposted using a single ethereum address, but discussions native to blobdata will need to happen across multiple addresses
  • Unfortunately, writing good indexers for blobdata is a problem for 1-to-1 chats and social media forums. Etherscan and blobdata have indexers already, but I found the code complicated.
  • Etherscan AFAIK does publish frontends based on their indexer that let you track 1-to-1 chats. As long as you trust etherscan (lol, lmao), you can already do 1-to-1 chats.
  • There is no native support for social media forums, however, so AFAICT you will have to write significant amount of code to build the indexer and frontend for this yourself.
• AFAIK you can't avoid writing an indexer that directly plugs into a full node.
  • Blobscan publishes IPFS archives but these are slow and have low upload bandwidth.
  • Full nodes (across all clients) store data in a complicated DB format, so it is not trivial to directly do bulk reads on this DB. There is Snappy compression and PeerDAS data column stuff and SSZ encoding.
• After you have built an indexer and a frontend, you will also need to build a user-facing search engine.
  • This search engine will need to implement heuristics like grepping for english dictionary words and magic numbers, since most of what gets posted to blobdata is only machine-readable (and hence considered spam from pov of a social media forum).
  • After it gets the human-readable content, then you will need to implement features like embedding search, full AI inference, and likes/follows-based search.

I decided not to work on this right now on this because this entire project seems like multiple months worth of effort, and I have other priorities right now. I will likely update this writeup if I ever end up building some of this myself.

Subscribe

To subscribe, enter "Subscribe " followed by Signal username, email address or whatsapp phone number. You will receive atmost one update per month.

Alternatively, you can subscribe via RSS, or a third-party service like blogtrottr

Comment

Enter comment