task_backlog.md

Task backlog for project management purposes

Priority set: Low, Medium, High Complexity set: Low, Medium, High

Contributions

UX

• Optimise sorting of topics on Google Sheet
  • Priority: Low. Context: right now the topics are a single written to each cell. I believe the best UX would be a dropdown list which is populated for each cell such that a given user could select (ideally from a single column) all the tags of interest.
• Optimise sorting of authors on Google Sheet
  • Priority: Low. Context: right now the authors are a single written to each cell. I believe the best UX would be a dropdown list which is populated for each cell such that a given user could select (ideally from a single column) all the authors of interest.
• Improve contribution UX for links beyond form
  • Priority: Low
• Create Twitter bot to contribute either a whole thread when called, else the links in the tweet content
  • Priority: Low
• Add a safe zotero parser and filter for MEV contributions
  • Priority: Medium
• Run routine / event-based script on Typeform Results content to be uploaded
  • Priority: Medium
• Create new URL patterns with LLM call given existing URL base
  • Priority: Medium
  • Context: adding url patterns is tedious and low added value. It would be great to have a script that can generate the URL patterns given a base URL and knowledge base of previous URL patterns.

Content

• Add the content from bookmarks dotted all over
  • Priority: High
• Create a community-led curation mechanism?
  • Priority: Medium
  • Context: It would be powerful for community to flag paper complexity and create a "MEV onboarding roadmap" from beginner to advanced. In the best case scenario, there would be a large amount of inbound content. We would hopefully need to curate its complexity sooner or later.

Research exploitation UX

• Scrape all websites for new uploads and add links to uploads in most recent order on a single worksheet

  • Priority: High

• Add a podcast tab sorted by most recent of content from the podcasts handle worksheet

  • Priority: Medium
  • Context: Right now besides going into each Podcast radio, there is no way to see the most recent content. This would be a powerful way to keep up to date with the latest content.

• Add Twitter bot to Tweet for each new scrapped research document

  • Priority: Low

• add GPT-generated tags to each link

  • Priority: Low

mev.fyi governance

• Deprecate typeform for scalable open-source solution with links waiting to be uploaded
• Decentralise domain name ownership
• Decentralise google sheet ownership
  • Context: for now the google sheet is updated by a single person via a google service account which is quite centralised. Technically if the .env were pointed to another sheet that can be replicated or become the new source of truth (would require a new url forwarding which is easy to do).
• Decentralise github repo ownership

Codebase

Devops

• Create Google sheet staging environment
  • Priority: Medium
• Host site on Google Cloud
  • Blocker: Streamlit LLM-app frontend
  • Priority: Low

Back-end

• Add support for Twitter threads

  • Should they be written as blogs first?
  • Priority: High

• Check https://apify.com/apify/website-content-crowler

  • Priority: Low

LLM chatbot and inference

• Design architecture for LLM-based chatbot for papers, blogs, podcast recommendation (backend)
  • Priority: Medium
  • Context: There is a trade-off between using off-the-shelf solution (IBM Watson Discovery for summarizing + service client like Intercom) vs. building from scratch with Langchain + vector database and HuggingFace or GPT backend. Latter likely more scalable with open-source contributions.
• Add LLM-based chatbot for recommending papers, blogs, podcast recommendation (backend)
  • Priority: Medium
• Generate MEV glossary, definitions based on Youtube transcripts, podcasts, papers
  • Priority: Low
• Improve transcript generation e.g. with LLM inference (AssemblyAI) on .mp3 youtube video content
  • Priority: Low

Frontend

• Add Streamlit frontend for LLM-based chatbot for papers, blogs, podcast recommendation
• Consider moving from Google Sheets to self-hosted website for analytics purposes (and UX?)
  • Priority: Medium
  • Context: it is hard to gauge how successful the project is without analytics. This is a trade-off between privacy (would rely on cookies) and analytics. Such analytics could help possibly (if approved/validated/not rejected) monetise the project e.g. by adding a job tabs / sponsored content and re-invest proceeds into the development of the project. Such analytics would help on the business development part.