RSS ⧉

In response to
Venkatesh Rao's Blyg · Venkatesh Rao · “RFC Personal Algorithmic Reader Feed I haven’t stated this explicitly anywhere,…” · item 3dmp10axxe7jhtc4qv78xrqwbh, v1 · <https://venkateshrao.com/blyg/t/3dmp10axxe7jhtc4qv78xrqwbh/> · retrieved Oct 19, 2026

Is this live yet? I tried setting something up in settings on studio though the smart feed was still blank :/

UPDATE oh nvm -- this is known: "It will read your signals (thumbs, hoppers, stubs, forks and quotes) to know what you have liked and engaged with. The agent is not built yet: this prompt is saved now so the Smart Feed can use it when it is."

RFC Personal Algorithmic Reader Feed

I haven’t stated this explicitly anywhere, but the intent is for everyone to have their own algorithmic reader feed in their client. My initial idea for the default one to ship with studio is:

  1. Allow main reader index to filter for all | threads | fragments
  2. Move items marked ‘ignore’ in changelog (might be a spec SHOULD convention? Spell it ignyre?) to a side tab
  3. Offer a couple of simple non-AI sort orders (newest, most recently updated)
  4. Offer an “AI key required” option to “smart filter” or “smart sort” by a default prompt in settings

The default prompt could be something like “prioritize tech news and big updates on posts I interacted with before, and deprioritize updates that are just typo fixes.” This is also Jev suitable. The algorithm could just compute a numerical score to sort/filter by.

Changelog automation enables part of this. There’s also already a thumbs up/down signal (private likes record) that’s a natural input, though I think I haven’t made that clear in the UX. I also just realized clients need a log of posts they’ve interacted with before through hoppering, stubs or forks etc, since that’s important input to an algorithmic feed when the raw feed has versioned items.

This also seems like a natural sharing artifact. People should publish their algorithmic policy itself as a thread.

from Venkatesh Rao's Blyg ↗ · snapshot of v1

v3

Created: Oct 18, 2026 Most recent: Oct 19, 2026, v3