𝕏 Feed Tune

Template id: x/tune · Trigger: Manual · Flow timeout: none (runs until you stop it)

𝕏 Feed Tune teaches an account's For you your topics. It searches the words people use when they post about your topics, reads the latest posts, likes the ones that are really about them, and now and then opens the author's profile: when their own recent posts are about your topics too, it follows them. Run it on the account your 𝕏 Auto Reply uses: within a few days its For you shows more of these posts, so Auto Reply finds more of them.

What you need

• Your topics in Settings › Brand › Topics (or the campaign's More topics). Without any, the run stops with "add them in Settings › Brand".
• A Dataset › Account with platform 𝕏, signed in in the campaign's browser. Pick the browser your Auto Reply campaign uses (campaign › Browsers).
• A Dataset › AI key.
• A campaign in its own group (e.g. tune): its memory of liked posts and checked people stays apart from Auto Reply's.
• Give it time slots that don't overlap the Auto Reply campaign's (e.g. Feed Tune 07:00–09:00 and 19:00–21:00): one account doing two things at once doesn't look like a person.

How it picks

• Posts: only posts where the person is really talking about one of your topics, not a word in passing; a round with none picks nothing and rests. Skip and Anything else work as in Auto Reply. A liked post is kept as liked and never picked again.
• People: of the posts it likes, Check the author (%) of them get a look: their profile (bio, followers) and latest posts are read, then the AI decides follow or skip: follow only when most of their own posts (not reposts) are about your topics and it's a real person, creator or community, never a store, seller, promo, giveaway, bot or spam account. Who to follow, more adds your own rules (e.g. at least 50 followers). Someone already followed or private is skipped, and a checked person isn't checked again that day.
• Limits: Likes at most and Follows at most (set here, or the account's own like and follow limits in Dataset › Account). Both count only a like or follow that happened.

Searches from your list, the AI, or both

Searches from (Collect, or Picking on Facebook) decides where its searches come from:

• My list: only the searches you write (timeline.query).
• The AI: an AI step writes Searches it writes short, varied searches from your topics (and What the AI should search, e.g. games popular in Indonesia right now). When they all run dry, or a round finds nothing on topic, it rests a minute or two and writes new ones. It's told how many on-topic results the last ones found and remembers what it gave before (per account), so it drops searches that found nothing.
• Both: yours, plus the AI's.

Options

The flow

flowchart LR
  manual["Manual"] -- next --> topicCheck["Topics set?"]
  topicCheck["Topics set?"] -- true --> topicStop["No topics"]
  topicCheck["Topics set?"] -- false --> signin["Sign in"]
  signin["Sign in"] -- next --> sourceIf["Searches from my list?"]
  searches["Searches"] -- next --> timeline["Open the searches"]
  timeline["Open the searches"] -- next --> collect["Collect posts"]
  collect["Collect posts"] -- next --> choose["Pick on-topic posts"]
  collect["Collect posts"] -- empty --> emptyIf["New searches?"]
  emptyIf["New searches?"] -- true --> thinkRest["A short rest before new searches"]
  emptyIf["New searches?"] -- false --> idle["Nothing new: rest"]
  idle["Nothing new: rest"] -- next --> collect["Collect posts"]
  choose["Pick on-topic posts"] -- next --> loop["Each post"]
  choose["Pick on-topic posts"] -- none --> noneIf["New searches?"]
  noneIf["New searches?"] -- true --> thinkRest["A short rest before new searches"]
  noneIf["New searches?"] -- false --> roundRest["Rest before the next round"]
  roundRest["Rest before the next round"] -- next --> collect["Collect posts"]
  loop["Each post"] -- item --> open["Open post in a new tab"]
  loop["Each post"] -- done --> roundRest["Rest before the next round"]
  open["Open post in a new tab"] -- next --> dwell["Read the post"]
  dwell["Read the post"] -- next --> like["Like the post"]
  like["Like the post"] -- next --> followIf["Check the author?"]
  like["Like the post"] -- limited --> closeLimit["Close tab, like limit"]
  closeLimit["Close tab, like limit"] -- next --> likeWait["Like limit: wait for the next slot"]
  likeWait["Like limit: wait for the next slot"] -- next --> collect["Collect posts"]
  followIf["Check the author?"] -- true --> profile["Read their profile"]
  followIf["Check the author?"] -- false --> close["Close tab"]
  profile["Read their profile"] -- next --> knownIf["New to us?"]
  profile["Read their profile"] -- missing --> close["Close tab"]
  profile["Read their profile"] -- error --> close["Close tab"]
  knownIf["New to us?"] -- true --> judge["Do they post about our topics?"]
  knownIf["New to us?"] -- false --> close["Close tab"]
  judge["Do they post about our topics?"] -- next --> judgeIf["Follow?"]
  judgeIf["Follow?"] -- true --> follow["Follow them"]
  judgeIf["Follow?"] -- false --> close["Close tab"]
  follow["Follow them"] -- next --> close["Close tab"]
  follow["Follow them"] -- limited --> close["Close tab"]
  close["Close tab"] -- next --> wait["Pause after the post"]
  wait["Pause after the post"] -- next --> loop["Each post"]
  sourceIf["Searches from my list?"] -- true --> listSet["My searches"]
  sourceIf["Searches from my list?"] -- false --> think["Write new searches"]
  listSet["My searches"] -- next --> searches["Searches"]
  think["Write new searches"] -- next --> aiSet["The AI's searches"]
  think["Write new searches"] -- invalid --> listSet["My searches"]
  think["Write new searches"] -- declined --> listSet["My searches"]
  aiSet["The AI's searches"] -- next --> searches["Searches"]
  thinkRest["A short rest before new searches"] -- next --> think["Write new searches"]

FAQ

Does a test run like or follow?

No. Like and Follow are effect nodes: a test run reads the searches and profiles, picks and decides for real, and logs what it would like and follow. Read the run's log to see each pick, each profile and the AI's reason.

How do I know it works?

Open the Auto Reply campaign's Charts: Posts seen and Posts on topic per day. A rising share of on-topic posts means For you is learning your topics. The Feed Tune campaign counts its Likes and Follows.

Why not just follow everyone it likes?

One post about your topic doesn't make a person who posts about it. Following people whose own posts are about your topics is what fills For you with them.