Setting up your sources

What a source is

loom scans the sources you give it 24/7 and drafts in your voice from new content. There are two kinds: X accounts you watch (it drafts replies/quotes to their new tweets) and RSS/news sites (it produces tweets/threads from new articles). A mix of 3–5 is a good start — a single source produces single-type drafts, a mixed set brings variety.

Watching an X account

Add a @username from Sources or the task center; loom scans that account's new tweets. You can paste several @handles into the box in the task center and assign dozens of watch tasks at once.

• +reply — a reply draft in your voice to the watched account's new tweet
• +quote — a quote draft commenting on the tweet (quotes always go to your approval)

Reply language: if you picked a language on the source card, that wins; otherwise loom detects the source tweet's language and the REPLY ALWAYS MATCHES IT (Turkish reply to a Turkish tweet, English to English — a reply not in the other person's language goes unseen). For very short tweets where detection fails, the global "reply language" setting applies. "Regenerate" on an approval card also lets you pick the language. These settings affect REPLIES only: quote, remix and thread drafts are always written in your content language — even when the source is English.

Topic focus (niche account): fill the "topic focus" field in settings and loom only drafts replies/quotes for posts in those areas — off-topic tweets (even ones mentioning 'AI' as a joke) are skipped without spending credits. Fill it to grow a niche account; leave it empty and loom drafts on any topic.

Adding an RSS / news source

Pasting a site address is enough. Even if you paste just the homepage (e.g. techcrunch.com), loom tries to find the feed automatically: first the page's RSS link, then common addresses (/feed, /rss.xml, /?feed=rss2...). If it finds one, it offers it to you as "did you mean this" — it doesn't add silently.

Engagement tier: core · rotation · discovery

For each watched X account you choose "how much to engage" — this single axis sets both the engagement rhythm and the daily volume:

• +core — favorite accounts: frequent + plenty of engagement (short wait, higher daily cap)
• +rotation — most accounts: balanced rotation, everyone gets a turn
• +discovery — opportunity pool / new accounts: rare + few (only on rising tweets)

This tier prevents piling onto the same account — loom distributes its engagement fairly across accounts. You pick the tier from the account card in the task center.

Conversation continuity (incoming replies)

If you enable "conversation continuity" in settings, loom regularly scans replies coming to your posts and drafts a follow-up for the ones worth continuing. While drafting, it also sees your original post as context — so the reply follows the flow of the conversation, written in the other person's language (Turkish to Turkish, English to English).

• +In the approval queue these drafts carry a "conversation" tag — none are posted automatically, you approve them.
• +No draft is produced for emoji-only / one-word reactions ("👍", "lol") — a like is enough for those.
• +Replies older than 24 hours don't get a follow-up (reviving a cold conversation looks unnatural).
• +Ping-pong with the same person stays human-scale — the per-account daily cap applies here too.

Answering replies is one of the most valuable signals on X: in X's open-source ranker a reply carries 10× the weight of a like — and a reply on your own post from someone you mutually follow ranks among the highest-weighted actions there is.

An honest limit: replies do not reach the feeds of people who don't follow you (the ranker filters out-of-network replies). So replies buy you RELATIONSHIPS, not reach — they're for the person you're answering and for those reading that thread. If you're after impressions, weight your own original posts.

Source limit and extra slots

On Power you can watch 30 X accounts + 10 RSS sources (separate budgets). If you need more, you rent extra slots with credits: 150 credits / 30 days (per source type, auto-renewing). If credits run short, only that source type's scanning pauses — your sources aren't deleted.