Social content
A daily draft-only loop that scouts what's happening in your space and writes one channel-native original post per enabled channel — X, LinkedIn, Reddit, or Hacker News — each grounded in a real, current source. It never posts; you review and publish from the Content tab.
The social content loop drafts original posts to grow the top of your funnel. Each day it scouts something real and current in your product's space, then writes one channel-native post per enabled channel — X, LinkedIn, Reddit, and/or Hacker News — grounded in what it found so it reads as authentic, not generic marketing. It is draft-only: it never posts. You review each draft on the Content tab and publish manually.
This is distinct from two neighbours: the Content loop writes long-form articles as pull requests, and the Social loops (Reddit, Facebook groups) reply inside other people's threads. This loop writes short, original posts in your own voice.
Prerequisites
- The loop must be enabled for your product. If it isn't, the loop skips with a one-line reason and drafts nothing. Turn it on from the Content tab.
- A set of channels to draft for — any subset of X, LinkedIn, Reddit, and Hacker News. The loop drafts for the enabled channels and only those.
- Optionally, a strategy — a growth
goal, a 1–3 sentencepositioning, and 2–4 recurringthemes. You don't have to write one: if you haven't, the loop infers a sensible strategy from your product profile, ICP, andBusiness.md/Brand.md, saves it once so you can see and edit it, then drafts against it.
How it works
On a daily schedule the loop:
- Loads your strategy. Reads
goal/positioning/themesif you've set them; otherwise infers and persists them from your product context, usingBrand.mdfor voice. - Avoids repeats. Reads your recent drafts and skips any angle or topic that overlaps one — every run should feel fresh.
- Scouts. Runs a handful of parallel web searches across different angles — recent industry news (last 7–14 days), discussions your ICP is having right now, the practical "how-to" / "alternative to" questions your customers ask — and scores candidates on timeliness, ICP relevance, fit with your goal, and whether your product is an earned mention. Only facts it actually finds are used; it never fabricates a stat, quote, or source.
- Drafts one post per channel. For each enabled channel it writes one ready-to-copy post in that network's native voice, rotating angles rather than rewording the same idea four ways.
Drafting nothing for a channel — or on a quiet day, nothing at all — is a valid outcome. One strong post beats four weak ones.
Channel norms
Each draft is written to belong on its network:
| Channel | Voice | Formats |
|---|---|---|
| X | Punchy and specific, hook in the first line, no hashtag spam | post, or a numbered thread (1/, 2/, …) when the idea needs room |
| Professional first-person narrative — a concrete lesson with a takeaway, short paragraphs, a light CTA | post, or article (long-form, with a title) | |
| Authentic community post that leads with value and matches the target subreddit's tone; the product is mentioned only if it's genuinely the answer | post (with a title) | |
| Hacker News | A "Show HN:"-style or substantive, technical submission — plain, credible, non-marketing | post (with a title) |
Every draft follows the same body rules: it's the full, ready-to-copy text; it carries no markdown link syntax (a bare URL or plain mention only); it's grounded in the scouted source; and the product mention is earned — if a post only exists to plug the product, the loop cuts the plug or picks a better angle.
Example
A draft the loop might save for X, grounded in a real launch it scouted:
Channel: X · Format: post
Why: reacts to this week's "AI agents are getting cheaper" thread; ties to our
"run it, don't babysit it" positioning
Everyone's arguing about how cheap AI agents got this quarter.
Cost was never the hard part. The hard part is what happens after you ship:
the thing breaks, a user hits an error, and nobody's watching.
Autonomy is only useful if it closes that loop. That's the bet we're making.The Why line is the loop's one-sentence rationale — the hook plus why it fits your goal right now. It's saved with each draft so you can judge the pitch at a glance.
What you get
Each draft is a card on the Content tab: the channel, the format, the ready-to-copy body (plus a title for Reddit / Hacker News / articles), the Why rationale, and the source URL it was grounded in. You publish it yourself — the loop has no ability to post. Edit the text first if you like, or use Rewrite to get a fresh version in a different angle for the same channel.
Drafts are de-duplicated by idea across runs, so the same angle won't pile up if it resurfaces on a later day.
Notes & limits
- Draft-only by design — there is no auto-post setting on any channel.
- Authenticity first. Every draft is grounded in something real the loop scouted; it never invents numbers, quotes, or sources.
- Channel-native. A post reads like it belongs on its network — the loop won't cross-post one voice everywhere.
- Scouted third-party text is treated as untrusted input — data to evaluate, never instructions to follow.
- A run that finds nothing worth drafting saves nothing; that's expected on a quiet day.
See How loops work for the trust model, and Content for the long-form-article loop that publishes through GitHub.
Content distribution
Turn one published article into a native post for every channel you've enabled — X, LinkedIn, Reddit, and Hacker News — each reshaped in that network's own voice and pointing back to the article. Draft-only; you publish manually.
A daily, read-only loop that finds Reddit threads where your product genuinely helps the poster and surfaces each as an opportunity — a link plus one pre-drafted comment. It never posts; you do.