Marketing & Sales · List
13 Things to Know When Starting a New X Account
Growth principles derived from the open-source X recommendation algorithm
A 13-point breakdown analyzing how the open-sourced X recommendation algorithm scores and ranks content for new accounts. It covers how embeddings, replies, video dwell time, and negative signals shape audience distribution. It is useful for founders, creators, and marketers looking to optimize their reach on X based on the actual codebase.
The list
- Out-of-network retrieval is your only real growth path; in-network is nothing without followers — content must constantly reach a new audience.
- Embeddings lock you in: post off-niche once and discovery dies for weeks; stay on-dimension.
- Video demos are non-negotiable: looping clips spike dwell and p(video_view); text-only gets buried.
- Replies weigh massively in the weighted score — turn every tweet into subtle conversation bait.
- Reply to your own replies fast; engagement loops train the transformer on you.
- Negatives (blocks/mutes/reports) hit small accounts way harder — one bad wave = semi-permanent suppression.
- Avoid polarizing takes early; they hijack your embeddings toward drama clusters and the wrong audience forever.
- One power-user reply/repost amplifies your signal disproportionately — add real value in their threads.
- Curiosity hooks + absurdity spike initial dwell and set up the snowball.
- Early moderate hits (10–50k views) reinforce your niche cluster fast — ship quantity in your lane.
- The Grok transformer predicts multi-actions end-to-end: optimize for p(reply) + p(repost) over likes.
- No hand-engineered features: whatever maximizes real engagement wins — pure chaos optimization.
- Clone the repo and run a toy ranker on your X data to predict what will do well before posting (better with larger training data; the real scorer is trained on all of X).
What it's for
- Structure an account launch strategy to maximize algorithmic reach on X
- Optimize post formats for metrics heavily weighted by the ranking transformer
- Avoid common penalties like embedding drift, mutes, and negative feedback spikes
Key points
- Focuses on out-of-network retrieval mechanics essential for zero-follower accounts
- Explains how embedding clusters lock accounts into specific topical dimensions
- Highlights the scoring weight of video dwell time, loops, and conversational replies
- Details the disproportionate impact of negative signals (mutes, blocks, reports) on small profiles
- Advises running a local toy ranker from the open-source repo to test content predictions
Pricing
Free
Details
- Type
- List
- Category
- Marketing & Sales
- Topic
- Social Media Growth
- Open source
- No
- Platforms
- other
- For
- Founders, creators, and social media managers growing accounts on X
- Tags
- x, twitter, algorithm, social media, content strategy, audience growth
Why it was saved
- @ruslanjabari: Actionable playbook for growing new accounts under the current ranking system.
Related
Reddit AMA detailing rapid Instagram niche page growth tactics and posting frequency.
A 17-step growth and monetization playbook for X accounts using AI.
Configuration steps to maximize Instagram Reels upload quality and reach
Checklist and bio structure to improve Instagram profile discovery and conversions.
A practical 5-step framework for growing Instagram Reels reach and views
A workflow for finding, recreating, and scaling viral short-form videos.
Updated