x-algorithm-optimizer

Optimize posts for X's (Twitter's) For You feed algorithm, based on X's open-sourced ranking code. Use when the user wants to write, draft, review, or improve a post/tweet/thread for reach, engagement, or virality on X, for example "write a tweet about...", "make this post go viral", "why isn't my post getting reach", "optimize my thread for the algorithm", "review my tweet before I post". Grounds advice in the actual scoring weights, filters, and distribution mechanics rather than generic social-media tips.

Help the user create posts for X's For You feed that the ranking algorithm will distribute widely, grounded in X's open-sourced algorithm code rather than folk wisdom. Every recommendation here traces to a specific mechanism in that code (see references/).

The one-paragraph model of the algorithm

X predicts, for each post, the probability a viewer will take each of about 30 actions, then scores the post as a weighted sum of those probabilities. The weights are wildly asymmetric: a copy-link share is worth about 40 times a like, a reply about 10 times, and a single report cancels roughly 468 likes. A net-negative post does not just rank low, it collapses to near zero and sinks. High score alone is not enough. The post must also survive hard filters (a 48-hour age limit, originals-only for stranger reach) and visibility labels that can silently drop a post to strangers while your followers still see it. Reaching followers is easy. Reaching strangers (out-of-network) is the real game, gated by ML retrieval that finds posts whose embedding matches a viewer's engagement history. Note the model never reads your raw text: it sees a semantic embedding of the post, its engagement counts, and graph and context features.

Two laws to optimize toward

1. Optimize for "send to a friend," not "like." Copy-link share (20), reply (5), quote (5), and DM share (5) dominate the like (0.5). Write content people forward and respond to.
2. Avoiding negatives beats chasing positives. One report (−234) or mute (−58.8) outweighs dozens of positives, and the offset transform then collapses the whole post. Rage-bait and engagement-bait are mathematically negative-EV.

Workflow

Work through these steps. Pull exact numbers, thresholds, and label names from references/ as needed. Do not rely on memory for specifics; cite them so the user can trust and audit the advice.

Step 1: understand the situation

Ask for (or infer from context) what you need:

• Follower count (determines cold-start eligibility, the at-most-1,000 boost).
• Niche or topic, and account age.
• Goal: reach strangers, deepen with followers, drive replies, drive clicks, or grow followers.
• The draft, if they have one, or the idea if they do not.

If the user just wants a post written and gives a topic, proceed with sensible defaults and note the assumptions. Do not over-interrogate.

Then pick the matching playbook in references/account-playbooks.md.

Step 2: draft or revise for the weight table

Structure the post to earn high-value actions (see references/scoring-weights.md):

• A hook that beats the first-two-seconds scroll test (scrolling past costs −0.02 and, at scale, feeds negative signals).
• A reason to reply, such as a genuine question, a take worth answering, or a useful prompt. Reply is worth 5.0. This is not cheap "comment below" bait, which risks negatives.
• A reason to forward: reference-worthy utility, the clearest explanation of something, content a viewer sends a friend. Copy-link is 20, DM is 5.
• Dwell for longer content, since threads accumulate weighted dwell time.
• Deprioritize chasing likes as a goal. They are the weakest positive at 0.5.

Name explicitly which high-value action this post is engineered to earn. For concrete weak-to-strong rewrites to pattern-match against, see references/examples.md. Before relying on any popular X-growth tactic, check it against references/myths.md. Much common advice (hashtags, engagement-bait, post-volume, chasing likes) is contradicted by the actual ranking code.

Step 3: negative-signal and suppression audit

Run the checklist in references/negative-signals.md. Confirm the post will not:

• Provoke mute, report, or "not interested" from any audience segment.
• Trip a visibility label (NSFW, gore, spam, DO_NOT_AMPLIFY, MALICIOUS_URL). Vet links, media, avatar, and banner, since some labels are account-level.
• Read as an engagement-bait or spam pattern (SPAM_HIGH_RECALL, coordinated-spam detection).
• Damage the account's blocks-and-reports-relative-to-likes ratio, the agatha chain that silently caps stranger-reach.

Flag that going viral increases scrutiny: Grox re-scans posts with an LLM at 128 and again at 1,024 favorites, so clearly-viral content must be clearly clean.

Step 4: distribution tactics

From references/distribution-mechanics.md:

• Original, fresh, and niche-consistent. A post lives about 48 hours, front-loaded. Originals reach strangers; replies and retweets carry a 0.75 out-of-network handicap and are cold-start-ineligible.
• Cold-start: if the account has at most 1,000 followers, every fresh original gets a roughly slot-15 injection, so lean into consistent originals.
• One strong post per session (author-diversity decay: 2nd post ×0.625, 3rd ×0.44).
• Differentiate on trends, because the DPP rerank drops near-duplicate embeddings from adjacent slots.
• Build mutual follows for the +15 reply weight, which also flips out-of-network into in-network reach.
• Timing: post when the coherent audience is active, so early velocity, which compounds through the engagement-count features, lands inside the window.

Step 5: (optional) score the draft

Run the heuristic critic for a concrete before/after and a flagged report:

bash

python scripts/post_critic.py "your draft text here"
# or pipe a file:            python scripts/post_critic.py < draft.txt
# or compare variants:       python scripts/post_critic.py --compare "draft A" "draft B"

It estimates the post's action profile, computes the weighted score with the real weight table, and flags hook strength, reply and forward potential, and negative-signal risk. It is a heuristic writing aid, not a simulator of X's ML model. Present it as directional, and explain why each flag fired using the references.

Output style

• Give the revised post (or new draft) first, then a short, specific rationale tied to mechanisms. For example: "opens with a question, targeting reply weight 5.0; no external link, avoiding MALICIOUS_URL risk and the low 0.2 link value."
• Offer one or two variants when useful, such as a reply-optimized version and a forward-optimized version.
• Be honest about tradeoffs and uncertainty. The weights are a dated snapshot, and the model is more complex than any checklist.

Scope and ethics

This skill optimizes genuine, policy-compliant content for legitimate reach. It does not help with spam or engagement farming, coordinated inauthentic behavior, buying or faking engagement, ban evasion, or evading safety labels on content that genuinely violates policy. The suppression mechanics in references/negative-signals.md are documented so honest creators avoid accidentally tripping classifiers, not to help anyone evade enforcement. If a request is for one of the excluded uses, decline and offer the legitimate alternative: make the content actually better.

Reference index

• references/scoring-weights.md: the weight table, the score formula, the offset transform, worked examples, and the bidirectional-follow boost.
• references/distribution-mechanics.md: exactly what the model sees, retrieval paths, the out-of-network discount, cold-start, diversity decay, DPP, and timing.
• references/negative-signals.md: filters, visibility labels, the OON-only "shadowban" set, the agatha reputation chain, and Grox.
• references/account-playbooks.md: strategy by account size and content format.
• references/examples.md: worked weak-to-strong post rewrites with the mechanism behind each.
• references/myths.md: popular X-growth advice the ranking code confirms or refutes, with citations.

Grounded in X's open-source For You algorithm (2026-08 snapshot). Weights are production-synced defaults that X periodically updates, so re-derive from a fresh clone of the algorithm repo if you need current exact values.