feat(hands): complete i18n fixes, SKILL.md enhancements, and README overhaul
- Fix French accent characters (é/è/ê/ç/â/ô) across all 14 HAND.toml files - Fix German special characters (ä/ö/ü/ß) across all 14 HAND.toml files - Add category translations to all 6 i18n language blocks in all 14 hands - Enhance SKILL.md content for 9 hands with practical examples and workflows - Trim bloated SKILL.md files (apitester 1400→892, devops 1301→870) - Rewrite root README.md with accurate stats, complete hand/integration tables - Update hands/README.md with full 14-hand listing and i18n documentation
This commit is contained in:
Evan Hu committed
1 parent
315f955ce2
commit
33d279889c
27 files changed
+10001
-78
No files matched your search
@@ -526,6 +526,313 @@ frequency = "hourly"
|
||||
token_consumption = "medium"
|
||||
default_active = false
|
||||
|
||||
# ─── Internationalization (optional) ─────────────────────────────────────────
|
||||
# All i18n sections are optional. Without them, the English values above are used.
|
||||
# To localize, add [i18n.LANG] sections (e.g. zh, ja, ko, es, fr, de).
|
||||
# Settings translations are also optional — omit to keep English labels.
|
||||
|
||||
# ─── Chinese (简体中文) ────────────────────────────────────────────────────
|
||||
|
||||
[i18n.zh]
|
||||
name = "Twitter Hand"
|
||||
description = "自主 Twitter/X 管理——内容创作、定时发布、互动管理和效果追踪"
|
||||
category = "通信"
|
||||
|
||||
[i18n.zh.settings.twitter_bearer_token]
|
||||
label = "Twitter Bearer Token"
|
||||
description = "Twitter/X 开发者平台的 Bearer Token，所有 Twitter API 操作均需此凭证。"
|
||||
|
||||
[i18n.zh.settings.twitter_style]
|
||||
label = "内容风格"
|
||||
description = "推文的语气和风格"
|
||||
|
||||
[i18n.zh.settings.post_frequency]
|
||||
label = "发布频率"
|
||||
description = "创建和发布内容的频率"
|
||||
|
||||
[i18n.zh.settings.auto_reply]
|
||||
label = "自动回复"
|
||||
description = "自动回复提及和相关对话"
|
||||
|
||||
[i18n.zh.settings.auto_like]
|
||||
label = "自动点赞"
|
||||
description = "自动为人脉网络中的推文和相关内容点赞"
|
||||
|
||||
[i18n.zh.settings.content_topics]
|
||||
label = "内容主题"
|
||||
description = "要创作内容的主题（逗号分隔，例如 AI、创业、效率提升）"
|
||||
|
||||
[i18n.zh.settings.brand_voice]
|
||||
label = "品牌调性"
|
||||
description = "描述你的独特风格（例如"用幽默简化复杂技术的创业者"）"
|
||||
|
||||
[i18n.zh.settings.thread_mode]
|
||||
label = "推文串模式"
|
||||
description = "在内容组合中加入推文串（多条推文组成的故事）"
|
||||
|
||||
[i18n.zh.settings.content_queue_size]
|
||||
label = "内容队列大小"
|
||||
description = "待发队列中保持的推文数量"
|
||||
|
||||
[i18n.zh.settings.engagement_hours]
|
||||
label = "互动时段"
|
||||
description = "检查提及和互动的时间段"
|
||||
|
||||
[i18n.zh.settings.approval_mode]
|
||||
label = "审批模式"
|
||||
description = "将推文写入队列等待审核，而非直接发布"
|
||||
|
||||
# ─── Spanish (Español) ────────────────────────────────────────────────────
|
||||
|
||||
[i18n.es]
|
||||
name = "Hand de Twitter"
|
||||
description = "Gestor autónomo de Twitter/X — creación de contenido, publicación programada, interacciones y seguimiento de rendimiento"
|
||||
category = "Comunicación"
|
||||
|
||||
[i18n.es.settings.twitter_bearer_token]
|
||||
label = "Token Bearer de Twitter"
|
||||
description = "Token Bearer del Portal de Desarrolladores de Twitter/X. Requerido para todas las operaciones de la API de Twitter."
|
||||
|
||||
[i18n.es.settings.twitter_style]
|
||||
label = "Estilo de contenido"
|
||||
description = "Voz y tono para tus tweets"
|
||||
|
||||
[i18n.es.settings.post_frequency]
|
||||
label = "Frecuencia de publicación"
|
||||
description = "Con qué frecuencia crear y publicar contenido"
|
||||
|
||||
[i18n.es.settings.auto_reply]
|
||||
label = "Respuesta automática"
|
||||
description = "Responder automáticamente a menciones y conversaciones relevantes"
|
||||
|
||||
[i18n.es.settings.auto_like]
|
||||
label = "Me gusta automático"
|
||||
description = "Dar me gusta automáticamente a tweets de tu red y contenido relevante"
|
||||
|
||||
[i18n.es.settings.content_topics]
|
||||
label = "Temas de contenido"
|
||||
description = "Temas sobre los que crear contenido (separados por comas, ej. IA, startups, productividad)"
|
||||
|
||||
[i18n.es.settings.brand_voice]
|
||||
label = "Voz de marca"
|
||||
description = "Describe tu voz única (ej. 'fundador sarcástico que simplifica la tecnología compleja')"
|
||||
|
||||
[i18n.es.settings.thread_mode]
|
||||
label = "Modo de hilos"
|
||||
description = "Incluir hilos de tweets (historias de múltiples tweets) en la mezcla de contenido"
|
||||
|
||||
[i18n.es.settings.content_queue_size]
|
||||
label = "Tamaño de la cola de contenido"
|
||||
description = "Número de tweets a mantener en la cola preparada"
|
||||
|
||||
[i18n.es.settings.engagement_hours]
|
||||
label = "Horario de interacción"
|
||||
description = "Cuándo verificar menciones e interactuar"
|
||||
|
||||
[i18n.es.settings.approval_mode]
|
||||
label = "Modo de aprobación"
|
||||
description = "Escribir los tweets en un archivo de cola para revisión en lugar de publicarlos directamente"
|
||||
|
||||
# ─── Japanese (日本語) ────────────────────────────────────────────────────
|
||||
|
||||
[i18n.ja]
|
||||
name = "Twitter Hand"
|
||||
description = "自律型Twitter/Xマネージャー——コンテンツ作成、スケジュール投稿、エンゲージメント管理、パフォーマンス追跡"
|
||||
category = "コミュニケーション"
|
||||
|
||||
[i18n.ja.settings.twitter_bearer_token]
|
||||
label = "Twitter Bearerトークン"
|
||||
description = "Twitter/X開発者ポータルのBearerトークン。すべてのTwitter API操作に必要です。"
|
||||
|
||||
[i18n.ja.settings.twitter_style]
|
||||
label = "コンテンツスタイル"
|
||||
description = "ツイートの語調とスタイル"
|
||||
|
||||
[i18n.ja.settings.post_frequency]
|
||||
label = "投稿頻度"
|
||||
description = "コンテンツの作成・投稿の頻度"
|
||||
|
||||
[i18n.ja.settings.auto_reply]
|
||||
label = "自動返信"
|
||||
description = "メンションや関連する会話に自動で返信する"
|
||||
|
||||
[i18n.ja.settings.auto_like]
|
||||
label = "自動いいね"
|
||||
description = "ネットワーク内のツイートや関連コンテンツに自動でいいねする"
|
||||
|
||||
[i18n.ja.settings.content_topics]
|
||||
label = "コンテンツトピック"
|
||||
description = "作成するコンテンツのトピック（カンマ区切り、例: AI、スタートアップ、生産性）"
|
||||
|
||||
[i18n.ja.settings.brand_voice]
|
||||
label = "ブランドボイス"
|
||||
description = "あなた独自の語り口を記述（例:「複雑なテクノロジーをわかりやすく伝える皮肉屋の起業家」）"
|
||||
|
||||
[i18n.ja.settings.thread_mode]
|
||||
label = "スレッドモード"
|
||||
description = "コンテンツミックスにツイートスレッド（複数ツイートで構成するストーリー）を含める"
|
||||
|
||||
[i18n.ja.settings.content_queue_size]
|
||||
label = "コンテンツキューサイズ"
|
||||
description = "準備キューに保持するツイートの数"
|
||||
|
||||
[i18n.ja.settings.engagement_hours]
|
||||
label = "エンゲージメント時間帯"
|
||||
description = "メンションの確認とエンゲージメントを行う時間帯"
|
||||
|
||||
[i18n.ja.settings.approval_mode]
|
||||
label = "承認モード"
|
||||
description = "ツイートを直接投稿せず、レビュー用のキューファイルに書き出す"
|
||||
|
||||
# ─── French (Français) ────────────────────────────────────────────────────
|
||||
|
||||
[i18n.fr]
|
||||
name = "Hand Twitter"
|
||||
description = "Gestionnaire Twitter/X autonome — création de contenu, publication programmée, interactions et suivi des performances"
|
||||
category = "Communication"
|
||||
|
||||
[i18n.fr.settings.twitter_bearer_token]
|
||||
label = "Jeton Bearer Twitter"
|
||||
description = "Jeton Bearer du Portail Développeurs Twitter/X. Requis pour toutes les opérations de l'API Twitter."
|
||||
|
||||
[i18n.fr.settings.twitter_style]
|
||||
label = "Style de contenu"
|
||||
description = "Ton et style pour vos tweets"
|
||||
|
||||
[i18n.fr.settings.post_frequency]
|
||||
label = "Fréquence de publication"
|
||||
description = "Fréquence de création et de publication de contenu"
|
||||
|
||||
[i18n.fr.settings.auto_reply]
|
||||
label = "Réponse automatique"
|
||||
description = "Répondre automatiquement aux mentions et conversations pertinentes"
|
||||
|
||||
[i18n.fr.settings.auto_like]
|
||||
label = "Like automatique"
|
||||
description = "Aimer automatiquement les tweets de votre réseau et le contenu pertinent"
|
||||
|
||||
[i18n.fr.settings.content_topics]
|
||||
label = "Sujets de contenu"
|
||||
description = "Sujets sur lesquels créer du contenu (séparés par des virgules, ex. IA, startups, productivité)"
|
||||
|
||||
[i18n.fr.settings.brand_voice]
|
||||
label = "Voix de marque"
|
||||
description = "Décrivez votre voix unique (ex. 'fondateur sarcastique qui simplifie la technologie complexe')"
|
||||
|
||||
[i18n.fr.settings.thread_mode]
|
||||
label = "Mode fil de discussion"
|
||||
description = "Inclure des fils de tweets (histoires à plusieurs tweets) dans le mix de contenu"
|
||||
|
||||
[i18n.fr.settings.content_queue_size]
|
||||
label = "Taille de la file de contenu"
|
||||
description = "Nombre de tweets à maintenir dans la file d'attente préparée"
|
||||
|
||||
[i18n.fr.settings.engagement_hours]
|
||||
label = "Heures d'engagement"
|
||||
description = "Quand vérifier les mentions et interagir"
|
||||
|
||||
[i18n.fr.settings.approval_mode]
|
||||
label = "Mode d'approbation"
|
||||
description = "Écrire les tweets dans un fichier d'attente pour révision au lieu de les publier directement"
|
||||
|
||||
# ─── German (Deutsch) ────────────────────────────────────────────────────
|
||||
|
||||
[i18n.de]
|
||||
name = "Twitter-Hand"
|
||||
description = "Autonomer Twitter/X-Manager — Content-Erstellung, geplante Veröffentlichung, Engagement-Management und Performance-Tracking"
|
||||
category = "Kommunikation"
|
||||
|
||||
[i18n.de.settings.twitter_bearer_token]
|
||||
label = "Twitter Bearer-Token"
|
||||
description = "Bearer-Token vom Twitter/X-Entwicklerportal. Erforderlich für alle Twitter-API-Operationen."
|
||||
|
||||
[i18n.de.settings.twitter_style]
|
||||
label = "Inhaltsstil"
|
||||
description = "Ton und Stil für Ihre Tweets"
|
||||
|
||||
[i18n.de.settings.post_frequency]
|
||||
label = "Veröffentlichungshäufigkeit"
|
||||
description = "Wie oft Inhalte erstellt und veröffentlicht werden"
|
||||
|
||||
[i18n.de.settings.auto_reply]
|
||||
label = "Automatische Antwort"
|
||||
description = "Automatisch auf Erwähnungen und relevante Gespräche antworten"
|
||||
|
||||
[i18n.de.settings.auto_like]
|
||||
label = "Automatisches Like"
|
||||
description = "Tweets im Netzwerk und relevante Inhalte automatisch liken"
|
||||
|
||||
[i18n.de.settings.content_topics]
|
||||
label = "Inhaltsthemen"
|
||||
description = "Themen für die Content-Erstellung (kommagetrennt, z.B. KI, Startups, Produktivität)"
|
||||
|
||||
[i18n.de.settings.brand_voice]
|
||||
label = "Markenstimme"
|
||||
description = "Beschreiben Sie Ihre einzigartige Stimme (z.B. 'sarkastischer Gründer, der komplexe Technologie vereinfacht')"
|
||||
|
||||
[i18n.de.settings.thread_mode]
|
||||
label = "Thread-Modus"
|
||||
description = "Tweet-Threads (mehrteilige Tweet-Geschichten) in den Content-Mix aufnehmen"
|
||||
|
||||
[i18n.de.settings.content_queue_size]
|
||||
label = "Größe der Content-Warteschlange"
|
||||
description = "Anzahl der Tweets in der vorbereiteten Warteschlange"
|
||||
|
||||
[i18n.de.settings.engagement_hours]
|
||||
label = "Engagement-Zeiten"
|
||||
description = "Wann Erwähnungen geprüft und interagiert werden soll"
|
||||
|
||||
[i18n.de.settings.approval_mode]
|
||||
label = "Genehmigungsmodus"
|
||||
description = "Tweets in eine Warteschlangendatei zur Überprüfung schreiben, anstatt sie direkt zu veröffentlichen"
|
||||
|
||||
# ─── Korean (한국어) ────────────────────────────────────────────────────
|
||||
|
||||
[i18n.ko]
|
||||
name = "Twitter Hand"
|
||||
description = "자율 Twitter/X 관리 — 콘텐츠 제작, 예약 게시, 소통 관리 및 성과 추적"
|
||||
category = "커뮤니케이션"
|
||||
|
||||
[i18n.ko.settings.twitter_bearer_token]
|
||||
label = "Twitter Bearer 토큰"
|
||||
description = "Twitter/X 개발자 포털의 Bearer 토큰. 모든 Twitter API 작업에 필수."
|
||||
|
||||
[i18n.ko.settings.twitter_style]
|
||||
label = "콘텐츠 스타일"
|
||||
description = "트윗의 어조와 스타일"
|
||||
|
||||
[i18n.ko.settings.post_frequency]
|
||||
label = "게시 빈도"
|
||||
description = "콘텐츠를 작성하고 게시하는 주기"
|
||||
|
||||
[i18n.ko.settings.auto_reply]
|
||||
label = "자동 답글"
|
||||
description = "멘션 및 관련 대화에 자동으로 답글 작성"
|
||||
|
||||
[i18n.ko.settings.auto_like]
|
||||
label = "자동 좋아요"
|
||||
description = "네트워크 내 트윗 및 관련 콘텐츠에 자동으로 좋아요"
|
||||
|
||||
[i18n.ko.settings.content_topics]
|
||||
label = "콘텐츠 주제"
|
||||
description = "콘텐츠를 작성할 주제 (쉼표로 구분, 예: AI, 스타트업, 생산성)"
|
||||
|
||||
[i18n.ko.settings.brand_voice]
|
||||
label = "브랜드 보이스"
|
||||
description = "고유한 스타일을 설명 (예: '복잡한 기술을 쉽게 풀어내는 유머러스한 창업자')"
|
||||
|
||||
[i18n.ko.settings.thread_mode]
|
||||
label = "스레드 모드"
|
||||
description = "콘텐츠 구성에 트윗 스레드 (다중 트윗 스토리) 포함"
|
||||
|
||||
[i18n.ko.settings.content_queue_size]
|
||||
label = "콘텐츠 대기열 크기"
|
||||
description = "준비 대기열에 유지할 트윗 수"
|
||||
|
||||
[i18n.ko.settings.engagement_hours]
|
||||
label = "소통 시간대"
|
||||
description = "멘션 확인 및 소통 활동을 수행하는 시간대"
|
||||
|
||||
[i18n.ko.settings.approval_mode]
|
||||
label = "승인 모드"
|
||||
description = "트윗을 직접 게시하지 않고 대기열 파일에 기록하여 검토"
|
||||
@@ -226,6 +226,72 @@ Do NOT auto-like:
|
||||
|
||||
---
|
||||
|
||||
## Advanced Engagement Patterns
|
||||
|
||||
### Quote Tweet vs Reply vs Retweet
|
||||
|
||||
Choosing the right interaction type determines whether you gain visibility or waste it.
|
||||
|
||||
**Use a Quote Tweet when**:
|
||||
- You have a distinct take or added context (not just "this!")
|
||||
- The original tweet has high impressions and you want to draft off its reach
|
||||
- You are crediting someone while adding your own insight for your audience
|
||||
- The original author has a similar or larger following (exposes you to their audience)
|
||||
|
||||
**Use a Reply when**:
|
||||
- You want to build a direct relationship with the author
|
||||
- Your comment only makes sense in context of the original
|
||||
- The original author has a much larger following (replies show on their thread, giving you visibility without looking self-promotional)
|
||||
- You are answering a question or adding a correction
|
||||
|
||||
**Use a plain Retweet when**:
|
||||
- The original says everything perfectly and you have nothing to add
|
||||
- You want to signal-boost a community member, customer, or partner
|
||||
- The content is time-sensitive (breaking news, event announcements)
|
||||
|
||||
**Avoid**:
|
||||
- Quote tweeting with only emojis or "this" -- adds no value, looks lazy
|
||||
- Quote tweeting someone with fewer followers just to dunk -- punching down
|
||||
- Retweeting more than 3-4 times per day -- dilutes your original content ratio
|
||||
|
||||
### Thread Repurposing
|
||||
|
||||
A thread that performed well contains 5-7 standalone content pieces. Extract them over the following week to maximize ROI.
|
||||
|
||||
**Process**:
|
||||
1. Day 0 (original): Post the full thread
|
||||
2. Day 2: Pull the single most quotable tweet from the thread. Post it standalone with slightly different wording. No link back to the thread
|
||||
3. Day 4: Turn a data point or example from the thread into a graphic or screenshot tweet
|
||||
4. Day 6: Post the thread's core thesis as a hot take (one tweet, punchy)
|
||||
5. Day 8+: If engagement stayed strong, post a "Part 2" thread that goes deeper on whichever tweet in the original got the most replies
|
||||
|
||||
**Rules**:
|
||||
- Change the wording each time -- copy-pasting feels like spam to followers who saw the original
|
||||
- Space extractions at least 48 hours apart
|
||||
- Stop if any extraction underperforms significantly -- the topic is tapped out
|
||||
- Never repurpose a thread that got low engagement; the content did not resonate
|
||||
|
||||
### Trending Topic Participation
|
||||
|
||||
**When to participate**:
|
||||
- The trend directly intersects one of your content pillars
|
||||
- You have a genuine, informed perspective (not a generic reaction)
|
||||
- The trend is still rising (check the "Trending" tab; if it has been trending for >12 hours, you are late)
|
||||
- The tone of the trend matches your brand voice
|
||||
|
||||
**When to avoid**:
|
||||
- Tragedy, disaster, or crisis events -- opportunistic posting destroys trust
|
||||
- Highly polarized political or social debates outside your expertise
|
||||
- Trends driven by outrage mobs -- associating your brand is high-risk, low-reward
|
||||
- You would need to force-fit your product or message into the trend
|
||||
|
||||
**Execution**:
|
||||
- Lead with your actual insight, not the hashtag. The hashtag goes at the end or is omitted entirely if the topic keyword is in your text
|
||||
- Be early or be different. If 50 people have already made the same joke, skip it
|
||||
- Tie back to your pillar: "Trend X is exactly why [your pillar topic] matters more than ever"
|
||||
|
||||
---
|
||||
|
||||
## Content Calendar Template
|
||||
|
||||
```
|
||||
@@ -254,6 +320,169 @@ Friday:
|
||||
|
||||
---
|
||||
|
||||
## Worked Examples
|
||||
|
||||
### Example 1: Product Launch Twitter Campaign (1-Week Plan)
|
||||
|
||||
**Context**: A dev tools startup is launching "FastDB," an open-source embedded database. The account has 2,400 followers, mostly backend engineers.
|
||||
|
||||
**Pre-launch (3 days before)**:
|
||||
- Seed curiosity without revealing the product name
|
||||
- Engage heavily in database-related threads to increase profile visits before launch
|
||||
|
||||
**Day 1 (Monday) -- Teaser**:
|
||||
```
|
||||
We've been heads-down for 8 months building something
|
||||
we think embedded databases have been missing.
|
||||
|
||||
Shipping it open-source this Thursday.
|
||||
|
||||
More soon.
|
||||
```
|
||||
Purpose: Create anticipation. No hashtags, no links. Let curiosity drive profile visits.
|
||||
|
||||
**Day 2 (Tuesday) -- Problem framing**:
|
||||
```
|
||||
SQLite is incredible for what it does.
|
||||
|
||||
But if you need concurrent writes, ACID transactions,
|
||||
AND sub-millisecond reads in the same embedded DB...
|
||||
your options get thin fast.
|
||||
|
||||
We've been living in that gap. Fix incoming Thursday.
|
||||
```
|
||||
Purpose: Define the problem space. People who feel this pain will follow for the reveal.
|
||||
|
||||
**Day 3 (Wednesday) -- Social proof / build-up**:
|
||||
```
|
||||
Shipped our embedded DB to 12 beta testers last month.
|
||||
|
||||
Results so far:
|
||||
- 4.2x faster concurrent writes vs SQLite WAL mode
|
||||
- Zero-config replication
|
||||
- Single static binary, 3.8 MB
|
||||
|
||||
One more day.
|
||||
```
|
||||
Purpose: Concrete numbers build credibility. "One more day" maintains tension.
|
||||
|
||||
**Day 4 (Thursday) -- Launch day thread** (6-tweet thread):
|
||||
```
|
||||
1/6 [HOOK]: Introducing FastDB -- embedded DB for concurrent-write-heavy
|
||||
workloads. Open source. Single binary. Here's why we built it:
|
||||
2/6 [PROBLEM]: SQLite = single-writer. Fine for reads, hits a wall on
|
||||
write-heavy apps (event logging, IoT, realtime sync). FastDB uses
|
||||
MVCC -- writers never block readers, readers never block writers.
|
||||
3/6 [PROOF]: Benchmarks (M2 Mac, 8 threads): concurrent writes 51K ops/s
|
||||
vs SQLite WAL 12K ops/s. Point reads on par at ~900K ops/s.
|
||||
4/6 [ONBOARD]: Getting started: `cargo add fastdb` then 3 lines of code.
|
||||
Full SQLite-compatible query layer coming in v0.2.
|
||||
5/6 [ROADMAP]: v0.1 ships ACID transactions, built-in replication, crash
|
||||
recovery, zero deps beyond libc. v0.2: SQL layer, S3 cold storage.
|
||||
6/6 [CTA]: Star the repo: github.com/example/fastdb -- open issues, roast
|
||||
the benchmarks, tell us what's missing.
|
||||
```
|
||||
Key structural choices: tweet 1 is a standalone hook, tweet 3 has hard numbers, tweet 6 ends with a specific ask (not just "check it out").
|
||||
|
||||
**Day 4 afternoon** -- Post a standalone tweet answering the most common reply question publicly (drives docs traffic). **Day 5 (Friday)** -- Reply to every substantive comment. Templates for common reactions:
|
||||
- "How is this different from X?" -> Concrete comparison, link to docs
|
||||
- "Benchmarks look suspicious" -> Link the reproduction steps, invite them to run it
|
||||
- "Will you support [feature]?" -> Link the tracking issue
|
||||
|
||||
**Day 6-7 (Weekend)** -- Repurpose: extract the benchmark tweet as a standalone with a chart image; post a "5 things I learned launching an open-source DB" reflection thread.
|
||||
|
||||
### Example 2: Building Thought Leadership from Scratch (Month 1)
|
||||
|
||||
**Context**: An individual ML engineer with 180 followers wants to become a recognized voice in applied machine learning. No existing audience. No viral content history.
|
||||
|
||||
**Core principle for month 1**: Do not broadcast. Contribute. Your first 500 followers come from being consistently useful in other people's threads, not from your own tweets.
|
||||
|
||||
**Week 1 -- Comment-first growth**:
|
||||
- Post 0 original tweets
|
||||
- Find 10 accounts in your niche with 5K-50K followers who post regularly
|
||||
- Reply to 5-8 of their tweets per day with substantive comments (not "great post!")
|
||||
- Goal: Get 3-5 of those authors to like or reply to your comments by end of week
|
||||
|
||||
**What a good reply looks like**:
|
||||
```
|
||||
Original tweet: "Fine-tuning LLMs is overrated. Most use cases
|
||||
are better served by good prompting + RAG."
|
||||
|
||||
Bad reply: "Agreed!"
|
||||
|
||||
Good reply: "Mostly agree, but there's a middle ground --
|
||||
LoRA fine-tuning on 500 domain-specific examples
|
||||
consistently beats RAG for structured extraction tasks.
|
||||
|
||||
We saw 23% higher F1 on invoice parsing after a 2-hour
|
||||
fine-tune vs our best RAG setup.
|
||||
|
||||
RAG still wins for open-domain QA though."
|
||||
```
|
||||
This reply adds data, shows experience, and invites further discussion. People reading the thread see your expertise and check your profile.
|
||||
|
||||
**Week 2 -- First original content**:
|
||||
- Continue the reply strategy (5/day minimum)
|
||||
- Post 2-3 original tweets. Keep them observational, not promotional:
|
||||
```
|
||||
Something I've noticed after fine-tuning 30+ models
|
||||
this year:
|
||||
|
||||
The quality of your eval set matters 10x more than
|
||||
the size of your training set.
|
||||
|
||||
50 carefully labeled examples with clear edge cases
|
||||
beats 5000 noisy scraped examples every time.
|
||||
```
|
||||
- Post 1 "ask the audience" tweet to start conversations:
|
||||
```
|
||||
ML engineers: what's the most counterintuitive lesson
|
||||
you've learned about deploying models to production?
|
||||
|
||||
I'll start: the model is almost never the bottleneck.
|
||||
Data pipelines are.
|
||||
```
|
||||
|
||||
**Week 3 -- First thread** (5-tweet authority thread):
|
||||
```
|
||||
1/5 [HOOK]: I've deployed 12 ML models to production this year. The ones
|
||||
that worked all had one thing in common. It wasn't the architecture.
|
||||
2/5 [THESIS]: Every success had a tight feedback loop -- predictions
|
||||
validated by a human within 24 hours, not "we'll evaluate next quarter."
|
||||
3/5 [EVIDENCE]: Model A (invoice classifier): accountants flagged errors
|
||||
same-day, retrained weekly, 84% -> 97% in 6 weeks. Model B (churn
|
||||
predictor): sales ignored outputs, no feedback 3 months, drifted to
|
||||
coin-flip accuracy.
|
||||
4/5 [FRAMEWORK]: The pattern: (1) deploy with human-in-the-loop review,
|
||||
(2) log every correction, (3) retrain on corrections every 1-2 weeks,
|
||||
(4) remove human review once accuracy stabilizes.
|
||||
5/5 [CTA]: If you're skipping the feedback loop, you're building on sand.
|
||||
What's your experience?
|
||||
```
|
||||
Notice the structure: personal credibility in tweet 1, a clear thesis in tweet 2, contrasting real examples in tweet 3, an actionable takeaway in tweet 4, and a discussion prompt in tweet 5.
|
||||
|
||||
**Week 4 -- Establish rhythm**:
|
||||
- Settle into a sustainable cadence: 1 thread/week, 1-2 standalone tweets/day, 5+ replies/day
|
||||
- Review metrics from week 2-3 content to identify which topics resonated
|
||||
- Double down on the topic that got the most replies (not likes -- replies indicate deeper engagement)
|
||||
|
||||
**Month 1 milestones**:
|
||||
| Metric | Target | Why it matters |
|
||||
|--------|--------|----------------|
|
||||
| Followers | 350-500 | 2-3x growth signals the approach is working |
|
||||
| Avg impressions per tweet | 800-2000 | Shows the algorithm is distributing your content |
|
||||
| Replies received per original tweet | 3-5 | People are engaging, not just scrolling past |
|
||||
| Mutual follows from target accounts | 5-10 | Your niche peers are noticing you |
|
||||
| Profile visits / week | 200+ | Your replies are driving curiosity |
|
||||
|
||||
**What to avoid in month 1**:
|
||||
- Posting 10 tweets/day hoping something sticks -- looks desperate, exhausts your ideas
|
||||
- Buying followers or using engagement pods -- Twitter's algorithm detects and penalizes this
|
||||
- Talking about yourself or your product -- earn attention through insight first
|
||||
- Getting discouraged by low numbers -- 180 to 400 followers in a month is strong growth
|
||||
|
||||
---
|
||||
|
||||
## Performance Metrics
|
||||
|
||||
### Key Metrics
|
||||
|
||||
Reference in new issue
Block a user