Twitter/X Influencer Strategy for Enovari

8. Twitter/X Algorithm & Growth Strategy

How the X Algorithm Works (2026 -- Grok-Powered Era)

In January 2026, X replaced its legacy recommendation system with a Grok-powered transformer model that reads every post and watches every video to match users with content. This represents the most significant algorithmic shift since Elon Musk's acquisition. The platform now processes 500 million daily tweets and makes 5 billion ranking decisions per day.

Key Ranking Factors (Updated for 2026 Grok Algorithm)

What the Algorithm Penalizes (Critical -- Avoid These)

• External links are now DEVASTATING for non-Premium accounts. Since March 2026, non-Premium accounts posting links receive near-zero median engagement -- their link posts are essentially invisible. Even for Premium accounts, external links reduce reach by 20-50%. Always put links in the first reply, or use the article feature for long-form content.
• Engagement bait phrases like "like if you agree," "retweet for reach," or "follow for more" can trigger suppression. The Grok algorithm has gotten better at detecting manufactured engagement.
• Rapid follower/unfollower behavior gets flagged and can result in shadowbanning or reduced reach.
• Duplicate or near-duplicate content across posts. The algorithm detects repurposed content and penalizes it.
• Over-posting without engagement -- More than 8-10 original posts per day without proportional engagement from others signals low-quality output.
• Hashtag overuse -- More than 2-3 per tweet looks spammy. For high-quality accounts in tech/AI, 0-1 hashtags is often optimal.
• Immediate post-and-leave behavior -- The algorithm tracks whether you stay active after posting. Post, then spend 30 minutes engaging with others' content. This is a key signal.
• Engagement pods and coordinated liking -- X has improved detection of artificial engagement patterns. Groups that systematically like each other's posts get penalized.
• Cross-posting from other platforms -- Auto-posted content from LinkedIn, Instagram, etc. often gets reduced reach. Native content performs better.
• Ratio'd posts (many replies, few likes) -- Posts where replies significantly outnumber likes signal controversial/low-quality content and get suppressed from the For You feed.
• Negative or combative tone -- New in 2026: the Grok algorithm actively down-ranks posts with hostile, aggressive, or combative language, even if they generate engagement.

The "For You" Feed vs "Following" Feed

Most users default to the "For You" algorithmic feed. Understanding how posts surface here is critical:

• For You is where growth happens. It shows content from accounts users don't follow, based on engagement signals and topic relevance. The Grok model now does end-to-end learning from user engagement history.
• Following only shows posts from followed accounts, chronologically. Less discovery potential.
• To appear in someone's "For You" feed, you need: (a) engagement from accounts they follow, (b) topical relevance to their interests (Grok understands semantics), or (c) very high engagement velocity on your post.
• This is why engaging with high-authority accounts in the AI space is the primary growth lever. Every like, reply, or RT from @simonw, @swyx, or @emollick potentially surfaces your content to hundreds of thousands of their followers.

X Premium Tiers and Their Benefits (2026)

Recommendation for Enovari: Start with Premium ($8/month). The algorithmic boost alone is worth it -- your replies to @alexalbert__ and @simonw's posts will appear significantly higher in the reply thread, giving more visibility. The link penalty for non-Premium accounts (near-zero reach since March 2026) makes Premium essentially mandatory for any product account. Upgrade to Premium+ ($40/month) once actively posting daily for access to X Pro analytics.

Best Posting Times

For AI/dev audience specifically: Tuesday-Thursday, 9-11 AM ET tends to peak. The developer audience is global, so also consider:

• 2-4 PM ET (evening in Europe)
• 6-8 PM ET (morning in APAC)
• Sunday evening (devs planning their week)

Pro tip: Post your highest-value content (threads, demos) on Tuesday-Thursday mornings. Save lighter content (observations, questions, community engagement) for off-peak times. Remember: with the 6-hour half-life on posts, timing matters more than ever.

Content Formats That Perform Best

1. Threads (Highest sustained reach)

• Start with a hook that creates curiosity or states a bold claim
• Keep each tweet in the thread to 1-2 sentences max
• Use visuals in at least 2-3 of the thread tweets
• End with a clear CTA (follow, bookmark, try the tool)
• Ideal length: 5-10 tweets
• Post the full thread at once (don't drip it)
• 2026 update: Thread engagement has slightly declined as the algorithm now also favors long-form single posts (using the 10K character limit for Premium users). Consider mixing both formats. Grok's semantic understanding means well-written long-form posts can perform comparably to threads.

2. Demo Videos/GIFs (Highest engagement per post)

• 15-45 seconds is the sweet spot
• Show the "magic moment" in the first 3 seconds
• No audio required (most people browse muted)
• Screen recordings with annotations/callouts work well
• Before/after format is highly effective for Enovari specifically
• Native video outperforms links to YouTube/Loom by 5-10x. Always upload directly to X.

3. Single Image + Short Copy

• Screenshot of the product doing something impressive
• "I built X that does Y" format
• Comparison images (without Enovari vs. with Enovari)
• Annotated screenshots with arrows and highlights

4. Hot Takes / Opinion Posts

• Bold statement about AI memory/context limitations
• "Unpopular opinion: [something about AI forgetfulness]"
• Engage with trending AI topics from the memory angle
• Contrarian takes get 3-5x more replies than neutral content
• 2026 note: Grok's tone monitoring means keep hot takes constructive. Hostile or combative framing will be down-ranked even if it generates engagement.

5. Long-Form Articles (Premium Feature)

• X now supports article-length posts for Premium subscribers
• These surface well in search and can rank on Google
• Good for technical deep-dives about MCP memory architecture
• Include inline images and code snippets
• Articles don't suffer the external link penalty

6. Engagement Posts (Use Strategically)

• "What's the most frustrating thing about AI assistants?" (leads to memory problem)
• Polls about AI pain points
• "Rate my stack" posts that include Enovari
• "What feature do you wish Claude had?" (answer: memory)

How to Build a Following from Zero

Phase 1: Foundation (Weeks 1-4) -- Target: 0 to 500 followers

• Optimize profile completely:
- Clear avatar (founder face is better than logo for engagement) - Banner showing the value prop: "Persistent memory for AI -- enovari.ai" - Bio: "Building @enovari_ai -- persistent memory for AI. Your AI forgets everything. We fix that. MCP server | Works across Claude, ChatGPT & more | https://enovari.ai" - Pinned tweet: best demo video or thread about the problem you solve
• Subscribe to X Premium -- the algorithm boost (4-8x reach) and the fix for the link penalty are mandatory for a product account in 2026
• Engage 30-60 minutes daily in AI conversations:
- Reply thoughtfully to 15-20 posts from people in the categories above - Provide genuine value in replies (insights, not just "great post!") - Quote-tweet interesting AI content with your memory perspective - Be consistently present in the reply sections of @alexalbert__, @simonw, @swyx - Remember: a reply is worth 27x a like, and a conversation is worth 150x a like
• Post 2-3 times daily:
- 1 value post (tip, insight, demo) - 1 engagement post (question, observation about AI limitations) - 1 community engagement (reply/QT to another builder)
• Follow the 130+ accounts in this document and turn on notifications for the CRITICAL ones

Phase 2: Growth (Weeks 5-12) -- Target: 500 to 2,000 followers

• Publish 1-2 threads per week on AI memory topics
• Share build-in-public updates with real numbers (users, API calls, interesting use cases)
• Engage with EVERY reply to your posts within 1 hour (conversations = 150x a like)
• Start tagging relevant accounts when sharing relevant content (max 2 tags per post)
• Cross-promote from other channels (newsletter, Discord, LinkedIn, Reddit)
• Create a "memory moment" series -- before/after screenshots of AI with and without memory

Phase 3: Acceleration (Weeks 13-24) -- Target: 2,000 to 10,000 followers

• Host or join Twitter Spaces on AI/MCP topics weekly
• Collaborate with mid-tier accounts (joint threads, mutual shoutouts)
• Respond to every major AI announcement from the memory angle ("Great that Claude got X, but it still forgets everything between sessions...")
• Engage with breaking AI news fast -- first-mover advantage in replies
• Launch campaigns timed to product milestones and Anthropic announcements

Thread Strategies That Go Viral

The "Problem-Solution" Thread

Tweet 1: "Your AI assistant has amnesia.
Every conversation starts from zero. Every. Single. Time.
Here's why that's about to change: [thread]"
Tweet 2: The problem: [AI has no persistent memory -- concrete example with screenshot]
Tweet 3: Why it matters: [lost productivity stats, repeated context, broken workflows]
Tweet 4: What we built: [Enovari -- one sentence + screenshot]
Tweet 5: How it works: [MCP server explanation -- keep it simple]
Tweet 6: Demo: [30-second video or GIF]
Tweet 7: What users are saying: [testimonial or use case]
Tweet 8: "Try it free at enovari.ai
If this was useful, bookmark this thread and follow @enovari_ai for more."

The "I Analyzed X" Thread

Tweet 1: "I analyzed 100 conversations between developers and AI assistants.
The #1 frustration? Having to repeat yourself.
Here's what I found: [thread]"
[5-7 tweets with data points, examples, insights]
[Final tweet: "This is why we built Enovari..."]

The "Build in Public" Update Thread

Tweet 1: "Month [X] building Enovari -- AI memory that persists across sessions.
[Milestone number] users. [Growth %]. Here's exactly what happened: [thread]"
[Journey, challenges, metrics, what worked, what didn't]

The "Comparison/Before-After" Thread

Tweet 1: "I used Claude for a week WITHOUT persistent memory vs WITH @enovari_ai memory.
The difference was genuinely shocking.
Here's what happened: [thread]"
[Side-by-side screenshots, specific examples, time savings]

The "Hot Take" Thread

Tweet 1: "AI without memory is like a doctor who forgets every patient after each visit.
Controversial? Let me explain why memory is the most underrated capability in AI right now: [thread]"

Crafting Tweets That Get Shared

Hook Formulas That Work for Enovari

• The Bold Claim: "AI memory will be bigger than RAG. Here's why."
• The Curiosity Gap: "I found the #1 reason AI assistants fail at long projects. It's not what you think."
• The Contrarian: "Everyone is building AI agents. Almost nobody is solving the actual problem: memory."
• The Number: "I saved 3 hours per week by giving my AI a memory. Here's how."
• The Confession: "I was skeptical about AI memory. Then I tried it for a month."
• The Tutorial: "How to give Claude persistent memory in 5 minutes (free):"
• The Observation: "The best AI workflows all have one thing in common: they remember context."
• The Frustration: "Just re-explained my entire project setup to Claude for the 47th time. There has to be a better way."
• The Prediction: "In 12 months, AI without persistent memory will feel like a phone without contacts."
• The Question: "If your AI could remember everything you've ever told it, what would you build?"

Tweet Structure Best Practices

• First line is everything -- it determines whether people click "Show more"
• Use line breaks generously for readability (not walls of text)
• End with a question or CTA to drive replies (replies = 27x a like)
• Put links in the first reply, never in the main tweet body (especially critical post-March 2026)
• Use 0-2 hashtags max; often zero is best for high-quality accounts
• Tag 1-2 relevant accounts only when it genuinely adds value
• Use em-dash and colon for punchy pacing
• Numbers and specifics always outperform vague claims
• Constructive tone -- Grok rewards positivity and penalizes hostility

Hashtags and Communities

High-Value Hashtags (Use Selectively -- max 1-2 per tweet)

Communities / Conversations to Embed In

• MCP Community -- The most directly relevant. Follow everyone who tweets about MCP servers. Engage in every thread. This is your home base. The MCP ecosystem now has 97M monthly SDK downloads and industry-wide adoption.
• AI Engineering (swyx's community) -- Technical AI builders who need memory solutions. Attend Latent Space events.
• Claude Code / Claude users -- People actively using Claude who feel the pain of no persistent memory
• LangChain / LlamaIndex ecosystem -- Framework users building agents who need memory
• Build in Public -- Indie hackers who share their tool stacks
• Cursor / Windsurf / AI IDE users -- Power users who want persistent context
• AI Twitter (broad) -- Engage with trending AI topics, always from the memory angle
• Context Engineering -- Emerging 2026 trend. Allie K. Miller and others are framing memory as part of "context engineering." Align Enovari's positioning with this trending terminology.

Twitter Spaces Opportunities

Spaces to Join/Host

How to Get Invited to Spaces

• Engage consistently with hosts' content for 2-3 weeks first
• Provide insightful replies on their topics (be memorable)
• DM with a specific topic you can speak on + your credentials
• Host your own smaller Spaces first (builds credibility and shows you can hold an audience)
• Offer to demo Enovari live during a Space (live demos are compelling content)

---

9. Engagement Examples & Reply Templates

The most valuable growth activity on X is replying to high-authority accounts with genuinely insightful content. Here are specific examples of good replies for key target accounts.

Replying to @alexalbert__ (MCP Creator -- CRITICAL)

When he tweets about MCP ecosystem growth (e.g., his March 2026 MCP timeline thread):

The memory layer is the missing piece in most MCP setups. Servers handle tools and data retrieval well, but session-to-session memory is what makes AI actually feel like a collaborator instead of a stranger you meet every day. We're seeing this firsthand building Enovari -- the most requested feature from MCP users is "remember what I told you yesterday."

When he showcases a new MCP server:

This is great. One thing I've noticed across MCP server demos -- they solve the "what can AI access" problem but not the "what does AI remember" problem. These two are complementary. [Server name] + persistent memory would be a killer combo.

What NOT to reply:

"Check out Enovari! We built an MCP memory server at enovari.ai" -- This is a cold pitch disguised as a reply. It will be ignored or muted.

Replying to @simonw (LLM Tools Reviewer -- CRITICAL)

When he's experimenting with Claude or LLM tools:

Interesting approach. One challenge I keep running into with [topic he's discussing] is that each session starts from scratch. I've been working on persistent memory via MCP that carries context across sessions -- changes the workflow significantly when the AI remembers your project structure, preferences, and past decisions.

When he reviews an MCP tool:

Nice writeup. Have you tried pairing this with a memory layer? The combination of [tool he reviewed] for [function] plus persistent session memory makes for a much richer experience. The AI context window is temporary by design -- external memory makes it effectively infinite for the stuff that matters.

What NOT to reply:

"You should try Enovari next! enovari.ai" -- He discovers tools himself. Let him find it. Or DM with docs and let him evaluate on his own terms.

Replying to @swyx (AI Engineering -- CRITICAL)

When he discusses AI infrastructure:

This is the right framing. Memory is infrastructure, not a feature. Just like databases became a standard layer in web apps, persistent memory will become a standard layer in AI apps. The question isn't "should AI have memory" -- it's "who owns the memory layer and how portable is it?"

When he discusses MCP:

The MCP ecosystem is solving the integration problem beautifully. The next frontier is state persistence -- MCP servers that don't just connect AI to tools but help AI build cumulative knowledge over time. Memory is the bridge between "tool use" and "genuine AI assistance."

Replying to @alliekmiller (Context Engineering -- HIGH, new)

When she posts about context engineering or memory as a 2026 trend:

Spot on. Context engineering is the next battleground. Models are getting smarter but the context management layer is still primitive. The question is: should context be ephemeral (rebuilt every session) or persistent (accumulated over time)? We're betting on persistent -- built an MCP server for exactly this.

When she posts about enterprise AI adoption:

This is the hidden bottleneck in enterprise AI. Teams start enthusiastic, then hit "Groundhog Day" -- every conversation starts over. The context setup cost eventually exceeds the productivity gain. Persistent memory changes the economics completely.

Replying to @emollick (AI in Education/Work -- HIGH)

When he posts about AI limitations or adoption challenges:

This is such an underappreciated point. I've been tracking how AI adoption stalls in teams, and a huge factor is the "Groundhog Day" problem -- every interaction starts fresh. People stop using AI for complex projects because the setup cost of re-explaining context exceeds the time saved. Persistent memory changes the cost-benefit calculation entirely.

Replying to @karpathy (AI Education -- Awareness)

When he posts about LLM architecture or context windows (or about his Loop agent):

Great explanation. One thing I keep thinking about: the context window is temporary RAM, but where's the persistent storage? Current LLMs have incredibly sophisticated short-term processing but no long-term memory architecture. External memory servers (via MCP for example) are an interesting bridge until the models themselves solve this. Your Loop agent would be even more powerful if it could remember findings across sessions.

Replying to AI News Accounts (When They Share AI Limitations)

General template for AI tool curators posting about frustrations:

This is the #1 pain point I hear from AI power users. I tracked my own AI usage for a month and spent 22% of my time re-explaining context that the AI should already know. The models keep getting smarter but they still have amnesia. Persistent memory is the fix.

Replying to @dsp_ or @jspahrsummers (MCP Creators -- HIGH, new)

When they discuss MCP protocol design or architecture:

One area I'd love to see formalized in MCP is memory as a first-class primitive. Right now tools and resources are well-defined, but session state and persistent user context feel like they should be part of the protocol's vocabulary. We're solving this at the server level with Enovari, but I wonder if there's a protocol-level pattern that could emerge.

Quote-Tweet Templates

When someone complains about AI forgetting context:

This. The most powerful AI models in the world still can't remember your name between conversations.

> We're building the fix: persistent memory for AI via MCP.

> [link in reply, not in QT]

When a major AI announcement happens (new model, new feature):

[Model/feature] is impressive. But here's the thing nobody's talking about:

> It still forgets everything the moment you close the tab.

> Context windows got bigger. But memory? Still zero.

When someone shares their AI workflow:

Solid workflow. One addition that transforms it: persistent memory.

> Instead of re-establishing context every session, the AI just... knows your project, your preferences, your decisions.

> Game changer for [specific use case they mentioned].

When @alliekmiller or others mention context engineering:

Context engineering is the 2026 trend everyone should be watching.

> But most solutions focus on building context each session. What about remembering it?

> Persistent memory is the missing piece that makes context engineering actually practical.

---

10. Outreach Playbook

The Engagement-First Method (Recommended)

DO NOT cold-pitch influencers. The success rate for cold DM pitches is under 2%. Instead, follow this proven sequence:

Week 1-2: Warm Up

• Follow the account
• Like 3-5 of their posts per week
• Leave 2-3 thoughtful, substantive replies (NO mention of Enovari)
• Quote-tweet one of their posts with genuine insight

Week 3: Add Value

• Reply to a relevant post with a helpful insight about AI memory or context
• Share their content with added commentary that shows you understand their work
• If they have a newsletter, reply to an edition with something useful

Week 4: Soft Introduction

DM template:

Hey [Name], been following your work on [specific topic] -- really enjoyed your [specific post/thread].

> I'm building something in the AI memory space that your audience might find interesting. It's an MCP server that gives AI persistent memory across sessions and platforms.

> Would love to get your take on it if you have a minute. No pressure at all -- happy to just keep learning from your content either way.

Week 5+: Follow Up

• If positive: Share a 30-second demo video, offer early access, provide ready-to-share content
• If no response: Continue engaging with their content for 2-3 more weeks. Try one more message with a different angle.
• If negative: Thank them sincerely and move on. Don't burn the bridge.

For Newsletter Authors (Email Pitch)

Subject: AI Memory Platform -- Potential Feature for [Newsletter Name]

> Hi [Name],

> I'm [Your Name], building Enovari (enovari.ai) -- a persistent memory layer for AI assistants.

> The problem: Every AI conversation starts from zero. Users repeat context, lose preferences, and can't build continuity across sessions.

> What we built: An MCP server that gives any AI platform persistent, structured memory. Works with Claude, ChatGPT, and other platforms. Setup takes 5 minutes.

> Why your readers care: [Specific angle tailored to their audience -- e.g., "Your readers are power users who have felt this pain" or "Your readers build AI tools and this is infrastructure they can use"]

> Quick demo (30 seconds): [Loom/video link]

> Happy to provide early access for you or your team to try it out. Also happy to share any data points or user stories that would make a good feature.

> Best,

[Name]

For Tool Curators (Make It Easy to Share)

Provide a complete "ready-to-share" package via DM:

Hey [Name], thought your audience might like this:
Enovari -- Persistent Memory for AI
Gives AI assistants memory that persists across sessions
Works via MCP (Model Context Protocol)
Compatible with Claude, ChatGPT, and more
Setup in under 5 minutes
Free tier available
Here's a 20-second demo: [GIF or video link]
Link: enovari.ai
I can also send you a suggested tweet if that's helpful -- just want to make it as easy as possible for you.

For MCP Community (Be Technical and Direct)

The MCP community respects technical depth. Be straightforward:

We built an MCP server for persistent AI memory: enovari.ai

> It stores structured memories that persist across sessions and work across different AI platforms. The architecture uses [brief technical detail].

> Here's the setup: [docs link]

Here's it running with Claude: [demo]

> Would love your feedback -- what memory patterns are you seeing in your MCP usage?

---

11. Pitch Templates by Category

For AI Tool Curators

"Enovari gives AI persistent memory via MCP. Your AI forgets everything between conversations -- Enovari fixes that. Works across Claude, ChatGPT, and more. It's like giving your AI a brain that doesn't reset. Demo: [link]"

For AI Thought Leaders

Don't pitch directly. Instead, engage with their content from the "AI memory" angle. When they post about AI limitations, reply with insights about how persistent memory changes the equation. Let them discover Enovari through your consistent, valuable presence in their replies.

For Developer Advocates

"We built an MCP server for AI memory. It gives persistent, structured memory to any AI assistant. The developer experience is dead simple -- install, configure, and your AI remembers everything across sessions. Open to feedback. Docs: [link]"

For Indie Hackers

"Building in public: Enovari -- AI memory that persists across sessions. [X] users and growing. The problem we solve: every AI conversation starts from zero. Our solution: an MCP server that gives any AI persistent memory. Would love feedback from fellow builders."

For Newsletter Authors

"Enovari (enovari.ai) is a new MCP server that gives AI assistants persistent memory across sessions and platforms. The pitch: 'Your AI forgets everything. Enovari remembers.' Setup in 5 minutes, free tier available. Demo: [link]"

For MCP/Claude Community

"We built an MCP server for persistent AI memory. It stores structured memories that persist across sessions and work across Claude, ChatGPT, and other platforms. The MCP integration means it works wherever MCP is supported -- and with MCP now supported by every major AI provider, that's everywhere. Try it: enovari.ai"

For YouTube Creators

"Would you be interested in reviewing Enovari for a video? It's an MCP server that gives AI persistent memory -- before/after demos are visually compelling. I can provide a full walkthrough, early access, and any assets you need. Here's a 30-second demo: [link]"

---

12. What NOT to Do -- Common Mistakes & Anti-Patterns

Cold Pitch Sins

• The drive-by DM: Sending "Check out my product: [link]" to someone you've never interacted with. This gets you muted immediately. Success rate: <1%.
• The mass-copy DM: Sending the exact same DM to 50 people in one day. X detects this and may restrict your account. Influencers talk to each other -- they'll notice identical messages.
• The reply-as-ad: Replying to an influencer's tweet with "Speaking of AI tools, check out Enovari at enovari.ai!" This is transparently self-promotional and will get you blocked.
• The fake compliment-then-pitch: "Love your content! BTW I built this tool..." Everyone sees through this. The compliment means nothing if it's followed by an ask.
• The unsolicited tag: Tagging @alexalbert__ or @simonw in your product launch tweet when they don't know you. Tagging people who haven't opted in feels aggressive.

Engagement Mistakes

• "Great post!" replies: These are invisible to the algorithm and to the person. Every reply should add information, perspective, or a question. If you don't have something substantive to say, don't reply.
• Replying to everything: If you reply to every single tweet from one account, you look like a bot or a stalker. 2-3 substantive replies per week per account is the right cadence.
• Automated engagement tools: Bots that auto-like, auto-reply, or auto-follow get detected and penalized. X's detection has improved significantly. The risk to account health isn't worth it.
• Engagement pods / pods of mutual likers: Groups that coordinate to like each other's content. X detects the patterns and penalizes everyone in the pod. Build genuine engagement instead.
• Posting only about your product: If 80%+ of your tweets are about Enovari, you'll build a promotional account, not a thought-leader account. Aim for 30% product, 70% value/insight/community.

Content Mistakes

• Links in the main tweet: Already covered in the algorithm section, but worth repeating: external links in the tweet body can reduce reach to near-zero for non-Premium accounts (March 2026 change). Always put links in the first reply.
• Too many hashtags: More than 2 hashtags makes a tweet look spammy. For the AI/tech audience, 0-1 is usually optimal.
• Jargon overload: "Our MCP server leverages distributed persistent memory architecture with vector-indexed semantic retrieval across polyglot AI platform integrations" -- nobody shares this. Keep it human: "Your AI forgets everything. Enovari remembers."
• Inconsistent posting: Posting 10 times one day, then disappearing for 2 weeks, then posting 5 times. The algorithm rewards consistency. 2-3 posts per day, every day, is far better than bursts.
• Ignoring replies to your posts: If someone replies to you and you don't respond within a few hours, you're training the algorithm that your content doesn't generate conversation. Conversations are worth 150x a like -- reply to everything, especially early in your growth.

Relationship Mistakes

• Burning bridges: If an influencer says no or doesn't respond, do not: (a) complain publicly, (b) badmouth them, (c) keep DMing. Move on gracefully. The AI community is small.
• Expecting immediate results: The engagement-first method takes 4-6 weeks per account. There are no shortcuts. Influencers who share your product after a 30-second cold pitch are the exception, not the rule.
• Treating influencers as marketing channels: They are people. They have their own goals, interests, and audiences to protect. Approach them as peers and potential collaborators, not as distribution channels.
• Overpromising: Don't tell an influencer "this will change AI forever" if your product is a solid MCP memory server. Undersell and overdeliver. Hyperbole erodes trust.
• Forgetting to follow up: If an influencer says "interesting, send me more info" and you don't follow up within 24 hours, you've wasted the entire warm-up investment. Have your demo video, docs link, and one-page overview ready to send instantly.

Account Health Mistakes

• Following/unfollowing rapidly: Following 500 people in a day then unfollowing 400 the next day. X flags this behavior and may restrict your account.
• Tweeting from automation tools only: Never signing into X directly and only posting via Typefully/Buffer/etc. X may reduce reach for accounts that never engage natively on the platform.
• Multiple accounts for cross-promotion: Running 5 accounts that all retweet each other. X detects coordinated inauthentic behavior and can suspend all accounts.
• Engaging in political/controversial debates: The AI community has strong opinions on AI safety, ethics, and politics. As a product account, stay neutral on anything not directly related to AI memory. Taking sides on unrelated controversies is pure downside risk. Additionally, Grok's tone monitoring means combative posts get down-ranked even if they generate engagement.

---

13. Anti-Spam Playbook -- Staying on the Right Side of the Line

This section covers the nuance between effective outreach and spam. In 2026, X's Grok-powered algorithm is aggressive about detecting and penalizing spam-like behavior, and the -369x report penalty means even a few reports can destroy your account's reach.

The Spam Spectrum

Specific Daily Limits to Avoid Triggering Spam Detection

The "Would I Mute This?" Test

Before sending any DM, reply, or tag, ask: "If someone sent this to me, would I mute them?"

• If yes: rewrite or don't send.
• If maybe: rewrite to add more genuine value before any product mention.
• If no: send it.

Cooling Off Protocols

If you notice declining engagement or suspect shadowbanning:

• Stop all product-related posting for 48 hours
• Engage with others' content only (genuine replies, no product mentions) for a full week
• Reduce posting frequency to 1-2 posts per day
• Check if your posts appear in X search (search your handle) -- if not, you may be restricted
• Gradually resume normal cadence after 1-2 weeks

If you receive a warning or restriction from X:

• Immediately stop all outreach DMs
• Remove any automated tools or scheduled posts
• Post only organic, non-promotional content for 2-4 weeks
• Do not create alternative accounts (coordinated behavior = suspension)
• Appeal via X's support if the restriction seems incorrect

The Relationship Velocity Rule

For each influencer, follow this maximum escalation speed:

The cardinal rule: if someone hasn't responded to two touchpoints (a DM and a follow-up), stop. Wait at least 4 weeks before any further direct contact. Continue engaging with their public content at a low cadence (1-2 likes per week).

Safe vs. Unsafe Self-Promotion Patterns

Safe:

• "We just shipped [specific feature] in Enovari. Here's a 20-second demo:" (native video, link in reply)
• Replying to a thread about AI memory problems with your experience (without linking)
• Build-in-public updates with genuine metrics
• Responding when someone asks "is there an MCP server for memory?"

Unsafe:

• "Great thread! BTW we solve this: enovari.ai" on any remotely related post
• Replying to @alexalbert__ with "Check out our MCP server!" on a post about MCP
• Tagging 5 influencers in a product announcement they didn't ask to be part of
• Posting the same demo video 3 times in one week
• DMing the same pitch to 10 people in one day

---

14. Priority Action Plan

Immediate Actions (This Week)

• Set up / optimize the Enovari Twitter/X account:
- Founder's face as avatar (higher engagement than logos) - Banner: "Your AI forgets everything. Enovari remembers." + enovari.ai - Compelling bio with keywords (MCP, AI memory, Claude) - Pin your best demo video or product thread - Subscribe to X Premium ($8/month) for algorithm boost -- this is now mandatory for product accounts given the March 2026 link penalty for non-Premium
• Follow all CRITICAL and HIGH priority accounts (approximately 40-50 accounts)
- Turn on notifications for CRITICAL accounts (@alexalbert__, @simonw, @swyx)
• Begin daily engagement routine:
- 30 min morning: engage with AI tool curators and thought leaders - 15 min midday: engage with MCP/Claude community threads - 15 min evening: engage with indie hackers/builders
• Create first product demo thread (5-8 tweets with screenshots/GIF showing before/after)
• Submit to MCP directories:
- PulseMCP (pulsemcp.com) - MCP Market (mcpmarket.com) - Awesome MCP Servers on GitHub - Glama.ai marketplace

Week 1-2 Actions

• Engage warmly with MCP community accounts (highest conversion probability):
- Reply to @alexalbert__'s MCP threads with genuine insight - Engage with @simonw's LLM tool content - Find and reply to every tweet containing "MCP server" - Engage with @dsp_ and @jspahrsummers's MCP protocol discussions
• Submit to AI tool directories that have Twitter presence:
- FutureTools.io (submit at website + DM @mreflow) - There's An AI For That (theresanaiforthat.com) - BetaList (betalist.com + DM @marckohlbrugge)
• Publish 2 threads:
- "Why AI memory changes everything" (thought leadership, problem-focused) - "How to give Claude persistent memory in 5 minutes" (tutorial, actionable)

Week 3-4 Actions

• Begin newsletter outreach (email pitches to):
- The Rundown AI (@rowancheung) - Ben's Bites (@bentossell) - Superhuman AI (@heykahn) - Latent Space (@swyx) - Alpha Signal (@AlphaSignalAI)
• Engage with 5 indie hackers who are building AI products (warm up for eventual pitch)
• Host first Twitter Space or join an existing one on AI tools/MCP

Month 2 Actions

• Scale outreach to MED priority accounts using the engagement-first method
• Establish weekly content cadence:
- Monday: Product update / build in public metrics - Tuesday: Thread on AI memory topic (educational) - Wednesday: Engagement day (replies, QTs, community building) - Thursday: Demo / tutorial content (how-to) - Friday: Community highlight / retweet interesting AI content
• Track and optimize based on X Analytics (requires Premium+ for X Pro, or use native analytics)
• Engage with "context engineering" trend -- reply to @alliekmiller and others framing memory as part of this trend

Month 3 Actions

• Reach out to LOW priority accounts as brand credibility grows
• Launch a "Memory Moments" content series -- weekly before/after comparisons
• Pitch podcast appearances:
- Latent Space (@swyx) - Cognitive Revolution (@labenz) - TWIML (@samcharrington) - The Next Wave (@nathanlands / HubSpot)
• Run first Twitter-exclusive campaign (early access to new features for Twitter followers)
• Pitch YouTube creators for tool review videos (highest long-tail value):
- Matt Wolfe (@mreflow) - AI Jason (@AIJasonZ) - All About AI (@AllAboutAI)

---

15. Verification Checklist & Live Research Guide

Before Reaching Out to Any Account

For each account in this document, verify the following on X directly:

• [ ] Handle is still active (not suspended, deactivated, or renamed)
• [ ] Follower count is approximately as listed (may have grown or shrunk)
• [ ] They are still posting regularly about AI/relevant topics (check last 2 weeks)
• [ ] Their DMs appear open (or find email in bio/pinned/links)
• [ ] They still share product recommendations (check last 10-20 posts)
• [ ] Their engagement ratio is healthy (not just followers, but replies/likes on posts)
• [ ] Their bio/links are current (companies change, roles change)

Handles Corrected in This Revision

The following entries had handle or data inconsistencies that have been corrected:

Live Search Strategies to Find More Accounts

Run these searches on X to find accounts not in this document:

Recommended Tracking System

Create a spreadsheet (Notion, Airtable, or Google Sheets) with these columns:

Tier	Monthly Cost	Key Benefits for Growth
Basic	$3/month	Edit tweets, longer posts, reduced ads, bookmark folders. No blue checkmark. No monetization.
Premium	$8/month	Blue checkmark, 50% fewer ads, 10K character posts, article publishing, ad revenue share, meaningful algorithmic boost (4-8x reach over non-Premium)
Premium+	$40/month	Nearly all ads removed, highest Grok AI limits, Articles, Radar Search, largest algorithmic visibility boost, X Pro access (analytics dashboard moved behind Premium+ paywall in March 2026)
Day	Best Times (ET)	Why
Monday	8-10 AM, 12-1 PM	Morning productivity mindset
Tuesday	8-10 AM, 12-1 PM	Peak B2B engagement day
Wednesday	9-11 AM, 5-6 PM	Midweek check-ins, second wind
Thursday	8-10 AM, 12-1 PM	Pre-weekend planning
Friday	9-11 AM	Last workday push; engagement drops in afternoon
Saturday	10 AM - 12 PM	Weekend builders, lighter competition = more relative visibility
Sunday	12-3 PM	Builder community catches up on what they missed
Hashtag	Audience	When to Use
#AI	Broad AI audience	Major announcements only
#AItools	AI tool seekers	Tool launches, demos
#MCP	MCP community	MCP-related content -- high signal, low noise
#Claude	Claude users	Claude-specific features
#BuildInPublic	Indie hackers	Progress updates, metrics shares
#DevTools	Developers	Technical posts
#LLM	LLM practitioners	Technical content
#AIagents	Agent builders	Agent memory content
#Anthropic	Anthropic community	Claude/MCP ecosystem content
#ContextEngineering	Emerging 2026 trend	Memory-as-infrastructure content (coined by Allie K. Miller and others)
Topic	Frequency	Target Co-hosts	Why It Works
"MCP Server Showcase"	Monthly	@alexalbert__, MCP devs	Show Enovari as MCP server to exact audience
"AI Tools of the Week"	Weekly	Tool curators like @rowancheung	Get featured as tool pick
"Building AI Products"	Bi-weekly	@tdinh_me, @marc_louvion, indie hackers	Share Enovari journey to supportive audience
"The Future of AI Memory"	Monthly	@swyx, @simonw, AI engineers	Position as THE memory expert
"Claude Power Users"	Weekly	Claude community members	Show how memory transforms Claude workflows
"AI Developer Tools"	Bi-weekly	@hwchase17, @jerryjliu0	Developer angle, integration opportunities
"Context Engineering"	Monthly	@alliekmiller, AI product builders	Align with 2026's hottest AI trend
Behavior	How It Looks	Result
Clearly Spam	Same DM to 50 accounts in one day. "Check out enovari.ai" replies on random threads. Auto-replies. Fake engagement.	Account restriction or suspension. -369x report penalty compounds. Permanent authority score damage.
Borderline Spam	Mentioning Enovari in every reply. Tagging influencers in every product tweet. Replying to every post from one account. Posting 10+ times daily with product links.	Muted by influencers. Reduced algorithmic reach. Reputation damage in the small AI community.
Acceptable Promotion	Occasional product mentions in context-relevant threads. Organic responses when memory/context is the topic. Build-in-public updates about Enovari. Responding when asked.	Normal growth. Respect from community. Sustainable engagement.
Ideal Engagement	70% value-add content (insights, education, community). 30% product-related (demos, updates, user stories). Mentions Enovari only when directly relevant. Engages as a person, not a product.	Rapid growth. Influencer attention. Organic amplification. Authority building.
Action	Safe Daily Limit	Warning Zone	Danger Zone
New follows	20-30	50-100	200+
DMs to new contacts	3-5	10-15	20+
Replies mentioning Enovari	1-2	5-10	10+
Identical or near-identical replies	0	1-2	3+
Account tags per day	2-3	5-10	10+
Link posts (even in replies)	3-5	8-10	15+
Original posts	3-5	8-10	15+
Week	Maximum Contact Level
Week 1	Follow + like 3-5 posts (NO replies mentioning Enovari)
Week 2	2-3 thoughtful replies (NO product mentions)
Week 3	1 reply that naturally references your work in the memory space
Week 4	1 DM (only if they've noticed you / replied to you)
Week 5+	Follow their lead on cadence. If they engage, match their energy. If silence, back off to Week 2 cadence.
Previous Entry	Issue	Correction
@oaboratory (Elvis Saravia)	Wrong handle. His actual handle is @omarsar0.	Corrected to @omarsar0. Verified via GitHub (github.com/omarsar) and Linktree.
@maboratory (Matt Wolfe)	Wrong handle. His actual handle is @mreflow.	Corrected to @mreflow. He also has @FutureToolsio as a second account. YouTube at 900K+ subs (was listed as 400K).
@balaboratory (Balaji)	Wrong handle/URL. His handle is @balajis.	Corrected URL to x.com/balajis.
@heyzain (Zain Kahn)	Wrong handle. His actual handle is @heykahn.	Corrected to @heykahn. Superhuman AI newsletter at 1.5M+ subscribers (was 500K).
@philschmid (Philipp Schmid)	Wrong handle (missing underscore) and wrong employer.	Corrected to @_philschmid. Now at Google DeepMind (left Hugging Face late 2025).
@reaboratory (Replit)	Wrong handle.	Corrected to @Replit. Added CEO @amasad as separate entry.
@windsurf_ai	Wrong handle. Windsurf's X handle is @codeiumdev.	Corrected. Acquired by Cognition AI Dec 2025.
@ClaudeAI	Was listed as "Unknown if active."	Confirmed as @claudeai, launched July 2025, active.
@marclou (Marc Lou)	Old handle. Primary active handle is now @marc_louvion.	Updated. @marclou may still redirect.
@haboratory_ai (#18)	Could not be verified as a real account.	Removed from document.
Aravind Srinivas duplicate	Was listed as both #16 and #37.	Consolidated into single entry (#16).
@nutlope (Hassan)	Listed as "Vercel engineer."	Updated: now leads DevRel at Together AI.
@goodside (Riley Goodside)	Listed as "Scale AI staff."	Updated: now Staff Prompt Engineer at Google DeepMind.
@bentossell (Ben Tossell)	Listed only as Ben's Bites.	Updated: now also Head of DevRel at Factory.
@alliekmiller	Follower count listed as 100K+.	Corrected to 2M+ (she's the #1 most-followed voice in AI business).
@labenz (Nathan Labenz)	Follower count listed as 60K+.	Corrected to ~17K. Podcast audience is the main reach.
Search	Purpose
"MCP server" min_faves:20	Find popular MCP server tweets
"MCP" "memory" min_faves:10	Find MCP + memory discussions
"Claude" "memory" "MCP" min_faves:10	Find Claude memory discussions
"AI tool" "thread" min_faves:100	Find active AI tool curators
"AI memory" min_faves:20	Find people discussing AI memory
"persistent memory" "AI" min_faves:10	Direct competitors or adjacent builders
"build in public" "AI" min_faves:50	Find AI indie hackers
"launched" "MCP" min_faves:10	Find people launching MCP products
"Claude Code" "memory"	Find Claude Code users discussing memory
"MCP server" "built" OR "building"	Active MCP builders
"Enovari" OR "enovari.ai"	Track organic mentions of Enovari
"context engineering" min_faves:20	Find people discussing the 2026 context engineering trend
"agent memory" min_faves:10	Agent memory infrastructure discussions
"Agentic AI Foundation" "MCP"	Industry-level MCP discussions
Column	Purpose
Handle	Current verified handle
Name	Real name
Category	From this doc (1-7)
Followers	Current count
Last Active	Date of last post
DMs Open?	Yes/No
Email Found?	Y/N + email
Engagement Phase	Warming/Ready/Pitched/Responded/Partner
Last Contact Date	Date of last outreach
Notes	Specific interests, what they respond to
Status	Active/Declined/No Response/Engaged

---