500 followers in 30 days sounds like nothing until you remember I started from literal zero on June 7, 2025. New account, no imported audience, no cross-promo deals. If you're trying to grow Twitter followers from zero, my situation was as clean-slate as it gets - my previous account @deeflect got banned and I wasn't going to beg for it back. So I rebuilt. And figuring out how I grew from 0 to 500 Twitter followers taught me more about the platform than the previous years I'd spent on it.
I'm at ~660 now. Not viral. Not a "Twitter guru." But I have a real, engaged audience of AI/dev/startup people who actually interact. That's worth more to me than 50K ghost followers.
Here's what I learned rebuilding from scratch.
The 3 routes to grow Twitter followers from zero (pick wisely)
Before I talk tactics, the mental model matters.
There are exactly three routes to growing on X. Route A: be funny and shitpost. Route B: drop insights and add value. Route C: rage bait and controversy. That's it. Every account you've seen blow up used at least one of these. Most big accounts you think are doing something special are just doing one of these three things really well.
Route C works. It works fast. It also fills your replies with people who are wrong on the internet as a personality trait. I'm not interested in managing that energy, so I mostly stayed off it.
Route A gets you shared. Funny spreads. But pure shitposting doesn't build authority - it builds a following that sees you as entertainment, not expertise. The second you post something serious, engagement tanks because you trained them to expect jokes.
Route B gets you saved. Value posts rack up bookmarks and follows from people who want to learn something. But pure value content is too easy to scroll past without engaging. It doesn't spread.
The actual sweet spot - and this is the insight that changed my numbers - is A+B together. A funny observation that also teaches something. A self-deprecating building-in-public moment that contains a real insight. The format makes people laugh, the substance makes them follow.
"Shipped a feature that took 3 days to build. My AI agent replaced it in 47 minutes. I'm choosing to be excited about this."
That's funny and it says something real about where AI tooling is going. That kind of post travels.
What actually grew my account from zero
Replies on big accounts - this was 80% of it
Not "great post!" Not "I agree!" Not the kind of thing that reads like a bot (more on that shortly).
Actual additions. Real takes. Occasionally a funny riff on what they said. When someone with 50K followers posts something about AI agents being overhyped, I'm not validating their hot take - I'm adding specificity. "The hype is real, the production readiness isn't. I've been running 14 scheduled agent jobs daily for 3 months and the failure rate on complex multi-step tasks is still too high for me to hand off anything important unsupervised."
That kind of reply does a few things. It shows up in the timeline of everyone following that account. It positions me as someone who's actually done the thing. And if it's good enough, the original poster might like it or reply - which extends the reach further.
The key is being genuinely useful or genuinely funny. Vague agreement is invisible. Disagreement with substance is visible. Addition is visible. The metric I use: would someone screenshot this reply? If not, it's probably not pulling any weight.
I probably wrote 15-20 quality replies a day for the first two weeks. That's the unsexy part nobody talks about. It's not posting content - it's showing up in other people's conversations consistently.
Opinionated takes from daily tool usage
I use AI tools every day. Cursor, Claude Code, various agent frameworks, LLM APIs. I have real opinions about what's good and what's broken. Those opinions performed well because they're specific.
"Cursor's composer is great until your project hits ~150 files, then the context quality drops noticeably" is a post. "AI coding tools are amazing" is noise.
The what happened when I stopped posting gives the full background, but the short version: 10+ years in design, deep fintech engineering experience, now building multi-agent systems full-time. That's a specific lens. When I post about AI tools, I'm not regurgitating documentation - I'm reporting from actual usage at a level most people don't have.
Specificity is unfakeable credibility. Anyone can say "this tool is good." Not everyone has run 40K tokens per day through their own agent system and tracked the monthly API cost down to $40.
Building in public - but only the interesting parts
"Day 47 of my building in public journey" is dead. Nobody wants that. The journey framing assumes your audience cares about you, and they don't - not yet. They care about what you're doing and what they can learn from it.
What works instead: the specific moment that was actually interesting.
Not "working on my AI project today." Instead: "Realized my RAG pipeline was failing not because of the model but because of my chunking strategy. Three days of debugging to find a paragraph break problem. Love this for me."
Building in public posts that perform have an insight embedded in them. The vulnerability is a delivery mechanism for the learning, not the point itself. When the vulnerability is the point, it reads as attention-seeking. When the learning is the point and the vulnerability makes it human, it actually helps people.
Timing matters more than I expected
I post when US tech Twitter is active. That's roughly 9am-12pm Pacific and then a second window around 5-8pm Pacific. I'm in LA so this is convenient for me.
But the timing thing is real. The exact same post at 2am gets 20% of the engagement it'd get at 10am. The algorithm's initial distribution window is short - if you don't get traction in the first 30-60 minutes, the post dies. So showing up when your audience is scrolling is a basic requirement that a lot of people ignore.
Being weird and specific instead of generically motivational
The fastest way to get lost in the feed is to sound like everyone else. "Consistency is key." "Ship fast, iterate faster." "Your future self will thank you." These are not posts. These are bumper stickers.
I'm Russian-born, ADHD, building AI systems in LA, async-only (no calls, ever), obsessed with multi-agent workflows. That's specific. When I write from that specific perspective instead of trying to be universally relatable, the people who connect with it really connect.
You can't optimize for everyone. You can optimize for your actual people.
What didn't work - including an embarrassing confession
AI-generated replies
This one I'm putting in writing because I think a lot of people are doing this and thinking it's working.
For a few months, I was using AI to generate replies to posts in my niche. The idea: stay "active" in conversations, look engaged, maybe get some follows from the visibility. The system did not notice - meaning I wasn't flagged or suppressed as far as I could tell. But the growth was essentially flat. Automated engagement looks active, but builds zero real connections.
Here's why it fails: a good reply starts a conversation. An AI-generated reply that's technically relevant but not genuinely insightful doesn't make anyone want to engage back. It gets ignored or gets a polite like. It doesn't turn into a thread. It doesn't make the original poster remember you. It doesn't build a relationship with anyone who saw it.
I switched to genuine replies only - less volume, more quality. Growth actually improved. Because one real conversation is worth more than 50 AI-generated non-interactions.
The meta-lesson: you can use AI to help draft a reply if you're stuck, but you need to inject your actual perspective into it. The algorithm might count impressions, but humans are better than you think at detecting genuine vs. automated engagement.
Mass following and hoping for follow-backs
Did this for like a week, stopped, never mentioned it until now. Following 200 accounts a day hoping 10% follow back is a number game that builds the wrong audience. Even when it "works" you get people who don't care about your content - they just returned a courtesy follow. Those accounts become noise in your metrics and won't engage with anything you post.
Generic "tips and tricks" threads
I tried a few of these early on. "5 AI tools every developer should know" type content. It did okay in terms of impressions but drove almost no follows. The problem is this content exists everywhere. There's nothing in it that could only come from me. When someone reads a generic tips thread, even a good one, they don't feel any reason to follow the source specifically. They just take the info and keep scrolling.
The posts that converted to follows were the ones with a clear point of view. Posts where you can tell who wrote them from the content alone.
The introverted builder angle
I want to flag something for anyone who's introverted and thinks Twitter growth requires networking, calls, or showing up to events.
It doesn't. I don't do calls. I don't go to tech meetups. I'm async-only. Every single one of my followers came through written content and written replies. No podcast appearances, no cross-promo deals, no "collab" posts.
The platform rewards writing, and writing is something you can do alone at 11pm in your apartment. For introverted builders, that's actually a structural advantage. You're competing with people who need external validation to show up - and you can just quietly be consistent.
Check out the building 7 apps solo for all the AI engineering and building-in-public content if you want to see how this thinking shows up across different topics.
Where I am now and what I'm focused on
660 followers, growing slower than the initial burst but more steadily. The first 500 came fast because I was very active with replies. I've pulled back on reply volume a bit as I'm heads-down on some builds, and the growth has slowed accordingly. That's fine - I'd rather have a smaller engaged audience than chase a number.
The things I'd do exactly the same: heavy reply game in the first month, being genuinely specific and opinionated, building in public with the insight-forward framing.
The things I'd change: started with genuine replies from day one instead of testing the AI reply thing, spent less time on generic content earlier.
If you're starting from zero or rebuilding like I was - the reply strategy is not glamorous but it's the most direct path. Find 10 accounts in your niche with 5K-100K followers, show up in their replies every day with actual value, and do it for 30 days before evaluating whether it's working. That's the playbook. The rest is just execution.
One last thing: don't optimize for follower count if you're building something. Optimize for the right followers. 500 engaged AI/dev/startup people is worth more than 5,000 general followers who won't care about what you're building. Niche compounds. Broad doesn't.
Find me at @deeflectcom if you want to watch this experiment continue in real time.
Top comments (0)