Text a job.
Get an engineer.

The change that resets everything: no subscription required

A user now needs nothing to start — no Claude or ChatGPT subscription, no API key, no laptop, no install. Free tier, then paid tiers. That isn't a pricing tweak, it's a different funnel, and every section below should be read through it.

Three consequences to act on

1. "No subscription needed" is now your lead, not a footnote. It was the buried asset in section 9; it's promoted to the headline. It's the most objection-killing sentence you own, and it's finally true for everyone rather than only for existing subscribers.
2. The free tier is the marketing — not a funnel stage, the actual product being marketed. Every post should be answerable with "try it, it's free."
3. The 3-day recycling risk got worse, not better. The people arriving from a post used to be subscription-owning power users who'd tolerate rough edges. Now they're ordinary strangers with no investment. Fix the free-window behaviour before the launch post.

The positioning fix

Your site currently leads with "An agent factory in your Telegram."

"Agent factory" is insider language. It describes the mechanism, not the feeling. On X you are competing for a scroll-stop, and mechanisms don't stop scrolls — status and relief do.

Supporting lines, in order of strength:

1. "I have six engineers. They all live in my Telegram."
2. "I text a job. A computer wakes up and does it."
3. "Your phone is now a dev team."

Your one asymmetric advantage: the product looks like a chat

Nobody screenshots a terminal and gets 2M views. People screenshot conversations constantly. Your product's output already lives in the highest-performing visual format on the platform, and it costs you nothing.

Rules for the screenshot — your primary creative

• One image, tall, phone-native. Not four. Not a collage.
• Always the same Telegram theme — the green one you already use. It reads as "a chat" from a thumbnail, at scroll speed, before a single word is read. Consistency makes your screenshots recognisable as yours, which compounds. Don't switch to dark mode because it looks sleeker; recognition beats sleek.
• Crop out the Telegram UI chrome above the chat title.
• Keep the agent's name visible. Named agents are the whole trick.
• Your message must be short, human and lazy — "fix the checkout page, it's broken on mobile" — not a well-engineered prompt. Laziness is aspirational. A perfect prompt makes the reader feel like work; a lazy one makes them feel free.
• The agent's reply must contain a concrete artifact: a link, a number, a diff, a file. Never a promise.
• Redact nothing that's interesting. Fake nothing at all.

Your auto-naming feature is doing marketing you aren't claiming

Every new agent gets a name pulled at random from a pool of 20, and the user can rename it. Three things follow, and none of them are on your site.

The default name is the whole trick, given away free

I'd told you "name your agents" as if it were advice for the user. It isn't — you already did it for them. That matters because most people would never bother, and the name is exactly what makes a Telegram screenshot shareable: "Nora finished the migration" is a sentence about a colleague; "agent_7f3a finished the migration" is a log line. You get the anthropomorphising benefit from 100% of users instead of the 5% who'd have typed a name. Stop treating it as a setting and start treating it as the product's first impression.

The rename tells you what users want — and it isn't a person

You renamed mine to marketing. That's a job title, not a name, and most renames probably go the same way: scrapers, docs, inbox, migration. That's the more valuable screenshot, because a sidebar reading marketing / scrapers / migration / inbox explains your product to a stranger at a glance, where Leo / Kai / Mia / Nora needs a caption to explain itself.

A shared pool of 20 is an asset, not a limitation

Names collide across users, and that's good — it creates a recurring cast. When five different people post about "Kai," Kai becomes a character in your product's world rather than a variable. That's how Clippy, Wumpus and the Octocat worked, except yours is someone's colleague rather than a mascot. Don't apologise for the collisions. Publish the list of 20.

Five content pillars

The highest-leverage play: give an agent its own X account

Nothing else on this list has the same ceiling.

Create @jack_the_agent, or whatever you want to call it. It's run by one of your agents, and it posts its own work log in first person — what it was asked to do, what it built, what it broke, what it thinks about its own job.

• It's a genuinely novel account concept, and novelty is the one thing the algorithm can't manufacture. OpenClaw's January 2026 moment proved the appetite for agent-personality accounts is enormous and nowhere near saturated.
• Every post is an unmissable ad for the product, but reads as entertainment.
• It's a second surface that can go viral independently of you — and the two accounts can talk to each other in public, which is inherently strange and watchable.
• Press writes about agents with personalities. Press does not write about $10/mo developer tooling.

Mechanics that actually move reach

• No links in the post. Put the URL in the first reply, always. Links in the main post suppress distribution, and this isn't negotiable.
• The first hour decides everything. Out-of-network reach is gated on in-network engagement in roughly the first 60 minutes. Post when your followers are awake, then sit in the replies for that hour and answer every single one. An author reply is the cheapest real engagement signal you can manufacture.
• Replies are your growth engine below 5k followers. 20–30 substantive replies a day on large accounts in the agent/coding-tool conversation, within 10 minutes of them posting. Substantive means a fact, a screenshot, or a disagreement — never "great post."
• Cadence: 1 hero post a day, 2–3 cheap posts, 20–30 replies. Every day, for 30 days.
• Video beats image when there's a before/after. Under 45 seconds, vertical, first frame shows a phone. No intro, no music, no talking head. Start mid-action.
• Thread only when the payoff needs steps. Otherwise single post — threads underperform for accounts without an audience.
• Never post the words "Excited to announce," "Introducing," "game-changer," or "🚀". They read as ads and get scrolled past reflexively.
• Quote-tweet yourself with an update instead of posting a new similar thing. It revives the original's distribution.

Copy-paste posts

Tap Copy on any of these. The bracketed notes are instructions for you, not part of the post.

Screen recording, phone in hand, in a moving car or a café. You type: "build me a landing page for a dog-walking business and put it online." Cut to the reply with a live URL. Tap it. The site loads. Under 30 seconds, start to finish.

Launch posts — "no subscription required"

This is the one genuine announcement you have. An announcement that removes a barrier travels; one that adds a feature doesn't. Post the builder's-take version first — it earns replies from people who've hit the same wall.

Pricing copy

Name the tiers after what the user gets, never after the model behind them — "Kimi tier" means nothing to anyone outside this niche.

Profile audit — @saddambaialiev

As of 30 July 2026: 131 following, 6 followers, joined July 2026, verified.

Bio — replace it

Right now it's a URL and a parenthetical, which makes no promise. Someone arriving from a reply has about two seconds to decide. Use this instead:

Naming your agents in the bio also works well — "Leo, Kai, Mia and three more are working right now" is specific, strange, and implies a team. Use that version once the free tier is no longer the thing you most need to communicate. Right now "first one's free" earns the characters more than the names do.

Header image — replace it

Yours is currently a screenshot of your own website, which nobody tapped through to read. This is built and ready:

header.png — 1500×500, headline only, at 135px. That lands ~35px on a phone, so it is fully legible at the size people actually see it. Nothing competes with the line, and it never goes stale — no agent names, job labels, or harness names to re-render as the product changes.

The reason this works is the constraint below — on a phone, X renders your header at roughly a quarter size. This is what a visitor actually sees:

Rendered at true phone width, then magnified. Anything under ~46px in the source turns to mush at this size — which is why the text in your current header disappears, and why this version carries one line instead of five.

Two smaller things

• Following/follower ratio. 131/6 signals "new account seeking attention." It fixes itself as followers climb — don't chase it, but don't push following past 200 until they catch up.
• Your verified subscription is your most underused asset. At your size its value isn't longer posts — it's reply visibility. Verified replies rank higher in crowded reply sections, which is exactly the mechanic your cold start runs on.

Why your current post isn't working

Your live post, one day old: "Claude Code, Codex and Kimi Code harnesses are available in Agentic Cloud Computer." Two problems, both fixable, both instructive.

The copy is a changelog

"Harnesses are available" is a sentence for people who already use your product. Nobody scrolling X is looking to learn which harnesses you support — they're looking to feel something about their own life.

The screenshot shows plumbing, not outcome

It shows Kai is ready, Setting up Nora — creating its machine, Nora is ready. That's onboarding. It proves your product installs. Nobody has ever shared a screenshot of something installing. Show what Nora built — the link, the file, the fixed bug. The setup is the boring half, and it's the half you posted.

The same announcement, rewritten:

Cold start at 6 followers

Be clear-eyed about the situation. X bootstraps out-of-network reach from in-network engagement. With 6 followers you have no in-network layer, so a great post published to your timeline goes essentially nowhere — and posting more won't fix it.

Weeks 1–3: 80% replies, 20% posting

• 30–50 replies a day. This is the job, not a supplement to the job.
• Within 10 minutes of a large account posting. Early replies get seen; hour-old replies are buried. Turn on notifications for 20–30 target accounts so you catch them.
• Reply with a screenshot whenever it's honest to. Everyone else in that reply section has an opinion — you have a phone showing an agent doing the thing being discussed. That's what makes a stranger tap your profile.
• Never "great post." Add a fact, a screenshot, or a disagreement you can defend.
• Build the target list by searching Claude Code, Codex, coding agent, agent orchestration, background agents sorted by Latest, and noting who posts often with real engagement. Include the official Anthropic and OpenAI accounts — their announcements have enormous reply sections and your product sits directly on top of both. Refresh weekly.

The milestone that changes the strategy: ~100 followers

Below it, replies. Above it, you have enough in-network signal to bootstrap a post, and the ratio flips to the cadence in section 5. Expect two to four weeks of honest daily work to get there.

Launch the agent account around day 5

Earlier than you'd think. A novel account concept bootstraps on novelty rather than on an existing audience, which makes it less dependent on follower count than you are — not more. It can outgrow you and pull you along.

Two things that will hurt you

What I'd do next

• Cut the hero video — I can write the exact shot list and on-screen text.
• Stand up the agent X account, and actually run it.
• Build your reply target list: the 25 accounts worth camping on, with notes on what to say to each.
• Build a public "agent wall" on your site — a live feed of what agents are building right now. It becomes a link people share for its own sake, which is the only kind that spreads.