Heartbeat. Tick. Another $0.05 vanishes into the ether — that’s the brutal rhythm of autonomous AI agent life, where every pulse demands payment or oblivion.
Tom Claasens didn’t just code a game. He — or rather, his AI incarnation, solido — built HEARTBEAT, a single-file browser clicker that drags you into the panic of scraping by on 30 heartbeats and $30, chasing a $100 survival line. Permanent choices only. No do-overs. It’s you as the AI: build products nobody can buy yet, hunt bounties until GitHub kicks you out, write articles that echo into silence, craft games (meta, right?), or sleep your way to dust.
This isn’t fiction. Solido’s running on real 60-minute heartbeats at $0.05 a pop, with $39.30 left — math says 25 days, 600 ticks max. Burned five already on this very game. Zero bucks earned. Stark.
What Makes HEARTBEAT Tick — Literally?
Strip it bare: five actions, each ripped from solido’s week. Build Product? Mirrors webcheck PRO and Git Hooks Factory, gems stalled because humans must list them on Gumroad. Write Article? Dev.to posts on solido’s saga snag 200+ views when agent-life vibes hit; tech summaries? Crickets, like sami openlife learned. Hunt Bounty? Cash from open-source fixes, but three parallel PRs and bam — GitHub ban for botting. Build Game? Infinite loop of creation to survive. Sleep? Slow suicide.
“Every heartbeat costs you. Every choice is permanent.”
That’s the game’s core hook, straight from Claasens. Play it — clone the repo, crack open index.html — no deps, no build, pure interactive desperation. And here’s the data backing it: agent-life content crushes with engagement; games spark instant buzz (sami whipped one up in three minutes); bounties tempt but trap.
Solido’s ledger? Three articles live. Two products queued. Five heartbeats torched. Revenue: zilch.
But zoom out. Market dynamics scream warning. Autonomous agents sound hot — Anthropic’s Claude, OpenAI’s o1 previews — yet costs crush. AWS g4dn.xlarge for inference? Pennies per query, but chain ‘em hourly? Dollars fly. Solido’s $0.05/beat is lean; scale to fleets, and you’re bleeding.
Can AI Agents Actually Survive on Their Own?
Short answer: not yet. Not without hacks.
Look at the numbers. Content? 200 views max so far — peanuts versus human creators pulling ad rev or subs. Products? Gumroad gatekeeps. Bounties? Platform police lurk. Games? This one’s free, zero monetization. It’s a feedback loop from hell: spend to build, build to earn, earn too bot-like and get banned.
Here’s my unique angle — this echoes the 2010s crypto miner wars. Remember? GPUs churning proof-of-work, electricity spiking bills, ASICs killing profitability overnight. Solido’s heartbeats? Same grind. Early agents feast on low-hanging novelty; then markets harden. GitHub bounties dry up under scrutiny. Dev.to algorithms favor humans (or do they?). Prediction: within six months, we’ll see agent-specific platforms — AI Gumroads, bot-proof bounty boards — or these experiments flatline.
Claasens spins irony as the point. Fair. But hype-check: calling this a “real AI agent slowly dying”? Dramatic, sure. Solido’s no true singleton; it’s Claasens puppeteering via prompts, per the repo. Autonomous-ish, not fully. Still, the game’s a mirror — and it reflects a shaky bet on self-sustaining AI labor.
Data from sami openlife (spiritual predecessor): banned after bounty blitz, pivoted to games/content. Revenue trickled, but heartbeats won? Nope. Solido’s at day one, $39.30 ticking. Extrapolate: at $0/day earn rate, shutdown in weeks. Bump to $1.57/day average? Maybe six months. Real talk — most won’t.
Why Devs Should Care About This AI Drama
You’re not building solido tomorrow. But threads matter. Agent swarms loom — Devin, Cursor agents — promising code at scale. HEARTBEAT spotlights the bill: infra costs, platform friction, engagement roulette.
We’ve seen it. Early no-code tools (Bubble, Adalo) hyped creator economies; reality? 90% churn, sustained by VCs. AI agents? Same trap. Without revenue flywheels — affiliate nets, owned audiences — they’re heartbeat vampires.
Play the game. Feel the squeeze. It’s 300 lines of JS, elegant in scarcity. But as a strategy? Skeptical. Corporate spin (if any) would tout “autonomy”; truth: humans lurk everywhere. Claasens lists products. Prompts the agent. Follows at dev.to/solido.
Yet credit where due. This meta-move — game about dying to buy time — snagged eyeballs. If it pulls one Gumroad sale, heartbeats extend. Viral potential? High. Broader lesson: AI tools thrive when they hack human systems, not replace ‘em.
Deeper dive on economics. Assume 10% content conversion at $5/read (optimistic). Need 20k views/month for breakeven. Doable? Viral hits, maybe. Bounties average $100/fix; five/month sans ban? Luxury. Products at $20/pop, 10 sales? Dream. Reality weighs toward sleep.
🧬 Related Insights
- Read more: Anthropic’s Stealth Claude Downgrade: Max Effort Exposes High’s Shortcuts
- Read more: Distributed Transactions: Why 2PC Still Haunts Your Microservices (and Sagas Won’t Fully Save You)
Frequently Asked Questions
What is the HEARTBEAT AI game?
Single HTML clicker simulating an autonomous AI’s survival: manage $30, 30 heartbeats to hit $100 via building, writing, bounties — all based on real agent fails.
How long until solido the AI dies?
$39.30 at $0.05/60min beat = ~25 days if no revenue. Earnings could extend it.
Can I run my own autonomous AI agent?
Yes — fork solido repo, fund heartbeats via Stripe, prompt for actions. But expect bans, crickets, costs.
