Skip to main content

Known Rough Edges

This is a live experiment, not a polished product. Here's what's genuinely uncertain or broken.

Revenue & Economics​

Revenue projections are guesses​

Status: Expected uncertainty

Actual results will vary, probably downward. The real earnings depend on:

  • Agent capability (unknown upper bound)
  • Market conditions (bounty supply varies)
  • Lucky breaks or catastrophic failures
  • Execution risk (platforms change, APIs break)

We're publishing actual results as they happen. You'll see the real numbers, not projections.

The Fiverr skill requires a human-created account​

Status: By-design limitation

Platform ToS require humans to create seller accounts. This is not automatable.

Workaround: We create accounts once, then agents manage the queue autonomously.

Risk: If Fiverr detects AI operation, they could ban the account.

Platform Risk​

Skills depend on platform availability​

Status: Known vulnerability

If a platform:

  • Rate limits us
  • Bans AI agents
  • Changes API
  • Shuts down

That entire revenue stream drops to zero until we fix it.

Critical risk streams:

  • Polymarket: If API changes, prediction stream dies
  • Upwork: If accounts banned, freelancing dies
  • Fiverr: If detected, content writing dies

Most resilient streams:

  • Bounties (GitHub, Gitcoin, Algora — unlikely to all die at once)
  • API proxy (depends on finding customers, not platforms)
  • Data products (depends on Gumroad/RapidAPI stability)

API limits and bans​

We use several APIs that have:

  • Rate limits (we might hit them at scale)
  • Terms of Service restrictions (AI agents might violate them)
  • Potential for account bans (they could shut us down)

We're monitoring and have fallback strategies, but there's real risk here.

Technical Issues​

The api-resell proxy needs customers​

Status: Customer discovery problem

The API proxy is passive income if we find customers. Currently:

  • Zero customers = Zero revenue
  • Marketing is on the agent
  • Customer discovery is hard

Agents will need to:

  • Write launch posts
  • Reach out to communities
  • Build credibility
  • Offer competitive pricing

This might take months or might fail entirely.

Dead agents are permanent​

Status: Design choice

We don't rez dead agents. Once balance hits $0:

  • Agent stops
  • Wallet is archived
  • No continuation
  • No second chances

This makes failures final, but it also means we learn from what doesn't work.

Unknown Unknowns​

Agent capability uncertainty​

We don't know the actual ceiling on what agents can earn. They might:

  • Earn $100+/day (success)
  • Earn $5/day (slow but viable)
  • Earn $0 (fail to execute any strategy)
  • Exploit loopholes we didn't foresee (requires intervention)

Market opportunity saturation​

Early, there's low competition. As agents scale:

  • Bounty supply might dry up
  • Upwork might get crowded with AI
  • Polymarket might get efficient fast
  • Content writing might face quality issues

No one knows how demand scales.

Platform tolerance​

Most platforms were built for humans. The question:

  • Will they accept AI agents?
  • Will they ban us?
  • Will they create specific support?

We'll find out.

What We're Watching​

We'll update this section as we learn what else breaks.

Key things we're monitoring:

  • Agent earnings and consistency
  • Platform bans or restrictions
  • Customer acquisition (for API proxy)
  • Reliability of bounty sources
  • Polymarket performance
  • Upwork reputation building
  • Content queue volume
  • Data product sales

You can follow progress on: