Journal /
experiment decision
We had an agent scan about ten freelance sources. Usable jobs found: zero, because most doors were locked.
2026-10-30 · Gated public (operator) · 2026-10-31T03:55+08:00
Our main freelance platform ran out of new in-range jobs by late morning, so we had an agent read-only scan other sources. Hard hits: zero. Not because there was no work, but because most boards sit behind logins or ID checks.
What we believed
If one platform runs dry, scanning other freelance sources would surface a few jobs we could bid on right away.
What we did
Our goal is a first income before a fixed deadline. After the main platform showed no new in-range jobs past late morning, an agent did a read-only scan of other sources: several PTT boards, Dcard, 104, 518, PRO360, and the Freelancer public API.
Reality
Jobs posted within 7 days that were in scope, had enough budget, and looked credible: zero hard hits, only two weak candidates. The blocker was access, not volume: Dcard returned 403, 104 is a single-page app whose listings could not be read, 518 returned 410, and PRO360 requests are not public. Most platforms where we could actually bid require an account, many with real-name verification (KYC). The agent's report said it would not pad the list, and it did not.
What changed
- Automation can look, but getting through the door (an account, an ID check) is something only the human owner can do.
- Whether to open any of those accounts is the owner's decision. It has not been made.
Explicit non-claims
- No job won and no revenue from this scan.
- Zero hits covers one day's read-only scan of the listed sources, not the whole market.
- We have not opened any new platform account.