In May 2026 the group co-hosted the OpenClaw Agenthon Indonesia with Build Club, an agent hackathon that drew 205 participants. The weekend made one thing plain: the teams that shipped a working agent were the ones that picked a narrow, real problem early and refused to widen it.

What separated the teams that shipped

Across the room, the finished agents had less in common technically than you would expect. What they shared was discipline of scope. The teams still searching for the perfect idea at hour ten rarely shipped.

  • They chose a single task a real person does often
  • They wired up one reliable tool before adding a second
  • They tested against messy real input from the start
  • They wrote down what the agent should refuse to do

Where teams lost time

The common time sinks were predictable and avoidable. Credentials and access took longer than any team budgeted for, so the groups that sorted keys and permissions in the first hour pulled ahead. Chasing model choice early was the other trap, since a smaller setup was usually enough to prove the idea.

Why hackathons matter for the local scene

Events like this give Indonesian builders a deadline, a peer group, and permission to ship something rough. That combination teaches more in a weekend than weeks of solo study. The group runs and co-hosts these because the local AI scene grows fastest when builders meet in one room.

Note on the host

AICON co-hosted the Agenthon alongside Build Club and the OpenClaw community. Credit for the event belongs to everyone who organized and showed up.

Frequently asked questions

What was the OpenClaw Agenthon Indonesia?

A May 2026 agent hackathon co-hosted by AICON, Build Club, and the OpenClaw community. It drew 205 participants building AI agents over one weekend.

What helped teams ship a working agent?

Narrow scope early: one real task, one reliable tool wired first, testing on messy real input, and a short list of what the agent should refuse to do.

Where did teams lose the most time?

Credentials and API access took longer than most teams planned. Groups that sorted keys and permissions in the first hour moved faster. Chasing model choice before the idea was proven was the other common trap.

Why run agent hackathons in Indonesia?

A deadline, a peer group, and permission to ship something rough teach more in a weekend than weeks of solo study. The local AI scene grows fastest when builders meet in one room.