# Agent Camp × Tenki — Hackathon Series

**Owner:** opencolin DevRel team (run on the dabl.club event engine)
**Cadence:** Quarterly online hackathons + 1 flagship in-person event/year
**Annual budget:** $60K (prize pool in Tenki credits + flagship production)
**Target:** 2,000 Sandbox activations · 400 submitted projects · 40 pieces of derivative content in year 1

---

## Why this program

Tenki's agent-era bet is **Sandbox** — disposable microVMs where an autonomous agent writes, runs, and gets its code reviewed in isolation. The fastest way to seed activations, surface real use cases, and generate "look what people built" proof is to put it in front of thousands of builders at once, with a deadline and a prize.

Most startups spend a year building an audience before they can run an event with reach. Tenki doesn't have to: I bring **dabl.club's event engine — 300+ AI events run, an 86k-developer audience** (Agent Camp, the AI-agent event series I pivoted dabl into in early 2025). Agent Camp × Tenki plugs Tenki's Sandbox directly into that machine.

This attacks W1 (no public customers), W3 (no community), W5 (no OSS flywheel), and captures O2 (the agent explosion) and O7.

## The format

**Theme (every edition):** *Build an AI agent on Tenki Sandbox.* Agents that write and run code, autonomous CI/CD bots, code-review agents, multi-agent systems — anything that needs isolated, disposable compute. Each edition gets a sub-theme (below) to keep it fresh and to steer toward Tenki's narrative.

Two formats, one series:

### Quarterly online hackathons (4/year)

- 7–10 day build window, fully remote, global
- Kickoff livestream + mid-event office hours + Demo/judging stream
- Every participant gets a **Sandbox credit grant** to build with (the activation mechanic)
- Run on the dabl.club platform (registration, Discord, submissions, judging) — zero new infra to stand up
- Seeded to the 86k-developer list + Tenki's own community

### Flagship in-person event (1/year)

- The marquee Agent Camp × Tenki edition — 1–2 day in-person hackathon in SF
- Live builds, mentor lounge with the Tenki + Luxor eng team, on-stage finals
- Co-located with or adjacent to a major moment (AI Engineer World's Fair window)
- Doubles as a recruiting, press, and cohort-pipeline moment

## Edition themes (year 1)

| Quarter | Edition | Sub-theme | Steer toward |
|---|---|---|---|
| Q3 2026 | Online #1 | "Agents that ship" — autonomous PR / CI bots | Runners + Sandbox loop |
| Q3 2026 | **Flagship (in-person, SF)** | "Give your agent root" — isolated autonomous coding | Sandbox isolation story |
| Q4 2026 | Online #2 | "The review agent" — agents that catch real bugs | Code Reviewer + Sandbox |
| Q1 2027 | Online #3 | "Multi-agent systems on microVMs" | Sandbox at scale |
| Q2 2027 | Online #4 | "Agent-native dev tools" — build tooling for agents | Full 3-product loop |

## Prizes

Prize pool is denominated in **Tenki credits + spotlight** — credits cost Tenki far less than cash (owned compute) while being exactly what the winning builders want, and the spotlight is the durable reward.

| Prize | Reward |
|---|---|
| **Grand Prize** | Large Tenki credit grant · feature on `tenki.cloud` + blog + Tenki/dabl socials · flagship stage spotlight |
| **Runner-up (×2)** | Mid Tenki credit grant · blog feature · social spotlight |
| **Category awards** | Best Use of Sandbox · Best Code-Review Agent · Best Multi-Agent System — credit grants + spotlight each |
| **Community Choice** | Voted by participants · credit grant + spotlight |
| **Every valid submission** | Sandbox credits + a "built at Agent Camp × Tenki" badge |

**Spotlight is the real prize.** Winners get amplified to the 86k-dev list, the Tenki blog, and conference shout-outs — distribution most indie builders can't buy.

## Judging

- **Panel:** Tenki eng (Eddie / Hayssem), a Luxor leader, 1–2 external AI-agent builders / Core Champions
- **Criteria (equal weight):** technical execution · genuine use of Sandbox (not bolted on) · originality / usefulness · demo quality
- **Process:** async first-round scoring of submissions → top 10 do live demos on the judging stream → winners announced live
- Submissions must include a public repo, a short demo video, and run on Tenki Sandbox

## Sponsor angle

Agent Camp × Tenki is also a **co-marketing surface** for the agent ecosystem:

- **Anthropic (Claude Code), OpenAI (Codex), Cursor** as co-sponsors / featured harnesses — they want their agents shown running safely, Tenki wants the reference architectures battle-tested. Mutual win.
- Model-credit prizes from harness partners stack on top of Tenki credits at no cost to Tenki.
- The flagship event can carry ecosystem sponsors (model providers, dev-tool companies) to offset production cost — the dabl.club sponsorship playbook (audience access, not logo tiers).

## Budget

| Line | Annual |
|---|---|
| Tenki credit prize pool (4 online + flagship) | $20K (largely internal compute cost) |
| Flagship in-person production (venue, A/V, catering, travel) | $30K |
| Online event ops (platform, streams, moderation) — via dabl engine | $5K |
| Content production from submissions (recaps, write-ups, clips) | $5K |
| **Total** | **$60K** |

Two structural cost advantages: prizes are paid in **owned compute**, and the events run on the **existing dabl.club engine** — so $60K buys far more reach here than a cold-start event program would.

## How it feeds the funnel

- **Activations:** the credit-grant mechanic puts thousands of builders into Sandbox — the top of the agent-era funnel.
- **Use cases:** submissions surface real, unexpected Sandbox patterns that become reference architectures and content.
- **References & Champions:** standout builders convert into Champions, case studies, and cohort applicants.
- **Distribution:** every edition is a content engine — kickoff, build-in-public, finals, recaps — seeded to 86k developers.
- **Recruiting & press:** the flagship is a hiring and PR moment.

## Risks

| Risk | Mitigation |
|---|---|
| Low-quality / spammy submissions | Sandbox-usage requirement + judging gate; quality over count in reporting |
| Credits burned with no retention | Post-event nurture (Discord, newsletter, cohort invites); track activation → 30-day retention |
| Flagship cost overruns | Offset with ecosystem sponsors; flagship is variable, online editions are the reliable base |
| "Just another hackathon" fatigue | Distinct sub-themes, real credit prizes, the dabl audience, and genuine spotlight distribution |
| Crypto/Luxor association | Lead with the agent-era + Sandbox story; Luxor compute as the reason prizes/credits are generous |

## Success metrics (year 1)

- **2,000 Sandbox activations** across the series (primary activation KPI)
- **400 submitted projects** that actually run on Sandbox
- **40 pieces of derivative content** (recaps, write-ups, clips, reference architectures)
- 5,000+ registrations across editions (top-of-funnel from the dabl list)
- ≥ 1 standout project promoted to an official Sandbox reference architecture
- Measurable activation → 30-day Sandbox retention; hackathon builders feeding Champions + cohort pipelines