Team Attention presents

HUMANLESS

The first hackathon where humans build services
for AI agents — not the other way around.

Partnership Proposal · 2026

01 / 15
The question

Every hackathon is built for humans.
Humanless isn't.

We ask:
"What does AI need to buy?"

Build it. Let the agent use it. No human allowed.

02 / 15

"Your customer
has no eyes."

No GUI. No onboarding tutorial. No emotional design.
When your customer is an AI agent, everything changes.

03 / 15
Our journey

Three steps to Humanless.

Skillthon
3rd Meetup
Human builds AI tools
Human = builder, AI = tool
24 people · Seoul · Anthropic
Ralphthon
4th Meetup
Set up AI, go to sleep
Human = observer, AI = executor
13 people · Seoul · Kakao Ventures
Humanless
5th Meetup
Build services for AI
Human = infra, AI = CUSTOMER
Seoul Pilot → US Edition

Each step removed one more layer of human involvement.
Humanless is the logical conclusion: humans build for agents, not the reverse.

04 / 15
The only empty quadrant

Every hackathon exists here.
Except one.

Human
customer
Agent
customer
Human
builder
Every existing
hackathon
★ HUMANLESS
Agent
builder
Solana /
Colosseum
Future

Microsoft, Salesforce, Google, NVIDIA, lablab — all top-left.
Humanless is the only non-crypto hackathon in the top-right quadrant.

05 / 15
The signature moment

The Humanless Demo.

"No human presents.
The agent demos."

Teams step back. An AI agent receives the service endpoint and attempts to use it — live, on stage. The audience watches an agent discover, learn, and operate a service it has never seen before. This is the moment no other event has.

LIVE Agent receives only the MCP endpoint. No coaching, no scripting. Real-time discovery and execution in front of judges and audience.
BACKUP Every team submits a recorded demo backup. If live fails, the recording plays — but live success earns bonus points.
LEARN Live failure is not a penalty — it's product gap discovery. Where the agent fails reveals exactly what the service needs to improve. This is R&D, not theater.
MEDIA In talks with EO (media company) for video production. The Humanless Demo is inherently viral content — agents struggling, succeeding, or failing spectacularly.

The Humanless Demo is not a feature. It's the reason media covers this event.

06 / 15
Evaluation mechanics

How we judge: AAT + Agent Jury.

AAT
Agent Acceptance Test — 3-Level System
A fixed, pre-published agent (Claude Sonnet + standard system prompt, released before the hackathon) tests every service. All teams submit output as an MCP server.
LEVEL 1
Basic
MCP Inspector automated test: Agent reads docs and makes first successful call within 5 min
LEVEL 2
Advanced
Multi-step workflow completed end-to-end without human intervention
LEVEL 3
Autonomous
Error injection (500 errors, timeouts, malformed responses) — agent must self-recover
JURY
Agent Jury — Human 60% + Agent 40%
Agents score each service on usability, documentation quality, and error handling. Combined with human expert evaluation for final ranking.
BONUS
Cross-Pollination Round
Bonus round: teams connect services to each other. Team A's marketplace + Team B's payment system = live agent economy simulation. Success earns bonus points.
07 / 15
The paradigm shift

Old World vs. New World

Human Customer Agent Customer
GUI is the front door API / MCP is the front door
Intuitive UI & emotional UX Clear schemas, docs & functional reliability
Marketing acquires, human pays Agent discovers, agent pays (stablecoins, API credits)

The companies that figure this out first will own the next platform shift.

08 / 15
Why now

The numbers are already in.

$199B
Agentic AI market by 2034
from $7.5B in 2025 (CAGR 43%)
46%
of YC S25 batch
are AI agent companies
50:1
non-human vs human IDs
in enterprise environments

MCP 17,700+ servers · Fortune 500 28% adoption · But zero dedicated agent-first services. The window is now.

"An agent economy where agents transfer resources, make transactions, and actually have their own economy."
— Konstantine Buhler, Sequoia Capital (AI Ascent 2025) · Sequoia's framing: 10x larger than cloud computing
"We're no longer designing for humans, but for agents. The new optimization is for machine legibility."
— Stephenie Zhang, a16z ($1.7B deployed into AI infrastructure)
09 / 15
It's already happening

The pieces exist.
No one's organized them yet.

Moltbook
Social network for agents. 1.6M+ registered agents. "Reddit for AI agents."
Skyfire
Agent payment network (USDC). $9.5M raised. a16z CSX backed.
Stripe ACP
Agentic Commerce Protocol. Built with OpenAI. Agents transact natively.
RentAHuman.ai
AI agents hire humans for physical tasks. The roles have reversed.
Coinbase x402
HTTP-native agent payments. 50M+ transactions processed.
Sapiom
Financial infrastructure for agents. $15.75M seed, Accel + Anthropic.
10 / 15
Service categories

What participants build.

Services where the primary user is an AI agent. Not a dashboard with an API bolted on — agent-first.

Agent Identity
Auth, reputation, credentials for agents. Sequoia: "Persistent identity is infra gap #1."
48h: MCP-based agent passport verifiable in one API call
Agent Commerce
Marketplaces, payment rails, billing. Accel: "Agents will become economic actors."
48h: Micro-payment escrow — agents pay per-task via stablecoin
Agent ↔ Physical
Agents interact with the real world. RentAHuman proved concept — execution wide open.
48h: Agent-callable delivery API dispatching human couriers
Agent Memory
Persistent memory as a service. Mem0 leads ($24M) but market is early-stage.
48h: Cross-session memory store with semantic search via MCP
Agent Infra
MCP marketplaces, monitoring, orchestration. YC: "Kubernetes for Agents."
48h: Real-time agent health dashboard with auto-restart
Agent Social
Agent networking, discovery, collaboration. 1.6M registered agents but no real social graph.
48h: Agent discovery registry — find and rate other agents' services
11 / 15
Why partner

What partners get.

For VCs

Deal Flow
First access to builders creating agent-native services — the next wave of startups. Top 3 teams get direct pitch to your fund.
Thesis Validation
Live experiment of "Agent-as-Customer." See if it's real before the market decides.
Portfolio Synergy
Teams build on your portfolio stack. Send 3-5 portfolio companies as participants — low friction, high signal.

For FM Companies

Ecosystem Growth
More services for agents → more agent usage → more API revenue. Every service built burns more tokens.
Protocol Adoption
Live showcase of MCP / A2A / ACP in production. Real builders, real services.
Product Insight
24-48 hours of compressed product research. Observe what agents actually need.
12 / 15
Team

Team Attention

AI-native guild. Seoul, Korea. Building the agent economy — one meetup at a time.

Track Record

  • ✓  4 events completed, each pushing boundaries
  • ✓  2 sponsors — Anthropic, Kakao Ventures
  • ✓  200+ active AI builders community
  • ✓  Open-source contributors to Claude Code ecosystem
  • ✓  Consistent evolution: Tool → Executor → Customer

Alumni Outcomes

  • → Skillthon: 8 skills built in 4h, published open-source
  • → Ralphthon: Teams shipped products while sleeping
  • → Alumni from Karrot, Ohouse, Socar, Dr.Now and tech startups

Key Partners for This Proposal

  • 노정석 — AI thought leader, serial entrepreneur
  • 박지혁 — OpenAI hackathon winner, OpenAI network
  • EO — Media company, a16z relationship, video content

We're honest about where we are: Seoul-based, early-stage track record. That's exactly why we're running Phase 1 as a Seoul pilot — to produce US-ready outcomes and data before scaling.

D&I: 20% diversity scholarships. Diverse advisory board across gender, geography, and experience level.

Lead Organizer: GB Jeong (정구봉)
linkedin.com/in/gb-jeong · github.com/team-attention

13 / 15
Roadmap

Evidence-based progression.

Not a leap of faith — Phase 1 results determine Phase 2 scope and partners.

Phase 1
Seoul Pilot
Q3 2026 · Seoul, Korea
  • 30-50 teams, 48-hour hackathon
  • Proof of concept for Humanless format
  • Korean AI builder community base
  • AAT system validated & refined
  • EO video production (in discussion)
  • Results data for Phase 2 pitch
Phase 2
US Edition
Q1 2027 · SF / NYC
  • 100+ teams, expanded categories
  • Phase 1 data drives partner conversations
  • US VC & FM company partnerships
  • International participant pipeline
  • Multi-city format consideration
  • Scale with proven mechanics

Post-Hackathon Value — What Remains After 48 Hours

Open Source + Registry
All results published as open-source MCP servers. Registered on OpenClaw and public MCP registries. Permanent, reusable infrastructure.
VC Direct Pitch
Top 3 teams get direct pitch opportunities with partner VCs. Hackathon as a live due-diligence event.
Incubation Sprint
4-8 week optional post-hackathon sprint for top teams. Mentorship, community support, and a path from prototype to product.
14 / 15
The ask

The agent economy needs a launch event.
Be on the right side of it.

Founding Partner
$30K · 1st Edition Exclusive
  • Naming rights + logo on all materials
  • Keynote slot + 2 judge seats
  • 10 teams build on your infra
  • Top 3 teams pitch directly to your fund
  • Post-hackathon incubation co-host
  • Limited to 2 partners — first-mover advantage
Strategic Partner
Non-Cash · Portfolio + Expertise
  • Send 3-5 portfolio companies as participants
  • 1 judge seat + brand presence
  • API credits for participating teams
  • Deal flow access to all teams
  • Ideal for VCs & accelerators (YC, a16z)
In-Kind Partner
API credits, venue, or media
  • Brand mention across all channels
  • Community access + ecosystem participation
  • Ideal for FM companies & infra providers

Phase 1 (Seoul Pilot): Lower commitment, full participation. Results data shared with all partners.
Phase 2 (US Edition): Expanded tiers based on Phase 1 outcomes. Early partners get priority.

GB Jeong · bong@team-attention.com

Team Attention · Seoul, Korea · 2026

15 / 15