Back to Hub
1 / 38
Builder Workshop

Sui Overflow 2026

3-Day Intensive — 4 Tracks to Win

Agentic WebDeFi & PaymentsWalrusDeepBook
Our Goal

This Is Not a Judging Panel

  • We are here to help you build better — not score you
  • Every track gets a critical deep-dive with actionable feedback
  • We study what won in 2025 so you know the bar
  • We warn you about judging criteria so you don't get disqualified or overlooked
  • By Day 3, every team leaves with a polished pitch and a sharper project

"The difference between a finalist and a winner is usually 3 small decisions made early."

The 4 Tracks

What We Are Building Toward

Core Track

Agentic Web (AI)

Agents that own objects, transact autonomously, and coordinate on-chain. $30K top prize.

Core Track

DeFi & Payments

Move-native financial primitives, real-time settlements, and payment rails. Sponsored by Pyth.

Specialized

Walrus

Programmable storage as a smart contract primitive. Not IPFS — on-chain access control + durability. $70K pool.

Specialized

DeepBook

Build on the shared on-chain orderbook. Novel interfaces, liquidity tools, and institutional-grade features. $70K pool.

Critical Warning

Project Eligibility — Read This Carefully

Build Period

  • Projects must be built during May 7 – June 21, 2026
  • Existing projects are allowed only if substantial new functionality was developed during this period
  • Judges will look at your commit history — a repo created in April with no new commits is a red flag

Disqualification risk: If your project was clearly built before May 7th with no meaningful new code, you will be removed from shortlisting. Don't waste your time.

Critical Warning

You Must Deploy to Testnet or Mainnet

  • Your project must be deployed to Sui Testnet or Mainnet at the time of shortlisting and demo day
  • You need a Package ID if your project has on-chain components
  • Local-only projects will not be shortlisted — no exceptions
  • Testnet is fine. Mainnet is better. No deployment = no review.
My project is deployed to Testnet or Mainnet
I have a Package ID ready for the submission form
I can access my app without localhost
Critical Warning

Submission Checklist — Missing = Rejected

Project Name — Clear and simple
Description — What it does, why it matters (2 sentences max)
Project Logo — 1:1 ratio, JPG or PNG
Public GitHub Repo — Must stay public during judging
Demo Video — Required. YouTube preferred. ≤ 5 minutes.
Website — Optional but highly recommended
Deployment — Testnet or Mainnet URL
Package ID — If you deployed on-chain

Most common failure: Teams spend 3 weeks building, then submit a 10-minute unedited screen recording with no voiceover. Your demo video is your first impression. Make it count.

Know the Bar

Judging Criteria — What Actually Moves the Needle

Real-World Application — 50%

This is half your score. Are you solving a real problem? Is there market relevance? Will this exist in 12 months?

Product & UX — 20%

Quality, usability, polish. A broken UI with great code loses to a smooth UI with decent code.

Technical Implementation — 20%

Code quality, reliability, and meaningful integration with Sui. Not just "we deployed a contract."

Presentation & Vision — 10%

Clarity, storytelling, and long-term vision. Can you explain your project to a non-crypto person?

Translation: A project that solves a real problem with decent UX and a clear Sui integration will beat a technically perfect project with no user. Build for people, not for judges.

Day 1 — Foundation & Core Tracks

Agenda

09:00Welcome & Eligibility Warnings
09:30Workshop: The Sui Advantage Audit
10:30Track 1: Agentic Web — Build AI That Owns Objects
12:00Lunch + 1:1 Office Hours
13:30Track 2: DeFi — Move-Native Design Patterns
15:00Day 1 Retro: What Changed in Your Project?
16:00Wrap-up & Homework
Pattern Analysis

What All 2025 Winners Had in Common

1. They Used Sui, Not Just Deployed On It

Winners leveraged object ownership, parallel execution, or zkLogin as core mechanics — not as afterthoughts.

2. They Had a Working Demo

Every winner had a live URL or video. Concepts without code never made the shortlist.

3. They Picked One Track and Owned It

Teams that tried to fit every track diluted their story. Winners went deep on one.

4. They Explained the "Why Sui" in 10 Seconds

Judges ask: "Why not Ethereum?" Winners had a crisp answer ready.

Workshop Exercise

The Sui Advantage Audit

Every team fills this out for their own project. Be honest.

Does my project break if I move it to Ethereum?
Am I using object ownership as a feature, not just storage?
Does parallel execution make my app faster or cheaper?
Have I used zkLogin or sponsored transactions for UX?
Is Walrus part of my data layer, or just a buzzword?
Can I explain my Move safety guarantees in one sentence?

Reality check: If you checked fewer than 3 boxes, your project is chain-agnostic. That's a problem for winning.

Core Track — $30K / $15K / $10K / $7.5K

Track 1: Agentic Web (AI)

The Winning Angle

  • Build agents that own objects, sign tx, and coordinate — not chatbots that read blockchain data
  • The bar: your agent should be able to hold a Sui object and act on it autonomously
  • Integrate zkLogin so agents have recoverable identity without seed phrases

2025 Winner: Suithetic — LLM-generated synthetic data with on-chain verification. The AI didn't just query Sui; it produced verifiable on-chain assets.

Critical Feedback

Agentic Web — Mistakes We See Every Year

❌ The "Wrapper" Trap

Off-chain AI calls an API to read Sui state. That's not an agentic web app — that's a dashboard with GPT.

❌ No On-Chain State

If your agent's decisions aren't recorded on-chain, judges can't verify autonomy or fairness.

✅ The Fix: Object-Centric Agents

Give your agent a Sui object (like a Cap) that authorizes actions. The agent's state lives on-chain.

✅ The Fix: Verifiable Output

Hash AI outputs to Walrus. Prove the model ran and the result is untampered.

Action Items

Agentic Web — What to Build This Week

Define one Sui object type that your agent owns and manipulates
Implement sponsored transactions so the agent acts without user gas
Add on-chain logging of agent decisions (events or shared objects)
Write a 3-sentence "why Sui" pitch for your demo day
Record a 2-minute demo video showing the agent transacting
Core Track — Sponsored by Pyth Network

Track 2: DeFi & Payments

The Winning Angle

  • Don't port EVM patterns — invent Move-native primitives
  • Use sub-second finality for real-time settlements or streaming payments
  • Integrate with DeepBook V3 or LBTC instead of building isolated pools

2025 Winner: Magma Finance — Programmable yield vaults with AI rebalancing. They didn't clone Yearn; they used Move's object model to let users compose yield strategies like LEGO blocks.

Critical Feedback

DeFi — Mistakes We See Every Year

❌ Forking EVM Logic

Uniswap-style AMMs on Sui miss the point. Move enables programmable objects — use them.

❌ Ignoring DeepBook

Building a new orderbook when DeepBook V3 exists signals you didn't research the ecosystem.

✅ The Fix: Object Composability

Design positions, vaults, or collateral as objects that other protocols can import and extend.

✅ The Fix: Real-World Settlement

Show a payment to a merchant, a payroll stream, or a BTC-backed stablecoin in action.

Action Items

DeFi — What to Build This Week

Audit your code: are you using Move's resource model for safety?
Integrate with DeepBook V3 or Pyth price feeds — don't reinvent
Add a real-world payment demo (even a mock merchant checkout)
Document your risk model — judges always ask about liquidation
Build a simple dashboard showing TVL or volume — metrics matter
Day 1 Retro

What Changed in Your Project Today?

Before

What was your project this morning?

After

What is it now? What did you cut or add?

Blocker

What's the one thing stopping you from shipping?

Winning Edge

What's your "why Sui" sentence?

Share with your table. The best feedback comes from other builders, not judges.

Day 1 — Closing

Homework for Tonight

  • Pick one 2025 winner in your track and study their repo or demo
  • Write down 3 things they did that you haven't done yet
  • Commit one meaningful change to your project before sleep
  • Prepare one question about your track for tomorrow morning

Tomorrow: Walrus, DeepBook, and cross-track composability.

Day 2 — Specialized Tracks & Integration

Agenda

09:00Day 1 Retro Share-Out
09:30Track 3: Walrus — Storage as a Primitive
11:00Track 4: DeepBook — Liquidity Layer
12:30Lunch + Office Hours
13:30Workshop: Cross-Track Composability
15:00Integration Pairing — Build Together
16:00Day 2 Retro & Demo Status Check
Share-Out

What We Learned from Day 1

Common Wins

[Teams fill this in — e.g., "3 teams added zkLogin today"]

Common Struggles

[Teams fill this in — e.g., "Move borrow checker confusion"]

Best Pivot

[Which team changed direction and why?]

Helpful Resource

[Share the doc, repo, or thread that saved someone]

Specialized Track — $70K Prize Pool

Track 3: Walrus

The Winning Angle

  • Walrus is not IPFS — it's programmable storage with on-chain access control
  • Winners treat storage as a smart contract primitive, not a file dump
  • Use Seal for encryption + Walrus for durability — that's the combo judges look for

2025 Winner: SuiSign — Decentralized document signing. They didn't just store PDFs; they created a signing workflow where access control, version history, and verification all live on Sui.

Critical Feedback

Walrus — Mistakes & Fixes

❌ "We Store on Walrus" (But Just Hash)

Hashing a file and putting the hash on Sui while the file sits on AWS isn't Walrus integration.

❌ No Access Control

If anyone can read your stored data, you missed the programmable part.

✅ The Fix: Seal + Walrus

Encrypt with Seal, store blob on Walrus, decrypt only for object owners. That's the pattern.

✅ The Fix: Storage as State

Your smart contract should reference Walrus blobs as part of its state transitions.

Action Items

Walrus — What to Build This Week

Store at least one meaningful blob on Walrus (not just a test file)
Implement Seal encryption so only authorized users can read it
Make the blob referenceable from a Move object — link on-chain and off-chain
Show a versioning or update flow (old blob → new blob via smart contract)
Document your cost model vs. traditional cloud storage
Specialized Track — $70K Prize Pool

Track 4: DeepBook

The Winning Angle

  • DeepBook V3 is a shared liquidity layer — don't compete with it, build on it
  • Winners create novel interfaces (prediction markets, RFQ, institutional tools) that plug into DeepBook pools
  • Show you understand flash loans, governance, and shared pools

2025 Context: DeepBook was embedded in winning DeFi projects. In 2026, it's a standalone track — the bar is higher. You need to show DeepBook as a platform, not just a price source.

Critical Feedback

DeepBook — Mistakes & Fixes

❌ Building a New Orderbook

Creating your own AMM or orderbook when DeepBook exists is a red flag. Use the shared pool.

❌ No UI Polish

Judges compare your UI to Binance or dYdX. A bare table of prices won't win.

✅ The Fix: Novel Interface

Build a prediction market, RFQ system, or institutional tool that routes through DeepBook.

✅ The Fix: MEV Awareness

Explain how your design avoids or mitigates MEV. Judges always ask.

Action Items

DeepBook — What to Build This Week

Read the DeepBook V3 docs and identify one feature you can expose
Build a UI that feels like a CEX — orderbook depth, fast execution, clean charts
Implement MEV protection or explain why your design avoids it
Show a liquidity bootstrapping mechanism (incentives, LPs, or airdrops)
Demo a cross-protocol trade (your app → DeepBook → another protocol)
Workshop

Cross-Track Composability

The best projects borrow from multiple tracks. Let's map how your project could plug into others.

Agentic Web + DeFi

AI agent that rebalances a DeepBook LP position autonomously based on market conditions.

Walrus + Agentic Web

Agent stores its decision logs on Walrus, with verifiable encryption via Seal.

DeepBook + Walrus

Store orderbook history or audit logs on Walrus for transparency and compliance.

DeFi + Walrus

Collateral documents, KYC blobs, or legal agreements stored on Walrus with on-chain access control.

Exercise: Pair with a team from a different track. Find one integration point. Spend 20 minutes sketching the shared object.

Build Session

Integration Pairing — Build Together

  • Form pairs across tracks (e.g., Walrus + DeFi, Agentic + DeepBook)
  • Spend 45 minutes building one shared interaction
  • Goal: a working function call or object handoff between your projects
  • End with a 5-minute demo to your pair partner

Why this matters: Judges love seeing ecosystem thinking. A project that composes with another team's work signals maturity and community mindset.

Day 2 Retro

What Changed in Your Project Today?

New Integration

What Sui primitive did you add today?

Cut Scope

What feature did you drop to focus on what matters?

Cross-Track Buddy

Who did you pair with, and what did you build together?

Demo Status

Is your 2-minute demo ready? If not, what's missing?

Day 2 — Closing

Homework for Tonight

  • Finish your 2-minute demo video — no exceptions
  • Write your README or one-pager as if a judge will read it at 11 PM
  • Test your app on Testnet with a fresh wallet — no pre-seeded accounts
  • Prepare 3 slides for tomorrow's dry-run pitch

Tomorrow: Pitch workshop, demo review, and dry-run pitches for all 4 tracks.

Day 3 — Polish, Pitch & Final Dry-Run

Agenda

09:00Day 2 Retro & Demo Video Review
09:30Workshop: The 3-Minute Pitch That Wins
11:00Demo Video Critique — All Tracks
12:00Lunch + Pitch Clinic
13:30Dry-Run Pitches — Agentic Web & DeFi
14:30Dry-Run Pitches — Walrus & DeepBook
15:30Final Feedback Round
16:00Wrap-up & Next Steps
Critique

Demo Video Review — What Works

Best Demo Flow

[Which team's video told a clear story from problem to solution?]

Common Video Mistakes

[Too long? No voiceover? No "why Sui" moment?]

Best README

[Which repo made you want to clone and run it immediately?]

Testnet Hero

[Who deployed something that actually works on Testnet today?]

Pitch Workshop

The 3-Minute Pitch That Wins

Structure

  • 0:00–0:20 — The Problem: One sentence, relatable, urgent
  • 0:20–0:50 — The Solution: Demo starts immediately. Show, don't tell.
  • 0:50–1:30 — Why Sui: Object model, parallel execution, zkLogin, Walrus — pick one and own it
  • 1:30–2:15 — The Traction: Users, integrations, or code quality. Be specific.
  • 2:15–2:45 — The Vision: Where does this go in 12 months?
  • 2:45–3:00 — The Ask: What do you need (grant, users, feedback)?

Common killer: Spending 90 seconds on "what is blockchain." The judges know. Start with your app.

Track-Specific Tips

Pitching Your Track — What Judges Ask

Agentic Web

Question: "Can this agent act without human approval?" Have a video of it transacting.

DeFi & Payments

Question: "What's the liquidation risk?" Have a risk model slide ready.

Walrus

Question: "Is the data actually on Walrus?" Show the blob ID and the on-chain reference.

DeepBook

Question: "Why not use DeepBook directly?" Explain your novel interface or tool.

Final Warning

Submission Prep — Don't Get Disqualified

Deployed? Testnet or Mainnet URL works right now
Repo public? GitHub is public and has commits from May 7 onwards
Video ≤ 5 min? YouTube link, with voiceover, showing the full flow
Logo ready? 1:1 ratio, not blurry, not a placeholder
Package ID? If on-chain, you have the ID for the submission form
Description clear? A non-crypto person understands what it does
Existing project? You can prove substantial new work during the build period

Last-minute disasters: Private repo, missing video, localhost-only app, logo is a screenshot. These are all disqualifiers. Check every box before you submit.

Dry-Run

Dry-Run Format — 5 Minutes Per Team

  • 3 minutes — Pitch (timed, no exceptions)
  • 1 minute — Demo video or live demo
  • 1 minute — Feedback from the room (2 things to keep, 2 things to fix)

Feedback rules: Be kind, be specific, be actionable. "Your UI is bad" is useless. "Your chart labels are unreadable on a projector" is gold.

Final Feedback

Last-Minute Fixes — 48 Hours to Submit

Deploy Now

If your app is still on localhost, stop everything and deploy to Testnet. Nothing else matters if you can't submit a URL.

Cut the Fluff

Remove features that don't work. A polished half-product beats a broken full product.

Record the Video

5 minutes max. Voiceover. Show the happy path. No blockchain tutorials.

Write the README

One page. What it does, how to run it, the Package ID, and a screenshot.

End of Workshop

You Are Ready

3 days of building, feedback, and iteration. Now go ship.

"The best projects aren't the most complex — they're the ones that clearly answer: why Sui, why now, and why you."
38 Slides4 Tracks3 Days

Final Resources

overflowportal.sui.io  •  docs.sui.io  •  go.sui.io/overflow26-tg