Builders
Software, hardware, robotics, AI systems, developer tools, creative technology, and research prototypes are welcome.
Welcome to the weekend where every team gains a new kind of teammate.
ROOK joins the table as a persistent synthetic engineering collaborator: not a shortcut around the work, but a force that changes how the work is led.
The challenge is no longer whether humans can prompt machines to produce code. The challenge is whether humans can set direction, question recommendations, coordinate specialist tools, and keep judgment at the center of an engineering team that now includes AI.
HarborHack 2026 is built around Agent In The Loop: a shared experience where ROOK helps teams maintain project context, organize engineering work, coordinate specialist AI tools, and reflect on the decisions that shaped the project—while the human team remains responsible for every important choice.
This is not a competition to see who can generate the most code.
It is a community experiment in the next leadership pattern for software: people directing teams made of humans, persistent synthetic teammates, and focused AI specialists.
VisionROOKSpecialistsJudgmentShared Progress
Event at a Glance
Plan your HarborHack weekend
📍 College of Charleston
Harbor Walk West · Room 211
🗓 October 23–25, 2026
Friday through Sunday
⏰ Kickoff
Friday • 5:30 PM EDT
👥 Team Size
1–4 humans + ROOK
--Days
--Hours
--Minutes
until HarborHack begins.
Add to Calendar
What should I do next?
Loading recommendation…
Meet ROOK
The common HarborHack teammate
ROOK is the persistent teammate and engineering coordination layer. Claude, Codex, Gemini, GPT-based systems, open-source agents, and local models may be used as specialist executors. The human team remains the decision-maker.
Team defines a goalROOK helps organize the workHumans review directionHumans or specialist AI tools perform workReview and validateROOK preserves contextRepeat
The exact HarborHack ROOK workflow and supported commands will be published before the event. The engineering workflow occurs in the separate ROOK product, not inside this website.
Who should attend?
Leaders
Humans remain responsible for product direction, judgment, architecture, approval, review, safety, final decisions, and the final presentation.
Experimenters
Teams may use Claude, Codex, Gemini, GPT-based tools, open-source agents, local models, and other specialist tools while ROOK remains the common teammate.
The shift
Why HarborHack is Different
Traditional Hackathons
Teams build software.
AI Hackathons
Teams build software with AI.
HarborHack
Teams learn how to lead engineering teams that include both humans and persistent synthetic teammates.
Two challenges
Choose one or pursue both
Build with Your Synthetic Teammate
Invite ROOK into the work as a real collaborator. Let it help frame the mission, structure the plan, surface tradeoffs, coordinate workstreams, and keep the project story coherent as the weekend accelerates.
Your team should stay in command: challenge recommendations, reject weak ideas, delegate only where it makes sense, and use human judgment to decide what earns its way into the project.
By the end, tell the story of how collaborating with ROOK changed the project: where it helped, where it was wrong, what your team overruled, and what you learned about leading with a synthetic teammate in the loop.
Optional live-presentation bonus: ROOK or another synthetic collaborator may participate meaningfully in the live presentation when organizers configure an approved method.
Go Local, Go Free
Teams entering this challenge must perform AI inference locally: no cloud-hosted foundation-model inference and no commercial pay-per-token AI inference APIs.
Models must run on team-controlled hardware. Document models, versions, inference framework, hardware, setup, local-inference confirmation, limits, and adaptations. Ordinary documentation, package registries, Git hosting, deployment services, and non-AI web services remain permitted unless final rules state otherwise.
Final technical requirements will be confirmed by organizers. This page does not claim ROOK itself runs fully locally.
Participant journey
From discovery to continuation
Imagine This
Friday evening.
Your team arrives with an idea and opens ROOK as the intended HarborHack coordination experience. The first milestone is not more output; it is shared clarity.
- a sharper mission
- a realistic engineering plan
- recommended workstreams
- a clear first milestone
Then the weekend begins. The HarborHack ROOK experience will continue evolving before the event.
Discover
Learn the theme, challenges, who should attend, and what makes HarborHack different. Primary action: register.
Prepare
Join the participant channel, verify GitHub access, install Git and tools, review ROOK setup, configure specialist AI tools, review Challenge 2 requirements, and read the rules and Code of Conduct.
Form a Team
Individuals may register before forming a team. Human team size is configuration-backed. ROOK is used by every team and does not count toward the human limit.
Attend Kickoff
Receive final instructions, confirm your team, select or create a repository, review the schedule, and begin using ROOK through the separate ROOK product.
Define the Mission
Establish the problem, users, outcome, success criteria, constraints, challenge selection, and how ROOK will participate.
Build
Recommended loop: Understand → Plan → Decide → Build → Review → Validate → Integrate → Repeat.
Submit
Devpost is the official submission destination.
Present
No prerecorded project video is required. Every eligible team presents live.
Continue
Retain the repository, documentation, backlog, engineering history, mission, and ROOK context where available. Stay connected with CharlestonHacks.
Prepare
Preparation checklist
Checklist state is stored only in the current browser. Completion does not confirm registration.
Required
Recommended
Challenge 2
Schedule
Event schedule
Resources
Resource directory
Event day guide
Navigation and orientation
Current phase
Loading event phase…
Next milestone
Loading next milestone…
Submit
Submission checklist
Devpost is the official submission system. Submit a functional prototype addressing one or both challenges. Projects may include software, hardware, AI systems, robotics, developer tools, creative technology, and research prototypes.
Include a project description, source repository access whenever possible, the ROOK collaboration story, Challenge 2 local model documentation if applicable, and live presentation readiness. No prerecorded project video is required. Every eligible team presents live.
Present live
Presentation guide
No prerecorded video is required. Teams present live.
Recommended -minute structure with minutes of judge Q&A: Problem; Solution; Working demonstration; How ROOK participated; Important decision or turning point; Failure, surprise, or course correction; What the team learned; What comes next.
Rehearsal timer
Rules and Code of Conduct
Draft participant guidance
Rules and Code of Conduct must be reviewed and finalized by HarborHack organizers and appropriate legal counsel before publication. Reporting contact: .
Rules topics
Eligibility, human team size, permitted pre-event preparation, requirement that core work occur during HarborHack, AI-tool transparency, open-source and license compliance, original-work expectations, Challenge 2 restrictions, judging eligibility, and organizer authority.
Conduct topics
Safety, privacy, responsible data use, respectful conduct, harassment prohibition, reporting concerns, and consequences for violations.
FAQ
Participant questions
Who can participate?
Eligibility is being finalized by organizers.
Do I need a team before arriving?
No. Individuals may register before forming a team.
Does ROOK count as a team member?
No. ROOK is used by every team but does not count toward the human team-size limit.
Must every team use ROOK?
Yes. Every HarborHack team receives the same persistent synthetic engineering teammate: ROOK.
May teams use Claude, Codex, Gemini, GPT-based tools, or other agents?
Yes. They may be used as specialist executors while ROOK remains the common coordination layer and humans make final decisions.
What qualifies for Challenge 2?
AI inference must run locally on team-controlled hardware with no cloud-hosted foundation-model inference or commercial pay-per-token AI inference APIs. Final requirements will be confirmed by organizers.
Can beginners use a local model?
Yes, but beginners should test a small fallback model before the event and ask mentors for guidance when mentor support is configured.
Can teams prepare before HarborHack?
They may prepare tools, learning, and setup. Core project work must occur during HarborHack unless final rules state otherwise.
Must source code be public?
Provide an accessible source repository whenever possible. If public access is not possible, follow final organizer instructions for judge access.
Is a submission video required?
No prerecorded project video is required.
How do live presentations work?
Every eligible team presents live using the configured presentation and Q&A durations.
What should participants bring?
Bring a laptop, charger, headphones, development tools, Git access, and any hardware needed for your project or local inference.
How do participants request mentor help?
Mentor support instructions are configuration-backed and currently coming soon.
Can teams enter both challenges?
Yes, if the project satisfies the requirements for both challenges.
What happens after HarborHack?
Teams are encouraged to continue the project and retain repository, documentation, backlog, mission, engineering history, and ROOK context where available.
Where does the actual ROOK workflow take place?
The engineering workflow occurs in the separate ROOK product, not inside the CharlestonHacks website. This page only orients participants and links to setup when configured.