Skip to content

Ship Something

Building the system that takes AI prototypes beyond the demo.

I’ve spent the last year building products with AI. And while getting something working has become dramatically easier, I’ve learned that building a prototype and shipping a real product are two very different things.

The Ship Something homepage: “Ship Something Real” above a mock dashboard listing four projects and their current status

AI can generate an interface and write code in minutes. But testing, security, payments, deployment, and everything that comes after are still problems someone has to solve.

I kept running into those same challenges with my own projects.

So I started building Ship Something, a platform that gives solo founders a way to go from an idea to a working product, without having to manage all the technical work themselves.

Instead of another coding environment, I wanted to turn the process I’d developed building my own products into something anyone could use.

shipsomething.com
Role
Product designer, builder
Timeline
2026
Models
Task-routed LLMs for planning, building, and independent review
Status
Public website live; core building and verification workflow tested on real projects

01 / ORIGIN

Turning my own workflow into a product

After building several products with AI agents, I’d developed a process that worked. We’d turn an idea into requirements and milestones, agents would build, and independent reviewers would catch what needed fixing.

The problem was that I still had to coordinate every step and keep things moving.

I started wondering whether I could turn that process into a product, giving other founders a way to move beyond the prototype without having to manage an entire team of AI agents themselves.

02 / EXPERIENCE

Designing around the founder, not the agents

I wanted founders to focus on their products, not on managing AI agents or learning another development environment.

The experience starts with a conversation. Someone describes what they want to build, reviews a proposed plan, and approves the work before it begins.

Behind the scenes, specialized agents handle implementation and testing. The interface keeps the founder informed without overwhelming them with the technical details.

03 / DECISIONS

Some decisions shouldn’t belong to AI

The hardest part wasn’t getting AI agents to do the work. It was deciding what they should never be allowed to do.

During testing, the system mistook a conversation about a milestone for permission to start building. The financial safeguards prevented a charge, but the misunderstanding exposed a bigger problem.

Even when a model correctly interprets a request, that doesn’t mean it should have the authority to act on it.

That distinction became a fundamental part of the product.

Ship Something has a live public website, and its core process, from requirements and planning through implementation, review, and testing, has been exercised on real projects.

The broader vision, including ongoing maintenance after launch, is still a work in progress.

The next challenge is turning a process that works under my direction into something other founders can use reliably on their own.

After spending so much time building products with AI, this feels like a natural next step: making the experience of building them a product itself.

From idea to production, with AI as a collaborator.

Glossy black robot character with glowing eyes

Discover

Understand the problem

  1. Step 1: Frame the idea

    Problem, audience, desired outcome

  2. Step 2: Explore together

    Questions, alternatives, tradeoffs

  3. Step 3: Validate assumptions

    Research, feasibility, evidence (AI research - I decide)

Define

Set the direction

  1. Step 4: Define the product

    Requirements and success criteria

  2. Step 5: Plan the work

    Milestones, tasks, acceptance tests. AI drafts

  3. Step 6: Approve the plan

    I revise scope before work begins. My approval.

Deliver

Build, verify, release

  1. Step 7: Build + review loop

    Implement, agent audit, fix, recheck (AI and agents)

  2. Step 8: Connect + test

    Services, env setup, staging QA

  3. Step 9: Ship + learn

    I approve release; we iterate. My approval

I use AI throughout the product development process, from exploring an idea and challenging assumptions to defining requirements, building, and reviewing the implementation.

But I don’t want a model that simply agrees with me. I expect it to question decisions, identify tradeoffs, and provide evidence that the work actually functions.

I review the plan, direct the experience, and use independent agents to help verify the implementation. Before release, I test the product and make the final call.

AI makes it possible to move much faster. My responsibility is making sure we’re moving in the right direction.