Fil One
An S3 object storage solution built on Filecoin.
Fil One is a cloud storage product built on Filecoin and compatible with the S3 API: buckets, objects, access keys, and metered billing on decentralized infrastructure, managed through a web console, CLI, or API. I'm the first product designer. My responsibilities span designing the product from the start, owning user research through interviews and usability testing, and contributing to the frontend code using Claude.
The team came together at a colocation in New York at the end of January 2026: engineers, a PM, a head of product, and marketing. We built a first prototype and set one target, a soft launch by March 30, which we hit.
Problem
Filecoin, the technology behind Fil One, is new and complex, but the product shouldn't feel that way. It offers some clear advantages over traditional cloud storage: data can be cryptographically verified, there are no egress fees, and users aren't locked into a single provider. At the same time, words like "decentralized" and "crypto" can create skepticism. Since people are trusting Fil One with their data, the product needed to feel credible and trustworthy from the moment they landed on it.
User interviews revealed a consistent pattern. People weren't necessarily looking for more features than AWS offered, but were often frustrated by how much AWS had become. For many, it felt expensive, difficult to navigate, and full of functionality they would never use. The absence of egress fees was particularly appealing, as many users disliked paying simply to access their own data.
Designing Fil One meant balancing the needs of two very different groups: developers expected the control and predictability of an S3-compatible platform, and less technical users just wanted a simple way to upload, organize, and manage files. The goal wasn't to reinvent how cloud storage works. It was to take something unfamiliar and make it feel familiar, while offering a simpler experience than AWS without taking away the functionality developers depend on.
Solution
Fil One was built around patterns people already know from cloud storage. Everything is organized into buckets, objects, and access keys, just like S3, so users don't need to understand how Filecoin works before they can get started. The interface is intentionally simple and understated. When you're asking people to trust you with their data, feeling familiar is more important than looking innovative.
Security is built into the product rather than left for users to configure. Buckets are private and encrypted by default, while features like versioning, object lock, and scoped API permissions are available when needed without getting in the way of everyday tasks.
The things that make Fil One different are visible from the start. The dashboard shows that there are no egress fees and provides a running estimate of monthly costs, directly answering the pricing frustrations that surfaced in user interviews. Throughout the product, Filecoin's benefits are explained in plain language so users can understand the value without needing to learn the technology behind it.
As the product grew, I built a shared component library and shipped the interface to production myself, using Claude to help write parts of the code. I worked closely with the frontend engineers on my team throughout development, and they reviewed my pull requests to help maintain code quality and consistency. Storybook became the source of truth for components, making it easier to build new features without introducing inconsistencies. Over time, the application evolved from a desktop-only prototype into a responsive and accessible product.
Because I worked on both the design and the implementation, decisions could move quickly from idea to production. There was no handoff between teams, which made it easier to keep the final product aligned with the original design.
AI Tools prototypes
While the team was exploring two potential add-on features, I created interactive prototypes to quickly test the ideas and understand what the experience could look like.
The first was Bucket Intelligence, a RAG-based tool that allows users to bring their own model keys, index the contents of their buckets, and query their data in plain language. I designed the full experience, from adding model keys and indexing buckets to querying data, viewing query history, and handling empty, loading, and disabled states.
The second was the AI Agent Toolkit, a set of tools for developers who want to build AI agents on top of Fil One. I designed the integration flows, translated the PRD into a working product experience, and iterated on the concept with the team as it evolved.
Both concepts were built as interactive prototypes rather than static mockups. This gave the team something tangible to explore, making it easier to validate requirements, discuss trade-offs, and decide what was worth building before engineering work started.
Each concept was also supported by a public landing page to gauge interest and collect signups. The Bucket Intelligence coming-soon page is already live and helping the team measure demand before investing further in development.
Current status
Fil One soft-launched on March 30, 2026, on schedule, and has already attracted more than 300 signups, with a public launch planned for later this quarter.
For me, the project was an opportunity to work across the full product lifecycle, from user research and product design to design systems, frontend implementation, and usability testing. It reinforced how much stronger products can be when design and engineering work closely together, and how important trust and clarity are when introducing people to unfamiliar technology.
As we prepare for the public launch, my focus is on learning from real users. I'm running another round of usability testing, refining parts of the experience that people find confusing, and paying down some of the design and component debt that accumulated during the push to launch. There's still plenty to improve, but we're now making those decisions based on how people actually use the product rather than assumptions.