PromptLoop Studio

## Day 2: PromptLoop Studio - Laying the Foundation Wow, Day 2 felt like a blur, but in a good way. A ton of foundational work got done, which is less visually exciting but absolutely critical. Feels good to have these core pieces in place. **Infra & Auth:** The biggest lift today was getting the user infrastructure properly hooked up. Integrating Clerk for managed auth (sign-up, login, profile management) is a massive time-saver, but still requires wiring it into the backend and frontend correctly. Successfully got users into their own private workspace environment. This is fundamental for handling user data, saving workflows, tracking usage, and eventually billing. It's the plumbing that makes everything else possible. Acknowledging the ease Clerk provides, but also the inevitable tiny config quirks that pop up. **UI & Canvas:** The other major push was getting the basic visual workflow builder live using React Flow. Spent a significant chunk of time getting the canvas set up – enabling drag-and-drop functionality, defining the initial node types (Input, Prompt, Output), and allowing basic connections between them. Right now, it's a simple linear flow capability, but seeing nodes appear and connect on screen felt like a real step forward. It's the core visual interface where users will build their workflows, so getting this basic interaction right was key. Still miles to go with node properties, complex connections, and real data flow, but the skeleton is there. **Wins:** Successfully shipped core user authentication and workspace provisioning. This feels huge for Day 2 – users can now technically 'exist' within the application. Also, got the very basic visual canvas working with initial node types. This is the heart of the product's UI. Having both the user shell and the core builder interface started makes the project feel much more concrete. Next up: making these nodes actually *do* something and persist workflow state. **Reflections:** Progress feels good, but it's a stark reminder of how much invisible infrastructure is needed before the flashy parts work. Auth and the basic canvas are solid wins, even if they just scratch the surface of the planned features. Staying focused on shipping core functionality first is the goal.

Read more →

## PromptLoop Studio - Day 1 Dev Log **Overall:** Day 1 is *done*. Feels good to get the absolute bare bones live and runnable. As a solo founder, the goal was just to get past the blank page and build the core foundational pieces that everything else will sit on. Authentication, check. A basic private workspace, check. The visual canvas where the magic *will* happen, check. It's ugly as sin right now, but it works. **Infra:** Got the entire infra stack provisioned and hooked up: Vercel for the frontend, Render for the Node.js/Express backend and the PostgreSQL database (managed via Neon, connected to Render). The absolute priority was user accounts and private workspaces for future usage tracking and billing (yawn, I know, but essential). Clerk is handling the managed authentication (signup, login, profile) and it was incredibly smooth to integrate – definitely a win opting for a managed service here to avoid that rabbit hole on Day 1. This infrastructure forms the spine; everything else connects to this user -> workspace -> data structure. **UI:** The biggest visual piece tackled was the core canvas using React Flow. This is where users will visually build their AI workflows. On Day 1, I got the drag-and-drop functionality working, allowing users to add basic 'Input', 'Prompt', and 'Output' nodes. Crucially, connecting these nodes via edges is also functional. Right now, it's just a visual representation – no actual data processing happens yet – but getting the interaction patterns and the flow logic laid out is a massive step. It proves the core visual builder concept is physically implementable within a day. **Wins:** The biggest win is simply getting the absolute core foundations in place. Auth is sorted, providing a secure base for users and future features. The visual canvas is operational at a basic level, which is the defining feature of PromptLoop Studio – enabling *anyone* to build AI workflows without code. Getting the basic node types and connections working confirms the feasibility of the core interaction model. This wasn't just wiring things together; it required thinking about the user journey and how nodes represent steps in a workflow. It's raw, but it's *real* progress towards bridging the gap between prompt engineering and automation for non-developers. **Missed:** Nothing specifically missed against the Day 1 goal. The goal was deliberately constrained to just these foundational pieces. Onwards to Day 2!

Read more →