Thanks for reaching out. Be sure to check us out on LinkedIn for all of our current news and announcements.
Oops! Something went wrong while submitting the form.
Highline Beta AI Builds: The Good, The Bad, The Ugly follows our own team building with AI in real time, what worked, what didn't, and what we'd do differently next time. No polished case studies. No pretending we had it figured out from day one.
Our first edition is based on Timothy Sun's experience. Read on to love, learn, laugh.
Tim Sun, Senior Product Manager
I built two AI-powered tools over the past few months:
AI Creative Studio The goal was to help people who may have limited design skills create materials such as social media content. Instead of starting from scratch every time, AI creates a first draft that a designer can review, refine, and improve. This makes the creative process faster and gives designers a starting point to work from.
Website Prompt Creator The goal was to help users create detailed prompts that they could copy and paste into tools such as Claude or Lovable to generate a micro-website concept. The prompt creator guides the user through 19 questions and then produces an established prompt that can be used to generate the website.
The problem I was trying to solve was that, for someone without a designer or developer, creating a website or prototype for concept or product testing can be time-consuming and frustrating. This tool is intended to reduce that barrier.
I primarily used Claude and ChatGPT.
I used Claude for the workflow, system diagrams, and coding because I find that it communicates technical logic very clearly. I am also on the Team plan, which made it a practical tool for my workflow.
I use ChatGPT primarily for natural language tasks and image creation. It is particularly useful when I need help communicating an idea clearly or generating visual concepts.
The first version looked very similar to what I had originally imagined and designed. The AI was able to translate my ideas into something that was fairly close to my initial vision.
Claude has been very effective at communicating the logic behind a solution, and I rely on it heavily for the technical aspects of development. Before starting development, I try to align my expectations with the development specifications and understand what Claude can and cannot realistically accomplish.
Claude is particularly helpful at providing a high-level understanding of the problem and proposing different development approaches. This helps me understand whether what I am asking for is technically feasible and whether it is worth the effort, or whether there may be a better path to achieve the same outcome.
In many cases, Claude was able to make what I asked for work and gave me the ability to go beyond what I initially imagined.
The bigger challenge for me is product maintenance and continuous improvement. I can create something, but with my existing workload, I often do not have enough time to maintain it and improve it the way I would like. Eventually, the product can fall so far behind that it is no longer worth using, especially when a third-party platform dedicated to solving the same problem can develop something much better.
The biggest problem was workload and the lack of dedicated time to continue working on a product after creating it.
It is extremely difficult to maintain and improve a product during my personal time, even when I try to work on it during evenings or weekends.
Ultimately, I had to let some of the products go.
However, there is a positive side to this. Every time I create a product, I become better at communicating with AI. I learn how to describe my ideas, define requirements, explain workflows, and communicate my expectations more effectively.
Using AI to build what I imagine, especially when I provide a clear Product Requirements Document, can take an idea a very long way.
The biggest unresolved issue is still time. Unfortunately, until I become a cyborg, I have a limited amount of time available to build, maintain, and improve products.
I know AI agents and automation can help improve efficiency, but I do not think AI can fully replace the creative thinking and continuous product judgment needed to independently improve a product over time.
Building a product is becoming much easier with AI, but committing to the product and maintaining it is much harder.
AI can significantly reduce the barrier to building something, but the real challenge is finding the time and commitment to keep improving it after the initial excitement of creation is over.
One thing I have learned is that AI is surprisingly resilient to small mistakes. A small typo or unclear phrase usually does not completely break its understanding of what I am trying to accomplish.
However, if you do not understand what you are building, AI can quickly turn the product into spaghetti. The more important lesson for me is that AI can help execute an idea, but you still need to understand the product, the logic, and the decisions behind it.
And one final lesson: swearing at AI does not help either.
