Ilus Blog

Blog

  • 24-25 Recap

    2024

    • Left my product manager job @ Kaufland eCommerce.
    • Travelled and island hopping in Indonesia for 4 months.
    • Joined 11th tier German Football club, and left after few matches due to injuries
    • Restarted Twentylab (Tech & Consulting Agency) after left them in 2020.
    • Built several SaaS, 0 MRR and decided to focus on the agency clients 😅

    2025

    • Joined a Dortmund-based company on part-time basis as digital strategist.
    • Co-founded TGC with my ex-boss and partnered with Twentylab
    • Co-founded Lucentis (an AI company focusing on medical training field) with 2 other cofounders
    • Twentylab officially got its business license in Germany.
    • Twentylab got its first long-term contract with a NRW-based company. (did the contract signing in a random classic german restaurant next to Ruhr River in Hagen)

  • Your Product Might just be 1-prompt away 

    Over the weekend I played around with the OpenAI API, trying to find out an idea to build, spent an hour on researching what can be built quickly, I ended up with a simple twitter profile summariser. 

    The goal is not building a great idea to hack on, but rather just building things from scratch with a backend language I am familiar with and with a little touch of OpenAI API. 

    You can check out the project in this repository: 

    https://github.com/apriandanailus/twitter-analyzer-with-openai-in-laravel

    So what it does is, I provide a form in which the user can put the URL of the twitter/X user, the backend will then validate the input, and a call to OpenAI user prompt API, and the API will give us the summary of the twitter user profile page. 

    As simple as it sounds, it’s basically what most AI Wrappers do. Definitely with more complex prompts, and more layers of prompt validation to catch different kinds of use cases. Years ago, it was a lot more difficult to build a service with this simple, NLP-aware engine or ML, a scrapper, and probably more. Today, it’s just 1 API call away, or 1 prompt away. 

    You might think, “this can be done on ChatGPT or DeepSeek Chat” . Yes, that is true. But, the idea is, if you simply change the objective of the sample-app I built, to a more complex, real-world use case, and streamline it, it can be a product someone or some company wants to use. For example, some people will not have a problem asking question to chatGPT on preparing an interview, they can upload a PDF of their CV, upload the job description and simply asking “Based on my CV and the job description, could you list points I need to highlight for this role” chatGPT will give you a nice answer using data you’ve fed to them. But, not everybody can ask a good question for some other use cases, not everybody is smart enough to actually come up with an idea that you can literally ask chatGPT anything. When looking further into the industry, there might be a lot of tasks and use cases that are waiting to be automated and streamlined by AI tools. 

    The whole idea of one-prompt-away is simple, to make it even easier for your user to get value. Chat-based interaction is nice, and easy enough to use, but there are still thousands of things a simple chat app cannot offer. As much as I believe that the future of app interaction is chat, I also believe that the traditional UI is there to complement it.