dilem.

    About dilem

    Describe your dilemma. dilem helps you choose.

    How it works

    1. Describe your choice in writing or by voice (1 min per recording). Dictation transcribes your words in real time, without translating them.
    2. A language AI (LLM) clarifies your dilemma and prepares the options. It only asks a question when the answer could change the choice; you can also ask for a recommendation right away.
    3. When recent facts matter, web research brings sources into the comparison.
    4. An AI specialized in decisions (System One) compares the options using this context and your own words, then returns a choice and a weight for each option.

    The LLM prepares the comparison

    An LLM is a language model: an AI that can understand and write text. In dilem, it turns your words into options and information useful for choosing, clarifying what matters to you. You do not need a perfectly structured problem to get started.

    System One compares the options

    System One refers to AI models specialized in structured decisions. In dilem, the model receives the prepared context and your own words, then returns a recommended option and weights rather than a freely written reply. The explanations you read are prepared by the language model, not written by System One.

    What this changes for you

    Understanding your dilemma and comparing your options are two distinct roles. Separating them lets dilem present a structured ranking with readable explanations. It does not make the decision objective or guarantee the outcome: if a priority is missing or changes, the recommendation can change too. You keep the final say.

    Percentages express a preference, not a probability of success. Open an option to see its reasons, trade-offs and what could change the choice.

    Explore further

    dilem uses OpenAI GPT-6 Luna for language and TypeSafe Jev 1.14, a System One model, to compare the options.

    Who is behind dilem?

    dilem is created and operated by Etienne Debost.

    Contact · [email protected]