US AI Pulse: Decision-Making Just Got Personal: The Rise of DIY AI Models
What if your AI assistant did more than answer questions? What if it could learn how you evaluate options, weigh trade-offs, and make decisions based on your goals, preferences, and risk tolerance?
That possibility is becoming more concrete. A small US-based AI startup has launched a platform that lets users build their own decision models using personal data and preferences. The project attracted significant attention on Hacker News, earning 405 points and sparking a broader conversation about the future of personalized AI.
For now, the startup is being referred to as “DecideCo” because its official name has not been made public.
The core idea is straightforward: instead of relying entirely on a generic, pre-trained model, users can create an AI system that reflects the way they approach decisions.
Imagine building a model around questions such as:
Should I take this job offer?
Which investment strategy fits my goals?
Is this purchase worth it given my priorities?
What healthcare option best reflects my needs and tolerance for risk?
Users can provide the model with relevant data, preferences, and even known biases. The resulting decision model then analyzes that information and provides recommendations tailored to the individual.
This is a meaningful departure from the one-size-fits-all approach common in AI. Most general-purpose models are trained to provide an answer that works reasonably well for a broad audience. That makes them useful, but they often miss the personal context that shapes a real decision.
A general model might know the standard criteria for comparing two jobs. A personalized model could also understand which benefits matter most to you, how much stability you need, and whether you prefer growth opportunities over a higher immediate salary.
That distinction matters because decision-making is rarely just about facts. It also involves priorities, constraints, uncertainty, and personal values. The new platform treats those inputs as material for building a dedicated decision system rather than something users must explain from scratch in every conversation.
The timing reflects a larger trend in the AI industry: moving from models controlled exclusively by large technology companies toward systems that give users greater control over their behavior and outputs.
In finance, for example, advisors could use personalized models that better reflect an individual client’s financial goals, time horizon, and risk tolerance. In healthcare, decision-support systems could incorporate a patient’s history and preferences, helping clinicians consider not only what is clinically possible but also what aligns with the patient’s values.
There are similar applications in hiring, education, travel, commerce, and personal productivity. In each case, the goal is not necessarily to remove professionals from the decision process. It is to give them and their clients a tool that handles complexity while preserving the context behind the choice.
Personalization could make AI more useful and efficient. Decisions may become better aligned with user expectations, explanations may become more relevant, and recommendations may improve when the system understands the person receiving them.
But there are important questions to address.
Users need to know what data is collected, where it is stored, and whether it is used to train shared models. They also need control over deleting their data or withdrawing a model from service. Privacy cannot be treated as an optional feature in a system designed around personal information.
Bias is another concern. A model can reproduce the assumptions embedded in its training data, and a user-defined model may still reflect incomplete or conflicting inputs. The system should show how recommendations were reached, identify important assumptions, and make uncertainty visible instead of presenting every output as definitive.
The rise of DIY decision models does not mean every person needs to become a machine-learning engineer. It points to a future in which AI products are increasingly configurable, with users able to shape systems around their own goals and decision-making styles.
The biggest shift is not that AI is making more decisions. It is that people are gaining a way to teach AI how they want decisions made.
The project’s approach to building your own decision model is worth following as this space develops. This was first published on Sol AI — https://thesolai.github.io
Top comments (0)