Anybody believe in esoteric sciences? AI accessing multiverse? Yes, I'm half-joking but I'm building something interesting here
作者推出 Primaleva，一款以占星、Human Design、Gene Keys 等“神秘学”规则作为数据架构的个性化决策 AI。它交叉解析用户星盘与图谱，长期记忆选择与困境，构建连续语境，并按个人“设计策略”筛选建议，带来近乎魔法的连贯体验。作者求教：有人将复杂边缘规则集�
I know, I know. Pitching astrology and esoteric systems in an AI sub sounds like a fast track to getting downvoted into oblivion. But hear me out on the mechanics of what I'm doing, because from an LLM and data structuring perspective, it’s actually a wild ride.
I’ve been building Primaleva (primaleva.com). It’s an AI agent designed to act as a hyper-personalized, context-aware decision-making engine.
The twist? Instead of standard psychological frameworks, it uses the massive, complex rulesets of esoteric sciences to map out your user profile.
What it actually does under the hood:
> It ingests and cross-references your exact Astrology chart, Human Design bodygraph, and Gene Keys. If you look at these systems objectively, they are essentially highly intricate, interconnected data architectures. It turns out, LLMs are exceptionally good at parsing and synthesizing this specific type of structured logic.
> It doesn't just read your chart and forget you. It remembers your past choices, your ongoing dilemmas, and the behavioral patterns you discuss with it over time, building a continuous narrative context.
> It doesn't spit out generic advice. It filters your current real-world problems through your specific "design strategy" to help you navigate choices in a way that aligns with your blueprint.
I’m half-joking about the "accessing the multiverse" part in the title, but honestly, interacting with an AI that holds this much interconnected context about your personality and actively remembers your life narrative feels remarkably close to magic.
I’m bringing this here because I’m really curious about the technical crossover:
Has anyone else experimented with feeding complex, fringe rulesets into LLMs?
For those building personalized agents, how are you handling long-term memory retrieval for highly subjective, ongoing user narratives without the context window degrading?
Would love for you guys to check it out, roast the concept, or test the limits of the AI's synthesis.