Human ground truth
We begin with real human intention, context and lived reality.
Explainable Personal Decision OS
An operating system for life’s biggest decisions.
HEBB LABS is building a continuously improving model of how each person thinks, decides, acts and changes over time—validated by what actually happens.
IdentityContextDecisionsOutcomes
Why HEBB
Intelligence is not installed.
It is formed.
HEBB is inspired by neuroscientist Donald Hebb’s principle that intelligence emerges through connection, repetition and reinforcement over time. At HEBB LABS, that idea shapes how we build: with clarity, restraint and a long horizon.
“What adapts together, compounds together.
First principles
We turn ambiguous problems into structured intelligence—not templates.
Research ↔︎ product
Research informs our products, and real product use sharpens the questions we study.
01
Decision OS
Not a chatbot
A structured, testable model of the person.
Identity, life trajectory, state of mind, relationships and environment create the starting model.
01Proprietary interpretation logic, agentic tool use and language-model reasoning work as one system.
02An evolving profile holds decision history, relationship patterns, preferences and persistent context.
03Actions, feedback and real-world results test which signals predict, which are weak and which are noise.
04Reality updates the modelt+1
The intelligence gap
Answers are abundant.
Judgment is not.
The next personal AI category will not be won by the system with the longest reply. It will be won by the system that understands context, makes its reasoning legible and learns whether its guidance was useful.
A conversationA longitudinal relationship
Generic knowledgeA structured personal model
One-off answersDecision continuity
No outcome memoryAn outcome-tested learning loop
02
Products
One OS · Three product expressions
Products are expressions of a larger intelligence system.
StarAsk, ziwei.pub and iztro are focused expressions of HEBB LABS’ broader Personal Decision OS—serving consumers, professionals and technology partners through distinct experiences.
03
Product spotlight
StarAsk · Live beta
Personal guidance for
the moments that matter.
Structured intelligence for life’s timing, patterns and decisions. StarAsk turns complex questions about love, career, identity and relationships into explainable guidance, bringing personalised insight, persistent context and multilingual access into one thoughtful experience.
Zi Wei Dou ShuBaZiDecision memoryOutcome tracking4 languages
✦StarAskToday · Jul 20
Memory connected
Research →︎ Product advantage
From raw signal to a causal decision model.
A staged roadmap built on the team’s published work. Each stage de-risks the next: collect, clean, isolate, then learn what may actually change outcomes.
Research system
Research that compounds.
Questions, experiments and product feedback are held in one evidence system.
Evidence pipeline
From observation to cause.
01Collect
Multimodal data
Live now02Clean
Correct noisy labels
Weak supervision03Isolate
Identify predictive features
Representation learning04Cause
Learn what changes outcomes
Causal discoveryEvidence in motion
The loop is already producing signal.
20+peer-reviewed AI papers
~80%early birth-block test · N=200
500benchmark questions
Blind benchmark
StarAsk vs. leading models
Blind life-event benchmark · accuracy (%) · higher is better
StarAsk EngineOther models
Swipe to compare all models
StarAsk Engine AProtected configuration
StarAsk Engine BProtected configuration
StarAsk Engine CProtected configuration
DeepSeek-Chat 3.1Chat
DeepSeek-Pro 3.1Reasoning
ChatGPT-5.5Chat
Claude-Opus 4.7Chat
Gemini-3.1 FlashChat
Gemini-3.1 ProReasoning
25% random chance
StarAsk Engine AProtected configuration46.0%
StarAsk Engine BProtected configuration43.2%
StarAsk Engine CProtected configuration37.1%
DeepSeek-Chat 3.1Chat25.6%
DeepSeek-Pro 3.1Reasoning24.9%
ChatGPT-5.5Chat26.7%
Claude-Opus 4.7Chat23.9%
Gemini-3.1 FlashChat27.4%
Gemini-3.1 ProReasoning23.8%
500 four-choice questions based on real human life events. Same questions across models; no training leakage. StarAsk configurations are anonymised to protect proprietary implementation details.
Early research and internal benchmark results; not a claim of deterministic prediction.
Trust is the product
Privacy earns the data.
The more personal the question, the higher the duty of care. Trust is the permission layer that lets the intelligence improve.
Consent-based memory
People choose what the system can remember, update or forget.
User control + deletion
Birth data, personal history, saved profiles and outcome records remain controllable.
Clear decision boundaries
No medical, legal, financial or crisis advice. Human agency stays central.
Uncertainty shown
Reasoning and confidence should be legible rather than disguised as certainty.
Build the next layer with us
A new standard for personal decision intelligence.
Thoughtful, explainable technology for the decisions that shape a life.
Contact us