AI processes conversations the way humans think
AWAI's pipeline is built on a 3-layer context model that structurally aligns with cognitive science. Designed by intuition, validated by research — here's how it works.
Five products stand on one analysis
Every AWAI product runs the same analysis underneath. What changes is only what gets fed into it, which is why they all read the same way.
- Meeting analysisInputVoices in the room, as they are transcribed
- Group analysisInputOne AI conversation per person you invited
- Process analysisInputText you already have, such as a chat log with an AI
- AI PersonalityInputA few minutes of talking to the AI
- AI CompatibilityInputTwo finished personality results
Three layers that construct the 'meaning' of thought
Immediate Context
Related information within the current topic. All thought nodes in the same message are referenced without filtering.
Proximal Context
Nodes from 1–2 messages prior. Not everything is pulled in — text similarity filtering ensures only relevant context is considered.
Distant Context
Past nodes with high cosine similarity, regardless of message distance. The "oh, this connects to something I said before" association.
We designed this by asking "how does human memory actually work?" — then discovered that established cognitive science theories describe the same structure. Built from intuition and self-observation, later validated by research.
Designed by intuition, validated by science
| AWAI's Implementation | Cognitive Theory | Proposed by |
|---|---|---|
| Layer 1: Immediate context | Working Memory | Baddeley & Hitch (1974) |
| Layer 2: Proximal context filter | Selective Attention | Broadbent (1958) |
| Layer 3: Distant recall | Cue-dependent Retrieval | Tulving (1983) |
| Distance-unlimited high-similarity recall | Spreading Activation | Collins & Loftus (1975) |
| 2-pass reinterpretation | Memory Reconsolidation | Nader et al. (2000) |
| Processing depth based on prediction error | Predictive Processing | Friston (2005) |
| Unaddressed topic detection | Zeigarnik Effect | Zeigarnik (1927) |
Predict, discover, reinterpret
Naive Interpretation
Using only Layer 1 and Layer 2 local context, generate an initial "straightforward" understanding.
Contextual Reinterpretation
When Layer 3 finds distant context, incorporate it to update the understanding.
Integrative Reinterpretation
Only thoughts whose interpretation changed significantly between Phase 1→2 receive deeper processing.
Explicit connections and hidden links
LLM Edges
Relationships the AI explicitly identifies as connected. These are ideas that were actually linked during the conversation.
Embedding Edges
Pairs with high vector similarity. Ideas that are semantically close but were never directly connected in the conversation.
Making invisible connections visible — that's one of AWAI's core values.
Six colors that reveal the nature of thought
Process analysis hands the engine your own text
Meetings and group analysis both turn talk into text inside AWAI first, and then run this same analysis over it. Process analysis is the entrance that skips the first step and hands it text you already have.
After an hour with AI, you remember the conclusion. But what happened to all those ideas along the way?
The ideas you raised are kept, grouped by the topic they belong to. The ones that came up once and never connected to anything else are marked, so a thought you set aside without noticing is still there to pick back up.
Four steps to the ideas you forgot
Paste
Copy any AI conversation
Wait
AI extracts and classifies your thinking
Explore
See the topics and how they connect
Realize
Rediscover ideas you'd forgotten
In the space between thoughts, structure emerges.
In the space between minds, shared understanding is born.
"AWAI" is an archaic Japanese reading of "between." In the uncertainty before things are settled, possibility lives — AWAI makes that space visible.