AI's Missing Primitive - The 15-Second Rule How micro-moments shape years of your life
ABOUT THIS EVENT
A notification at 11:47 PM. A scroll that lasts 15 seconds too long. A decision you make six months later.
These aren't separate events. They're the same event — just separated by time.
Behaviour doesn't happen in moments. It happens across timescales — milliseconds, hours, days, decades — each layer shaping the next in ways that no current AI system can see, let alone model.
Your LLM knows what you typed today. Your recommender knows what you clicked this week. Neither knows that 90 days of your data captures only 10% of who you are.
This is the temporal blindspot at the heart of every AI product ever built.
This Saturday, we're mapping it. And we're building tools to close it.
WHAT WE'LL DISCUSS
▸ Why behaviour has a temporal hierarchy — and why ignoring it makes every model wrong by design.
▸ The 15-second micro-moment: how short-timescale inputs create long-timescale outcomes.
▸ What LBM's multi-scale temporal modeling looks like in a live demo — working architecture, not slides.
▸ Why 90 days of behavioural data tells you only 10% of who someone is — and what the other 90% requires ▸ The causal chain from a single notification to a life-altering decision.
▸ What the base unit of behavioural time actually is — and why the field hasn't agreed on one yet.
▸ Where the model breaks — honestly, specifically, technically.
SCHEDULE
11:30 — Arrival & LBM Primer
11:45 — Thesis: Multi-scale Episodic Memory
12:05 — LBM Live Demo #6: Episode compression algorithms
12:20 — Hands-on: Design your episode schema
12:45 — Debate: What should AI forget?
13:10 — Episodic Schema Working Group formation
13:25 — Closing
WHO THIS IS FOR
▸ AI/ML researchers working on causal inference, RL, or foundation models
▸ Behavioural scientists, computational psychologists, cognitive neuroscientists
▸ ML engineers building personalisation systems and hitting the ceiling
▸ Professors and PhD candidates exploring modalities beyond language and vision
▸ Deep tech investors evaluating foundational AI infrastructure
WHAT YOU GET
▸ A LBM architecture overview — discussed, debated, taken apart
▸ The hardest questions answered without corporate polish
▸ For qualified participants: early access to full technical white paper under NDA
▸ Pathway to contribute to open-source and closed-source research tracks
WHERE WE ARE
This is Session 3 of an ongoing build series.
The foundation has been laid — the primitive defined, the architecture introduced, the community seeded. 20+ active members. Temporal modelling working group now forming.
If you've been in the room, you know what's been built. If this is your first time — the primer has you covered.
HOSTED BY
Suraj Biswas — Founder & CEO, Dots-In
Saurav Gupta — Co-founder & CTO, Dots-In
🌐 dotsin.ai
DETAILS
📅 Saturday, 25 April 2026
🕐 11:30 AM — 1:30 PM IST
📍 Dots-In Office, Bangalore
30 seats. No recordings. Chatham House Rule.