MENTRIA_

mentria

A feed that teaches.

Cards you read, answer, and come back to. Import a pack from a book or your notes, or start with ours. Everything runs on your device.

Browse packs
Source

Source · 8 cards · 6 min

How a model decides what to look at.

Open
Source

Source · 8 cards · 6 min

How a model learns where each word stands.

Open
Source

Source · 8 cards · 6 min

How a word becomes a point in space.

Open
Source

Source · 8 cards · 6 min

How text becomes the pieces a model reads.

Open
Source

Source · 8 cards · 6 min

How a row of scores becomes a confident choice.

Open
Source

Source · 8 cards · 6 min

The one dial between safe and surprising.

Open
Source

Source · 8 cards · 6 min

How a model learns — by minimizing its own surprise.

Open
Source

Source · 8 cards · 6 min

How a model improves at all — one small step downhill.

Open
Source

Source · 8 cards · 6 min

How a model assigns blame across a billion knobs at once.

Open
Source

Source · 8 cards · 6 min

Why making it bigger keeps making it better — predictably, then suddenly.

Open
Source

Source · 8 cards · 6 min

Why a model makes things up — and sounds so sure doing it.

Open
Source

Source · 8 cards · 6 min

How a model learns what we want — by being shown what we prefer.

Open
Source

Source · 8 cards · 6 min

Why each next word stays cheap to generate.

Open
Source

Source · 8 cards · 6 min

Why twice the context costs four times as much.

Open
Source

Source · 8 cards · 6 min

How a model holds many specialists — and wakes only a few.

Open
Source

Source · 8 cards · 6 min

Exact attention, sped up by never writing the giant grid down.

Open
Source

Source · 8 cards · 6 min

How a model makes a picture by removing noise.

Open
Source

Source · 8 cards · 6 min

How a giant model learns your task by training almost nothing.

Open
Source

Source · 8 cards · 6 min

Why 'think step by step' buys real accuracy.

Open
Source

Source · 8 cards · 6 min

How a model picks up a new task from the prompt — without changing a thing.

Open
Source

Source · 8 cards · 6 min

How one tiny wire lets a network go truly deep.

Open
Source

Source · 8 cards · 6 min

How a model keeps every layer's signal from exploding or fading.

Open
Source

Source · 8 cards · 6 min

Why handicapping a model on purpose makes it generalize.

Open
Source

Source · 8 cards · 6 min

The line between memorizing and understanding.

Open
Source

Source · 8 cards · 6 min

Why one neuron means many things at once.

Open
Source

Source · 8 cards · 6 min

How understanding arrives long after the memorizing is done.

Open
Source

Source · 8 cards · 6 min

How a model runs in four bits without falling apart.

Open
Source

Source · 8 cards · 6 min

How a small model inherits a giant's skill.

Open
Source

Source · 8 cards · 6 min

How a small model drafts and a big one just checks — same words, far faster.

Open
Source

Source · 8 cards · 6 min

How a model encodes where a word sits — by turning, not tagging.

Open
Source

Source · 8 cards · 6 min

Why scale and search keep beating our cleverest tricks.

Open
Source

Source · 8 cards · 6 min

How a model reads the whole without weighing every word against every other.

Open
Source

Source · 8 cards · 6 min

How a model reads on and on without re-reading a thing.

Open
Source

Source · 8 cards · 6 min

How attention heads share their memory — and shrink it many times over.

Open
Source

Source · 8 cards · 6 min

Where a model's facts actually live.

Open
Source

Source · 8 cards · 6 min

Two ways to learn a language — and what each one becomes.

Open
Source

Source · 8 cards · 6 min

Turn 'A beats B' straight into a better model — no reward model, no RL.

Open
Source

Source · 8 cards · 6 min

Give the model an open book: let it look things up instead of making them up.

Open
Source

Source · 8 cards · 6 min

How a model learns to see by reading.

Open
Source

Source · 8 cards · 6 min

Why a model gets smarter the longer it thinks.

Open