Nicholas Poon – Medium

Nicholas Poon

Pinned

Published in

The First 1-bit LLM — BitNet 2B

Optimised for < 5% memory usage

Jul 1

The First 1-bit LLM — BitNet 2B

Jul 1

Pinned

Published in

A Breakthrough in Video Generation — Test-Time Training

The math behind Test Time Training

Jun 1

A Breakthrough in Video Generation — Test-Time Training

Jun 1

Pinned

Published in

Generate LoRAs in Under 1 minute.

Text-2-LoRA

Sep 16

Generate LoRAs in Under 1 minute.

Sep 16

Pinned

Published in

Meta Drops Tokenisation in Transformers

Byte Latent Transformers ✨

Feb 7

Meta Drops Tokenisation in Transformers

Feb 7

Pinned

Published in

Is ModernBERT really that good

An explanation you won’t wanna miss.

Mar 10

Is ModernBERT really that good

Mar 10

Published in

Benchmarking LLMs on PhD-Level Problems

The Curie Benchmark

Oct 7

Benchmarking LLMs on PhD-Level Problems

Oct 7

Published in

Self‑Improving AI Meets the Theory of Evolution

The Darwin Gödel Machine

Aug 2

Self‑Improving AI Meets the Theory of Evolution

Aug 2

Published in

Anthropic Fires Back at Apple’s Research Paper

The Illusion of Thinking

Jun 17

Anthropic Fires Back at Apple’s Research Paper

Jun 17

Published in

100 Days after DeepSeek-R1

The DeepSeek aftermath

May 15

100 Days after DeepSeek-R1

May 15

Published in

LLM of the day: The BEST Fully Open-Sourced Model—OLMo 2

A dive into OLMo 2’s technical report

Apr 30

LLM of the day: The BEST Fully Open-Sourced Model—OLMo 2

Apr 30

Nicholas Poon

Nicholas Poon

Your friendly AI Scientist, turning labyrinthine AI knowledge into easy-to-read articles.

Following

Help

Status

About

Careers

Press

Blog

Privacy

Rules

Terms

Text to speech