2084

Trajectory-Conditioned SFT for writing Reddit Posts really well.

Or; How To Train Your Models

Sep 19, 2025 • Lukas Nel

2

1

2084: Object Oriented LLMs: Better Tool Calling in LLMs and building Powerpoints with AI

How I learned to stop worrying and let AI build PowerPoints

Sep 10, 2025

1

July 2025

2084: The soul of the LLMs

All the LLMs are converging to the same representation

Jul 1, 2025 • Lukas Nel

3

2

June 2025

Deep Reasoning With Tools: Toolcalling in the reasoning trace

A framework I've been working on for the past while.

Jun 1, 2025 • Lukas Nel

3

3

May 2025

April 2025

2084: DeepStock V2 - Predicting the market better than 50%

Training AI to Predict the Market with GPRO

Apr 17, 2025 • Lukas Nel

5

6

February 2025

2084: Diabetes Seek

Predicting diabetes with DeepSeek from a patient's medical history.

Feb 25, 2025 • Lukas Nel

4

1

1

2084: A Dive into DeepSeek

While deepstock is training, an intro on DeepSeek V3

Feb 11, 2025

6

2

2084: Deepstock - can you train deepseek to do stock trading?

A short guide to using deepseek to do stocktrading.

Feb 3, 2025 • Lukas Nel

5

26

January 2025

The AI Shift: What It Means for India's Outsourcing Empire

Exploring the capabilities of agentic AI and the effect this will have on India's current outsourcing empire

Jan 27, 2025 • ShailenDesai

Bureaucracy is all you need.

A manifesto for paying more attention to how AI agents communicate.

Jan 19, 2025 • Lukas Nel

2

1

June 2024

2084: BitNet overhyped? Training BitNet Mamba on the Tiny Shakespeare dataset

Applying single bits to solve my problems.

Jun 4, 2024 • Lukas Nel

2