Riley Goodside
Also known as goodside, prompt engineering
Riley Goodside is a public practitioner of prompt engineering. He worked in data science before publishing GPT-3 prompt examples and screenshots in 2022. In August 2023, Goodside was a staff prompt engineer at Scale, where he tested and integrated language models. His technical talk covered context limits and the changes caused by instruction tuning and reinforcement learning from human feedback. In 2024, he co-authored research on human attacks across multiple conversation turns. In an Interconnects interview, he discussed prompting, reasoning models, post-training and evaluation.
Timeline
-
Before public prompting
A data science background
Goodside worked in data science at OkCupid, Grindr and CopyAI. After Grindr, he took time to learn about large language models.
-
2022
Publishing prompt demonstrations
Goodside began posting GPT-3 examples and screenshots in 2022. The demonstrations made model behaviour visible through individual cases. They gave readers a practical way to inspect how wording and supplied context can change an answer.
-
By August 2023
Prompt engineering at Scale
Goodside was a staff prompt engineer at Scale. His work also included comparing Claude and ChatGPT.He tested and integrated language models and used public examples as a major part of his teaching practice.
-
August 2023
Explaining how prompting changed
His Scale talk covered prompting across pretrained models, instruction tuning, and reinforcement learning from human feedback. The programme also discussed limited context windows and longer-context systems. The event connects his examples with a broader explanation of how different model training stages affect interaction.
-
August–September 2024
Testing multi-turn model defences
Goodside co-authored LLM Defenses Are Not Robust to Multi-Turn Human Jailbreaks Yet. The paper examined attacks carried out by people across a conversation. The team released a dataset and discussed why tests based on a single automated exchange can miss weaknesses in the studied systems.
-
September 2024 onward
Discussing evaluation and model changes
In an Interconnects interview, Goodside discussed prompting, post-training, reasoning models, and evaluation. He stressed hands-on experience and the effects of prompt choice on measured results.
Sources and credits
Photo: Riley Goodside / GitHub profile. License: Not stated by source. Source: github.com. Cropped to a square and resized. Ask us to remove or replace this photo.