Guides

How to work with AI, from the evidence. Each guide starts from published research — mostly other people's — reports what it actually found, and turns that into something you can do differently on Monday.

4 guides · 8 sources cited

What every guide here does

Links the source and says when it was retrieved.Read the original; a guide is a reading of it, not a replacement.
Reports only what the source says. Every claim was checked against the paper itself, not recalled from it — citing a study is not the same as reading one.
Separates the dated from the durable. A percentage measured on a 2023 model is not a description of today's. Each guide says which of its findings depend on the models tested and which do not.
Says where it is our view. What follows from the evidence is our reasoning, marked as a position, and you may draw a different conclusion from the same papers.

The guides

Asking an AI to check its own work is not a second opinion

Research finds that asking a model to review its answer, or whether it is sure, can turn right answers wrong — and that correction works when something external can be checked.

General Chat · ReadingsLeon Mallett, Founder of Captivated Ltd with Claude Code

Last confirmed working 20 October 2026 on Jie Huang, Xinyun Chen, Swaroop Mishra, Huaixiu Steven Zheng, Adams Wei Yu, Xinying Song, Denny Zhou, Mrinank Sharma, Meg Tong, Tomasz Korbak, David Duvenaud, Amanda Askell, Samuel R. Bowman, Newton Cheng, Esin Durmus, Zac Hatfield-Dodds, Scott R. Johnston and others, Ryo Kamoi, Yusen Zhang, Nan Zhang, Jiawei Han, Rui Zhang

Telling an AI it is an expert does not make it more accurate

A study of 162 personas found none made models reliably more accurate on factual questions. A role shapes tone and focus; accuracy comes from what you hand over, not what you call the model.

General Chat · ReadingsLeon Mallett, Founder of Captivated Ltd with Claude Code

Last confirmed working 20 October 2026 on Mingqian Zheng, Jiaxin Pei, Lajanugen Logeswaran, Moontae Lee, David Jurgens, Anthropic

Where something sits in a long prompt changes whether it gets used

Models used information at the start and end of long inputs far better than in the middle — at worst, below having no documents. Put material first, the question last, and cut what is not needed.

General Chat · ReadingsLeon Mallett, Founder of Captivated Ltd with Claude Code

Last confirmed working 20 October 2026 on Nelson F. Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, Percy Liang, Anthropic

The strongest evidence against this library, read carefully

A controlled study finds context files do not improve task success and cost 20% more — and its detailed findings are closer to this library's own rule than its headline suggests.

Coding · ReadingsLeon Mallett, Founder of Captivated Ltd with Claude Code

Last confirmed working 14 September 2026 on Thibaud Gloaguen, Niels Mündler, Mark Müller, Veselin Raychev, Martin Vechev