Season 1 · Episode 72 · September 30, 2026 · 1h 5m
Has Claude’s Output Got Worse? Here’s What You Can Control | Re-release with Rob Webster
Show Notes
Claude’s output has felt harder to work with recently. It ignores instructions, produces overlong, overwritten answers and can get worse over a long conversation. But if you want to improve your AI output, the answer might just be sitting in your Documents folder.
In this re-released episode, Jess and Kyle speak with Rob Webster, founder of TAU Marketing Solutions, about managing AI context. The tools and product names may be moving fast, but the practical problem has not changed: AI performs better when it has clear, well-managed context.
In this episode, Rob explains how to set up a few plain text files and folders so your AI knows who you are, what you do and what good looks like before you type a word. He also explains why giving your AI everything at once tends to make answers worse, how to spot when your setup has got bloated, and why the same files are the foundation for Claude Code and AI agents.
Follow Rob:
- LinkedIn: https://www.linkedin.com/in/marketingaivisionary/
- TAU Marketing Solutions: https://taums.ai/
What You'll Learn
- Why AI forgets instructions in long conversations, and how better context can improve output
- The difference between persistent and situational context
- How to organise plain-text and Markdown context files
- Why more connected documents do not always produce better AI answers
- What to retain permanently and what to remove when a project ends
- How to identify context drift early
- Why this approach provides a foundation for Claude Code, Claude Cowork and AI agents
Timestamps:
00:43 Introduction and Weekly Updates
03:09 Claude’s Output Has Got Worse, So We’re Revisiting Our Setups
11:12 Markdown Files and the Two Types of AI Context
14:24 Reintroducing Rob Webster
16:31 The Goldfish with a PhD: Why AI Forgets Everything
20:17 What You're Losing Every Time You Open Claude Without Context
26:11 Context Drift: Why AI Outputs Get Worse the More You Use It
29:26 How to Start Giving Your AI Context About Who You Are
35:35 Making It Stick: Naming Chats, Pinning Projects, Starting Your Folders
39:02 What a Working Folder Structure Actually Looks Like
40:38 Where Your Files Should Live
50:14 Global vs Project-Specific: The About Me Folder and Beyond
57:21 How AI Reads Context
59:47 How This Scales Into Agents, Claude Code, and Vibe Coding
Resources:
- https://tech-insider.org/anthropic-engineer-claude-writing-quality-worse-2026/
- https://news.ycombinator.com/item?id=49348163
Get in touch with Early Adoptr: hello@earlyadoptr.ai
Follow Us on Socials & Resources:
IG: https://instagram.com/early_adoptr
TikTok: https://tiktok.com/@early_adoptr
YouTube: https://www.youtube.com/@early_adoptr
Substack: https://substack.com/@earlyadoptrpod
Hosted on Acast. See acast.com/privacy for more information.



