Descriptions:
Full post with Agent Maintenance Guide: https://natesnewsletter.substack.com/p//ai-agent-maintenance?r=1z4sm5&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true
OpenAI and Anthropic aren’t the only ones who figured this out: Vercel made its AI agent better by deleting most of its tools. The real skill in 2026 is not building agents, it is maintaining the harness around them as the model improves and your data drifts.
My Links 🔗
👉🏻 Newsletter: https://natesnewsletter.substack.com/
👉🏻 X: https://x.com/natebjones
👉🏻 TikTok: https://www.tiktok.com/@nate.b.jones
👉🏻 Instagram: https://www.instagram.com/nate.b.jones
What’s really happening inside the AI agents companies are racing to build?
The common story is that agents get better when you add more tools, memory, and access, but the real question is whether you can maintain the setup around the model as it keeps changing underneath you.
In this video, I share the inside scoop on why agent maintenance, not agent building, is the real skill of 2026:
– Why Vercel improved its agent by deleting eighty percent of its tools
– How a model getting better can quietly break your agent
– What a harness actually is, and why every AI user already has one
– Where Codex and Claude Code get their real edge
For operators and builders, the upside is real, but an unmaintained agent does not fail loudly, it just keeps working on top of stale truth until it costs you.
Chapters:
00:00 Vercel made its agent better by deleting 80% of its tools
04:34 The model changes inside, the world changes around the agent
10:18 The art of a good harness and the two teams building them
13:30 Your harness is the setup that makes a model useful
18:03 Why you should read The Maintenance of Everything
Listen to this video as a podcast.
– Spotify: https://open.spotify.com/show/0gkFdjd1wptEKJKLu9LbZ4
– Apple Podcasts: https://podcasts.apple.com/us/podcast/ai-news-strategy-daily-with-nate-b-jones/id1877109372
