What is the primary security concern regarding 'AI sprawl'?

AI sprawl refers to users inadvertently sharing sensitive information, such as usernames, passwords, identity details, and source code, within AI prompts, which can lead to unauthorized access or token misuse.

How will enterprise AI agents change how employees interact with software?

Enterprises are shifting toward agentic AI that handles task-specific duties. Instead of logging into individual software UIs, employees will use a 'conversational harness' to interpret their chat or voice commands to perform tasks.

What is the 'playbook gap' in automation lifecycles?

The playbook gap occurs when the original creator of an automation playbook leaves an organization, leaving the successor to deal with 'drift' and a lack of context, which can cause automation to fail or result in the accidental exposure of private data.

Jay Cuthrell April 12, 2026 WTF

What The Fudge for April 12, 2026: The Two-Day Late Edition

AI Sprawl, Agentic execution, and the state of hardcoded secrets

Subscribe and Listen: https://youtu.be/05HZBFMUUJc

🎙️ Better Late Than Never

Welcome to What the Fudge for April 12, 2026—the two-day late edition. I’m late, and that’s okay. This is a podcast and it’s mine, so I can sort of make the rules. Just catching up on the week in the rear view mirror, this was the week where AI agents and secrets really started to get into the zeitgeist.

🛡️ Security and the “AI Sprawl”

Security came up again—not the usual “AI will make the industry moot” argument, but a realization that people aren’t being as careful as they should be.

Sharing Secrets: We are seeing an “AI sprawl” where users are inadvertently sharing secrets—usernames, passwords, identity, and source code—within AI prompts.
GitGuardian Report: The “State of the Secrets 2026” report found a massive amount of hardcoded secrets leaked publicly on GitHub.
Token Burning: If these are AI-related secrets, someone might “burn your tokens up” using the platform you’ve already paid for.

🤖 From Pilot to Execution: Agentic AI

Enterprises are moving from the “pilot” phase to actual execution with AI agents.

Task-Specific Agents: In the sprawling world of enterprise applications, imagine the things you hate logging into at your day job.
Agentic Interaction: About half of those tasks might soon be handled by something “agentic” on your behalf.
Conversational Harness: Instead of logging into a UI, you’ll work in your favorite “harness” and have a “conversation” with software that attempts to interpret as your chat based or transcribed voice commands.

⚖️ The Leadership Spectrum: “No-No-No” vs. “Oh Fudge”

Following Simon Wardley’s concepts, leadership is currently split into two camps:

Headlong Ambition: C-suite leaders with outsized ambitions about what AI will do.
The Pragmatic “No”: Those in the “no-no-no-no” category until they hit the “Oh Fudge” tipping point.

The risk remains that there isn’t a centralized platform in place to manage the growth of these AI agents.

🔄 Automation Lifecycles and Drift

How survivable is automation without attention to its lifecycle?

The Playbook Gap: Problems arise when the person who created an automation playbook leaves, and the person who inherits it realizes “drift” has occurred.
Context Matters: Computers will do exactly what you ask them to do, but if they lack the relevant knowledge base or context, they may inadvertently share private data in a public repository.

🚀 Space Value Chain: Artemis 2

Rounding out the week, the stories coming out of Artemis 2 are absolutely amazing. Between the great memes and the incredible soundbites, it is delightful to be alive on this little blue sphere while folks on the “high ground” put our existence into a greater context.

That’s a wrap for the What the Fudge Digest for April 12, 2026. Thanks for checking in.

View this page on GitHub.