More Beyond Coding episodes

The Most Future-Proof Engineering Skill Nobody Wants to Learn thumbnail

The Most Future-Proof Engineering Skill Nobody Wants to Learn

Published 4 Nov 2026

Duration: 00:38:04

Rapid AI exploit times demand deterministic security guardrails, while engineers must adapt to new roles and highlight human traits in hiring.

Episode Description

What is the most future-proof engineering skill to learn now with AI and Coding agents? Time is up for traditional development cycles as Agents shrink...

Overview

The conversation covers how AI agents are accelerating cybersecurity threats and transforming engineering work. Anna MacDougall, Field CTO at IBM HashiCorp, discusses how exploit times have dropped from roughly 200 days to about 18 hours, making traditional response timelines obsolete. She emphasizes that skills files or instructions.md are insufficient guardrails for agentic systems, and that deterministic, policy-scoped credentials and agent identity are essential.

The discussion also explores the emotional and career impact of AI on engineers, including grief over the loss of hands-on coding, existential dread about role survival, and the need for empathy in how the industry treats those struggling to adapt. MacDougall advocates for security as a future-proof career path, noting that 88% of security exploits involve credentials and that agentic runtime security is an emerging, high-value domain.

Career advice centers on networking as connecting people, finding a memorable "pumpkin" differentiator in your CV, and focusing on achievements that signal initiative and depth beyond generic job descriptions.

What If

  • What if you treated your solo development environment as an agentic system requiring identity and scoped credentials?

    • Move: Propose implementing policy-scoped, just-in-time credentials for any automation or agent tooling you run locally, so that "credentials injected at runtime and scoped" apply even to your own workflows.
    • Why Now? Exploit windows have compressed to "something like 18 hours", and "88% of security exploits have to do with credentials", meaning a leaked long-lived token in your solo stack is a realistic breach vector.
    • Expected Upside: You would reduce your personal attack surface and build habits aligned with the "agentic runtime security" practices MacDougall identifies as increasingly relevant, positioning you as a practitioner rather than just a consumer of these tools.
  • What if you built a security-focused side project demonstrating you can find and fix vulnerabilities in your own systems?

    • Move: Propose auditing one of your existing projects for credential exposure, dependency vulnerabilities, and missing authorization checks, then documenting the fixes as a public write-up.
    • Why Now? MacDougall notes that "trying to find those vulnerabilities and essentially hack your own systems" is "more fun" and more interesting than just patching, and that closing a security gap is "hugely valuable to a company" and a strong promotion or job-change signal.
    • Expected Upside: This could give you "highly visible work that has highly provable value" to showcase, differentiating you in a market where "so many brilliant engineers want nothing to do with" security.
  • What if you added a "pumpkin" story to your professional profile that signals a work-relevant trait?

    • Move: Propose identifying one personal achievement (a completed distance, a language learned through a hobby, a certification earned outside work) and framing it in your CV or LinkedIn as evidence of depth, commitment, or initiative.
    • Why Now? MacDougall explains that hiring managers look for "signals that this is someone who takes an extra step" and that a specific achievement like "you've run a half marathon" or "you're a yoga instructor" tells a story that generic hobbies do not.
    • Expected Upside: This could make your profile memorable in a stack of AI-tailored resumes and give an interviewer a conversation hook that reveals your work ethic, as MacDougall describes with the anime-to-Japanese example.

Takeaway

  • Replace long-lived credentials in any agent or automation workflow with just-in-time, policy-scoped tokens that expire after one use, so a leak is already invalid by the time it matters.
  • Implement deterministic, binary guardrails (allowed versus not allowed) for any system an agent can access, rather than relying on context files like instructions.md, which an agent "can ignore."
  • Direct a portion of your learning toward security - specifically agentic runtime security, credential management, and vulnerability discovery - because MacDougall calls it "one of the most future proof parts of engineering right now."
  • Network by connecting two other people who need each other, rather than asking for a job; being "the person who connects other people" builds reciprocal goodwill across your career.
  • Add one specific, verifiable personal achievement to your CV or profile (a completed distance, a language learned, a certification) that signals commitment and depth, instead of listing generic hobbies.

Recent Episodes of Beyond Coding

30 Sept 2026 Why Coding Agents Keep Making Your Codebase Worse

"Explores McDonald's World Menu Heist, AI's coding limitations (like poor long-term maintenance and refactoring), critiques unrealistic AI hype, and emphasizes engineering best practices, human oversight, and incremental productivity improvements."

19 Aug 2026 How Amazon Turns Real Failures Into Better AI Models

AI's evolution and challenges are explored through agentic systems like Amazon's Nova, model routing inefficiencies, simulated learning, enterprise tool integration, dynamic benchmarks, and the balance between specialized and generalized models.

More Beyond Coding episodes