NA010 - The Future of Platform Engineering with Stuart Clark

Network Auto Magic Podcast  

The Future of Platform Engineering with Stuart Clark

Episode Overview  

Stuart Clark left school, spent eighteen years cutting hair, then taught himself network automation and ripped through CCNA, CCNP, CCIP, and Juniper certs in three years. That career launch took him to Cisco for nearly eleven years, including 4.5 years on DevNet as one of the first five DevNet Experts (now CCIE Automation), then to AWS, and now to Senior Developer Advocate at Spotify. In this episode Steinn and Urs interrogate Stuart on what actually changed at Cisco DevNet after Susie Wee and Mandy Whaley left, why DevRel keeps getting cut first, what Spotify’s Backstage and the new Portal SaaS offering actually solve, why an internal developer portal becomes a “voice to chaos” the moment AI agents start committing code, and the gap between AI hype and AI in production. Plus the standout Ford RS Turbo metaphor for organizations bolting on horsepower without ever upgrading the brakes — and Stuart’s closing hot take on not letting your job define you.

Episode Guest  

• Stuart Clark: Senior Developer Advocate at Spotify — previously AWS, Cisco DevNet (one of the first five DevNet Experts / CCIE Automation), with 11 years at Cisco preceded by an unlikely pivot from 18 years of hairdressing into self-taught network automation. Eight years across DevRel covering Cisco, AWS, and Spotify.
  • LinkedIn: Stuart Clark

Listen to the show on YouTube:  

Listen to the show anywhere:  

• YouTube: @networkautomagic
• Spotify: Network AutoMagic
• Apple Podcasts: Network AutoMagic
• RSS Feed: Anchor.fm

Show notes resources:  

• Backstage — The open source IDP framework that Spotify built and donated to the CNCF
• Spotify Portal — Spotify’s hosted SaaS offering for teams that don’t want to operate Backstage themselves
• Site Reliability Engineering (Google) — The book Stuart credits with shaping how he thinks about building production systems, even from a non-software-engineer background

What we cover:  

From Hairdresser to CCIE Automation  

• The Career Pivot Nobody Plans: Stuart left school, started cutting hair, did it for eighteen years, then completely retrained — moving into NOC work, then a three-month Cisco contract that turned into eleven years
• The Cert Sprint: CCNA, CCNA Security, CCNP, CCIP, plus Juniper certs in three years — entirely self-taught, pre-DevNet, pre-AI
• Onto DevNet: Joined DevNet at the start of 2017 and stayed 4.5 years, becoming one of the first five DevNet Experts (now CCIE Automation)
• Cisco → AWS → Spotify: The cloud pivot, then the platform-engineering pivot — driven by curiosity, not a career plan

What Happened to Cisco DevNet  

• The Riding-High Years: 550,000 community members, certifications launching, Chuck Robbins on the Cisco Live San Diego stage saying DevNet is the future of Cisco
• Leadership Exit: Mandy Whaley left first, then Susie Wee. The original founding voices were gone within months of each other
• Messaging Whiplash: The pivot — from “the future is cloud” to “the future is applications” to “DevOps and applications” to “DevOps, applications, and SRE” — left the community confused about who DevNet was actually for
• Org Reshuffles: DevNet got moved out of Learning & Certification, into emerging tech / Outshift, and the alignment between the certification track and the DevNet team broke down

AI as Productivity Tool, Not Marketing Slogan  

• Working Faster, Not Harder: Stuart’s core take — he’s working less, not more, despite turning out work faster. AI assistants pair with strict boundaries to keep him out of burnout
• 221 Broken Links to Zero in an Hour: A real Spotify docs migration win — Claude Code knocked 221 broken links down to 59, then 29, then 30 (don’t ask), then 0, in roughly an hour
• Faster Than Scripting It: Stuart’s point — even writing a custom shell or Python script for the link fix would have taken a full day. Doing it manually would still be ongoing
• Human Review Still Found Three: After the script said zero, a senior reviewer caught three more nuanced issues — proof that AI accelerates the volume work and humans handle the judgment calls

The AI Mania Trap  

• “Why the F Are You Doing AI From the Gym?”: Stuart’s pushback against the always-on AI culture — people bragging about prompting from the gym, on dog walks, while on holiday
• The Two-Phone Rule: Stuart used to run two phones to keep work apps off his personal one. Now he just signs out of Slack/WebEx/everything when work ends
• Steinn’s Hot Tub Confession: Self-aware moment — iPad in the hot tub, cigar lit, twelve cloud instances running, scrolling through output convinced everyone would be impressed Monday. Spoiler: they were not
• Boundaries Are the Productivity Tool: The unspoken truth — most “AI productivity gains” disappear if you let them eat your evenings and weekends

DevRel Pipeline, Not Vanity Metrics  

• Vanity vs Revenue: Likes, stars, shares, clicks — DevRel built on those metrics is the first to get cut when budgets tighten, because it can’t defend itself with a revenue number
• The Last-Mile Visibility Problem: Most DevRel teams can track top-of-funnel (people showing up to webinars, blog readers) but lose visibility once those leads enter the sales pipeline. Tracking that last mile is the job
• Office Hours as Attribution: Spotify runs bi-weekly office hours — UTM-tracked links from blogs and LinkedIn posts let DevRel measure exactly which content drove someone toward becoming a customer
• You’re Part of the Team: Stuart’s framing — you’re not solely responsible for closing a deal, but you’re a real part of the customer journey, and that’s what gets the headcount renewed

Backstage and the IDP Story  

• The Onboarding Problem at Spotify (2014): New hires were taking six weeks to land their tenth pull request — Stuart calls that “archaeological excavation,” not onboarding
• Backstage as the Single Pane of Glass: Built internally to give engineers full lifecycle visibility — every commit, every deployment, ownership maps, golden paths — and shipped externally via the CNCF
• Portal — Backstage as a Service: Spotify’s hosted SaaS for teams that don’t want three React engineers babysitting a Backstage instance — the same “we built it, we’ll show you the way” pattern as Kubernetes coming out of Google
• Plug-and-Play Plugin Ecosystem: AWS, Netbox, Cisco SD-WAN, anything with an API — Stuart built a personal Backstage during onboarding and integrated DevNet sandboxes, vManage, and AWS templates as a learning exercise

IDPs as Voice to Chaos in the AI Era  

• Five Engineers, Three Agents Each: Suddenly you have a team of thirty checking in code. Visibility, ownership, version compliance, naming conventions — the abstraction layer becomes load-bearing
• Skills Files as Accelerated Golden Paths: The next evolution beyond manual templates — encoded standards that AI agents respect when generating code
• CI/CD Doesn’t Replace Standards: Pipeline failures cost time and context. A skills file that respects pipeline rules upfront catches issues before they become commit-then-fail loops
• Trust Is Still the Bottleneck: The 99/2 reality — at conferences, 99% of engineers raise their hand for using AI personally; only 2% say it runs in production. The same trust gap that held back early network automation

The Ford RS Turbo Metaphor  

• More Horsepower, Same Brakes: Ford bolted a turbo onto the Escort and went from 80 to nearly 200 brake horsepower — but never upgraded the braking system
• The Platform Engineering Lesson: Companies bolt AI productivity onto existing engineering orgs without investing in the foundational pieces — observability, governance, security, standards — that act as the brakes
• One Security Incident from Lockdown: Steinn’s contrast — extremes go both ways. Race forward without brakes and you crash; one AI security incident and management locks everything down

Closing Wisdom  

• Don’t Be Defined by Your Job: Stuart’s hot take — he fell into the trap where his developer-advocate persona swallowed who he was as a person. Setting boundaries between professional and personal isn’t optional
• CCIE Fundamentals Still Beat Shiny Things: Locking in 9-14 months for a fast-moving cert track is risky. A traditional CCIE — knowing OSPF, BGP, IS-IS deeply enough to call out bad design — ages well
• You Can’t Skip Python: Stuart interviewed at major AI companies after AWS. They didn’t let him use AI in the interviews — they made him write Python by hand. The lesson: when you take AI out of the loop, you find out how much you actually know
• AI as Today’s Scapegoat: Layoffs blamed on AI when the real driver is recession optics. The honest read

Key Takeaways  

1. Internal Developer Portals Solve Onboarding First, Then Everything Else: Backstage didn’t start as a productivity tool — it started because Spotify hires were taking six weeks to ship their tenth PR. The platform-engineering use cases came after
2. AI Without Standards Just Multiplies the Mess: Five engineers with three agents each is a thirty-person team checking in code. Without governance, that’s chaos with extra steps
3. DevRel Survives by Building a Pipeline: Vanity metrics get you cut in the first round of layoffs. Tying content to revenue keeps the team funded
4. Working Faster Doesn’t Mean Working Longer: The boundary discipline matters more than the AI assistant. Stuart’s working less and producing more
5. Trust, Not Tooling, Gates AI Production Adoption: The 99/2 conference-room stat. The hard sell isn’t the technology — it’s convincing leadership to let it loose
6. Race Forward, Upgrade the Brakes: Foundations and standards are not optional. Bolt a turbo on a car with stock brakes and the only question is when you crash
7. Your Job Is Not Your Identity: Set the guardrails for yourself the same way you’d set them in a system