I'm done with the AI hype

About this video

Chuck says he’s tired of ‘AI-powered’ slop since ChatGPT and asks whether AI is solving real problems for IT and network engineers, or just marketing. Reporting from around Cisco Live, he contrasts Cisco’s approach with Juniper Networks, which invited him to its event. Juniper pitches a self-driving network. He traces Juniper’s AI to Mist Systems, founded in 2014 by former Cisco staff to reinvent enterprise Wi-Fi with cloud and machine learning trained specifically on wireless LAN data — what good and bad networks look like and how to troubleshoot them — long before generative chatbots. Juniper acquired Mist in 2019. The core argument, attributed to Juniper speakers, is that AI quality depends on data and context. Chuck contrasts vendors feeding an LLM telemetry from many separate tools and hoping it figures things out with Juniper’s claim to be ‘AI native’: all wireless, switching, routing, client, application and data-center context fed consistently into the Mist AI cloud, with a microservices architecture, Apstra intent-based networking and a contextual graph database for the data center. On top sit Marvis Minis — digital-experience twins that act like virtual clients, authenticating, getting IPs, using DNS and apps to test client journeys and service-level expectations, especially after config changes — and Marvis AI, with an assistant that translates human questions for it. In his worked example, a CEO’s bad Zoom call could already be correlated in that single store to a switch port with rising CRC errors, yielding a specific recommendation to replace a patch cable, then re-tested by a Mini. Juniper staff also claim trend analysis can predict cable or optic failures before they hit applications. A partner integrator, Allan from Nexum, tells Chuck customers can ask who is having a bad day, drill down, and look back at client state to answer ’last Thursday at 3pm’ complaints. Chuck stresses he has not used it in an outage himself; conceptually he finds the single-context approach interesting and less complex than stitching many legacy sources into an LLM, while asking viewers whether it is real and whether automating network-engineer work worries them.

Key takeaways

• Chuck frames the test for networking AI as real results for engineers — fewer outages and 3am calls — not an ‘AI’ sticker added since ChatGPT.
• Mist, founded in 2014, used specialised machine learning trained only on wireless LAN data; Juniper acquired it in 2019 and built on that head start.
• The central claim is data/context design: one Mist AI cloud holding network state and relationships vs. feeding many separate telemetry sources into a general LLM.
• Marvis Minis act as virtual clients to proactively test authentication, DHCP/DNS, apps and experience, including after configuration changes.
• Chuck has not personally operated the system in an outage — the effectiveness claims come from Juniper and a partner/customer voice in the video, not his own hands-on test.

Episode details: published July 1, 2025 · 20:18 long. Summary written by Self Growth Videos from the episode itself. Business results and any income figures mentioned are NetworkChuck’s own claims, shared for education — they are not typical or guaranteed.

← Previous: got AI anxiety? Do this RIGHT NOW! · Next: You NEED to Use n8n RIGHT NOW!! (Free, Local, Private) →

All NetworkChuck videos · NetworkChuck profile · NetworkChuck on YouTube