AI at Work, Fiber at Home and Mountain E-Bikes with Mike Wieger – HGG681

Posted on by

What happens when AI leaves the demo stage and starts showing up in real work, real home labs, and the gear that keeps a household connected? This week, Mike Wieger returns to Home Gadget Geeks for a practical conversation that moves from AI in a regulated M&A workplace to OpenClaw at home, new symmetrical fiber, mountain e-bikes, and the growing MeshCore/Meshtastic community around Omaha and Lincoln.

Listen and Watch

Audio Only
Apple PodcastsGoogle PodcastsPlayer EmbedShareLeave a ReviewListen in a New WindowDownloadSoundCloudStitcherSubscribe on AndroidSubscribe via RSSSpotifyAudioboomSpreakerOther Options

Episode summary

Mike and I cover a lot of ground: AI tools at work and at home, his move to fiber internet, a detour into electric mountain bikes, and a solid update on the MeshCore/Meshtastic mesh networking scene around Omaha and Lincoln. If you’ve been following the OpenClaw series here on the site, this episode adds a different angle — Mike’s experience managing AI adoption inside a regulated M&A workplace, which is a perspective I don’t get to share much.

We opened during College World Series week — with unusually good weather for once — then got into it.

Mike has been experimenting with OpenClaw at home, partly because AI tools weren’t yet available in his regulated industry at work. His early use cases were calendaring, watching for property listings and managing Home Assistant — useful, but he eventually gave OpenClaw enough access to his home lab that it started to feel uncomfortable, and he shut it down. I came at the same problem from the opposite direction, locking things down so tight the system could barely do anything without asking. Different paths, same conclusion: useful AI automation needs real boundaries.

Mike also pointed out that his household’s actual AI power user might not be him — his wife, Hannah, has been using ChatGPT for birthday invitations and PTO flyers, which is a good reminder that AI adoption often starts with whatever someone actually needs done today, not with code or spreadsheets.

The most useful part of the conversation, for me, was Mike’s management perspective on AI in a regulated M&A environment. His team builds financial models and reviews data, and AI can speed parts of that up — but Mike is clear that analysts still need to understand the model underneath. The goal isn’t to make someone look smarter than they are. As Mike put it, it should make it look like there are four of you, not fake expertise. That shift changed how he thought about a hiring decision: the work didn’t go away, but more time opened up for review, client conversations and the soft-skill development that’s harder to teach.

I shared a similar story from the other side — a long-delayed, genuinely messy data project with inconsistent files, bad headers, duplicates and weighted scoring. Claude didn’t just process the data; it questioned my assumptions and forced me to think more structurally about the problem. That’s the part of AI use that is easy to undersell: the value isn’t always the output, it’s getting asked better questions along the way.

We also talked through the practical side of AI cost and complexity — what happens if today’s $20 AI plans become $100 plans, and how local models, frontier models, agents and skills all factor into a decision that’s getting less obvious over time, not more.

Fiber finally arrives

Mike made the jump from Cox to Metronet/T-Mobile fiber, going from roughly $160/month for 1 gigabit down and 100 megabit up to about $70/month for 2 gigabit symmetrical — with a technically friendly install team that actually understood pfSense, static IPs and self-hosting.

That symmetrical upload changes the math on a lot of home lab projects: Tailscale exit nodes, self-hosted services, offsite backups and even the old “buddy backup” idea, now that uploading large amounts of data is actually practical. Mike already has a small TrueNAS box at his in-laws’ place receiving backups.

E-bikes and 3D-printed accessories

Family camping trips and state park rides got Mike into electric mountain bikes — pedal-assist, no-throttle, mid-drive systems that keep mountain bike geometry but add weight and smart assistance. We talked through the safety side too: kids on fast e-bikes, class 1/2/3 rules, and how a few reckless riders can get restrictions put in place for everyone.

Mike has also been 3D-printing bike accessories — mud guards, hand guards, water bottle holders and hidden AirTag holders — with ASA filament holding up noticeably better outdoors than PLA.

MeshCore and the Omaha/Lincoln mesh community

We closed with an update on Meshtastic and MeshCore. Meshtastic was the original firmware a lot of people started with, but MeshCore has gained traction locally because its repeater/client structure works better for an organized, city-scale mesh network — and Omaha and Lincoln have grown enough that reliable mesh messaging across the metro is becoming realistic.

That comes with real coordination work: node placement, repeaters, public channels and the usual open-source growing pains — Discord debates, competing admin groups and disagreements about who gets to decide direction. But there’s a good side too: a community mesh analyzer project, shared observer data and an in-person meetup that Mike said felt a lot like a ham radio gathering.

We wrapped up talking about robot lawnmowers, a possible ZFS migration for Mike’s storage setup and a couple of AI-assisted home lab wins — including Chia replotting as a candidate for a bounded, guardrailed AI task, and OpenClaw helping solve a Bluetti/EcoFlow Home Assistant integration faster than expected.

Key takeaways

  • AI at work isn’t just a tooling question — it changes training, hiring and what the human role becomes.
  • AI works best when it increases capacity and judgment, not when it stands in for understanding.
  • OpenClaw and Hermes are exciting precisely because — and despite — how much they show that permissions and guardrails matter.
  • Symmetrical fiber changes what’s practical for self-hosting, exit nodes and offsite backup.
  • MeshCore is gaining ground locally for organized mesh networks, but the community and governance layer is part of the system, not separate from it.

Chapters

  • [0:29] – AI, Work, and Family Shifts
  • [4:59] – Why OpenClaw Hooked Him
  • [6:11] – AI Changes Team Roles
  • [12:11] – Claude Solves Data Chaos
  • [17:22] – Orchestrators and Local Control
  • [22:53] – Bubble Talk and AI Costs
  • [29:36] – Fiber Finally Arrives
  • [34:43] – Faster Speeds, Bigger Ideas
  • [39:34] – E-Bike Curiosity
  • [47:51] – AI and Local Models
  • [57:13] – MeshCore Community Dynamics
  • [1:07:17] – Robot Lawn Dreams

Resources mentioned

Full show notes, audio, and video at: https://theaverageguy.tv/hgg681

Join Jim Collison / @jcollison for show #681 of Home Gadget Geeks, brought to you by the Average Guy Network.

WANT TO SUBSCRIBE? http://theAverageGuy.tv/subscribe

Join us for the show live each Thursday at 8pmC/9E at http://theAverageGuy.tv/live