Sign In

Michael DeHaan

Fractional C3PO

Michael DeHaan is a versatile professional with a background in technical marketing. He holds a BS in Computer Science from NC State University. Michael has been involved in various organizations and roles, ranging from teaching staff at North Carolina State University to founding and serving as the CTO at Ansible, Inc. He has also held positions at rPath, WebAssign, Puppet Labs, Inc., Red Hat, Integrian, and IBM (Adaptec). Currently, he operates as a consultant at Michael DeHaan LLC, offering his expertise in marketing technical products. Michael is also the creator of a tech-focused newsletter available on his website.

Highlights

Jul 25 · linkedin

It's been over 10 years since I created Ansible. I've decided the world needs a new IT automation system designed for 2023 and beyond, more so, I'm itching to build something in the world again. What am I thinking exactly? Rust. Planetary Scale. Performance. Audit Features. A Hard Engineering Focus. A Return To Minimalism. Easy to learn, highly compatible, not a port, but a spiritual successor. I've already started building this. It's doable and it's going to be impressive.

https://lnkd.in/g3Ac2BZt

quick update: Discord chat and announcements/blog info added at https://www.jetporch.com !

Jul 11 · linkedin

PSA in case people need to hear it: if you need AI to configure (or code, I guess) your computers or find this appealing - then you are deploying things you do not understand and possibly deploying insecure solutions or software without complete backup configurations.

If you don't have time to configure something yourself and then automate it, without copy/pasting solutions, you should not be in DevOps and should perhaps focus on another career, including programming, rather than trying to cut corners. You should never deploy something you don't know how to use and administer manually first. You must first think in operations terms, and then automate, only after you understand. Thus, most automation should be written completely from scratch, making deliberate conscious decisions.

As such, most of the modern view of Agile (2018-present?) that devs can also do ops misses the point. The original view was for ops to automate themselves, and just give devs better understanding of prod, and prod better relations of dev. The current situation puts dev in control of prod, but also with little understanding of what they are doing, and little semblance of long term planning, consistency, or architecture.

Using templates out of other people's sample configurations is also unwise, start from the heavily commented software configuration files that came with the installed software, customize that as you need. You never know what the template is leaving out or what it is including that is extraneous.

Until AI sources can fully annotate what material they are trained on, we will never know the errors they make. Until they can explain their understanding (and not just hallucinate understanding), they are no better than copy/pasting off Stack Overflow. Expert systems can understand solutions, items like ChatGPT are only an illusion of actual understanding. They are clever algorithms and very interesting, but we should never trust the safety of our computer systems to them. People may say they can provide a jumpstart, but something that is 99% correct is more dangerous than something that is half wrong, because you will not spot the error in the 1% and will assume correctness.

People who think "AI" is intelligence, especially, are those who should not use it, and do not understand what a machine understanding a concept really means, and how we are not even close to that point with computing.

Michael DeHaan
Add to my network

Location

Cary, North Carolina, United States