I talk about my work as founder of @snowbeesoftware, a logistics ERP SaaS. Wrote a book: a.co/d/03A7sT4E. Raising 4 kids.

Joined October 2012
590 Photos and videos
At least I've started noticing when I design for potential future needs. Next step on path to enlightenment is to not actually do it.
53
Norwegian tech X is dull. I was hoping to see founders of Visma, Tripletex, 24sevenoffice and other industry peers active on here. But alas, no such thing.
1
79
We're building an operstional logistics ERP at @snowbeesoftware Place your bets: how many years until we build our own database engine? Seems like a rite of passage for all self respecting B2B enterprise SaaS, especially ERPs
71
Feature flags 💪
1
51
Happy with this AWS cost graph Golive was middle of April, we overscaled everything to be on the safe side Downscaling RDS reserved instances = happy wallet Cloud costs doesn't have to be runaway!
1
62
Has anyone ratioed @HunterBiden yet? Generational run!
44
Should all code have to know about and handle edge cases in all other code? Me: no LLM: yes
1
1
92
"Yes son, when grandpa was young, the dark side of the moon was completely black at night."
1
98
Feels like our (expensive) db server on RDS
Japanese electronics giant Toshiba introduces a mini-notebook PC, the “Libretto L1,” equipped with a 600-MHz Crusoe processor on its CPU and a 10-inch 1,2800 x 600 pixels LCD display. The retail price is ¥150,000 ($1,480).
114
CloudWatch alarms saved our asses today. If our database is using over 80% CPU 8 of 10 times when polled every 60 secs, we get an alarm. Got an alarm today after pushing a borked query. Fix out in 3 minutes.
95
We're adding "after prod" validations. Ideally we'd catch bugs before prod. Or not make bugs at all! But catching bugs like "customer search just became super slow" can only be done in prod. Also, we have a lobster AI that knows our codebase and comments on commits to main.
1
64
Currently "training" an openclaw on our repo. I've asked it to analyze our code and quiz me relentlessly. This ends up being ultra low effort, with ultra high quality output. Recommended! Also made it add an "index" that we link to from AGENTS.md.
46
I've been working in Windows for the last month or so. Poor Omarchy, dormant on a separate drive. I occasionally need Windows because of our POS app. So I just end up staying there. I've never really felt "oh this OS makes me so much more productive" for any setup.
1
80
LLMs loves to add dumb tests. gpt-5.5 writes lots of great tests, in my limited experience, by looking at our 2 year corpus of manually written tests. But it also loves to add tests for insignificant things, like "when we call the function, do we get the hard coded label" etc.
82
Setting up a new CI pipeline is always like this
57
I always imagined small green Yoda was a sign of his old age. Disappointing that Disney made Yodas into a species that just looks that way.
43
Did you know (DB edition): triggers can be "after statement". If you have "after insert", and insert 100 rows, the trigger runs 100 times. With "after statement", it runs once.
36
Good: realize 2 characters makes LIKE search slow, so disable it Bad: realize core data in your system has 2 characters and people want to search for it Did not expect that to be a breaking change..
50
Good: measure P95 on a hot spot and be satisfied that it's 500ms Bad: forget to set up webhook so polling makes it take 5 seconds anyway.
47
Super cool agent skill: readonly connection to production db. Use 1password - secure, and you get a popup when it wants to conect. Super efficient way to debug an issue - just paste a stack trace, screenshot, etc, and get perfect feedback in mere minutes.
70