Profile avatar
carlfinmo.bsky.social
Studied physics, works with software, ham radio fan. 🇸🇪 Works at Airpelago - Automated power line inspections with drones.
37 posts 18 followers 89 following
Regular Contributor
Active Commenter
Posts 13 Comments 24

Are there any tools that takes a diff, a coverage report and some instrumentation you can run in your staging / acceptance test environment and then tell you what code paths have not been touched?

submitted 64 days ago • 0 comments

Is this vibe coding? I always wanted to make a augmented reality Kubb game, where you make the throw with your phone. Maybe with Claude et al I will. At any rate coding with these tools is really fun since you can get started and play around with an idea that would otherwise not happen.

submitted 75 days ago • 2 comments

Two bad traits we have to keep in check as engineers: - Overemphasis on a solution: e.g your favourite technology - Overemphasis on requirements: e.g from your specialty field I really like the mental model of solution space and problem space. I liked solutionspace.blog/2021/10/29/p...

submitted 89 days ago • 1 comment

I’ve been trying to use @obsidian.md for a while. Here is my process: 1. Write a note every day 2. Go through the new notes weekly and refine 3. Explore your notes monthly No special results but brain feels less clogged. I like letting ideas go and letting them resurface later.

submitted 89 days ago • 0 comments

Good start of my day. Thanks for all the hard work #neovim developers!

submitted 91 days ago • 0 comments

Does the LLM shops need to filter the generated content from other LLMs, that just seems to increase everywhere, before training new models? How to determine that the source material is high quality? Do they have the best ‘generated by ai detectors’?

submitted 92 days ago • 0 comments

Guessing that this will lead to poisoning the data lake tactics. And any rate with so much generated content nowadays maybe they have already trained on the cleanest possible data?

submitted 92 days ago • 0 comments

Wrote some code to read a well know binary format this week and the chatbots I used probably expedited the work compare to rtfm. But they also made fundamental mistakes with logic conditions and number of bytes to read etc. Is it a fundamental limitation of current architectures? (E.g tokenisation)

submitted 103 days ago • 0 comments

Is cyber security services&products more of an insurance than a tech? Is the goal of csec work is not primarily to increase security, but to show you did what you have to and should not be sued? Similar to how you lower your home or car insurance by having the right lock or device.

submitted 110 days ago • 0 comments

Normal programs are brittle on bad input unless we take a lot of caution. Now we have fitted models to be ‘helpful’ on bad input. Similar to how you eventually can’t fix bugs without introducing new ones I’m guessing you can’t train away protection from bad input?

submitted 110 days ago • 1 comment

Postgres bulk insert and get IDs for further processing: COPY to temp table UPDATE temp set db_id = nextval(real_tbl_id_seq) INSERT into real table SELECT prel_id, db_id from temp for further processing Any better ideas?

submitted 117 days ago • 0 comments

Knocked a couple of % off someone’s AWS bill today by telling them about docs.aws.amazon.com/vpc/latest/p... If you have a large line in your bill under Nat gateway and you download data from S3 to a VPC this might help.

submitted 118 days ago • 0 comments

When making software you’re always just a one line change away from a catastrophe

submitted 236 days ago • 1 comment