Norwegian Coder

Norwegian Coder

@dnaleetos.bsky.social

DIY | IT | Selfhosting | Homelab | Coding | React | dotnet | PHP | CSS | AI | LEGO | Support #Wikipedia | Let's discuss! | I'm I installed 3.11 with floppys years old.

99 followers156 following195 posts

What are you using AI for that you never believed could be done with AI? I'm using it to guide me to NIS2, building a subwoofer, managing #unraid, handling personal CRM and missed a Toro-pack for Béarnaise last night so Claude guided me to something even better using what I had on hand!

Just the tip of the iceberg. It only does things i could have done myself, give time. It just does it faster and mostly better. Mostly much better. Happy for now, afraid for the future (of humanity).

I'm about to quit programming.... Just had a #WordPress project. Added Ubuntu to ssh-keys on the server. Told #ClaudeAI what I wanted. And that was all i had to do... I've build "countless" Wordpress-sites over the years, but there's no point anymore.

Claude does fonts, spacing, contrast, #WCAG, #SEO, usability and design way better than me. Even more so, when you account for the speed. Content is the only thing that we still have to make sure is "human". For now. It's all exciting and scary!

1. In just a few hours #Claude and I were able to knock out a webserver that saves images from a webcam and prepare for fast motion (web, instagram, facebook) for a thing I'm involved in. 2: Also built an MCP for #MonicaHQ CRM where I can add people and ask for info directly in Claude Voice Chat...

My ideas and projects actually gets realized, documented, tested and works! The speed of which I can get measurable real world results and value now is amazing.

I'm a Windows fanboy. But. Who decided the start meny is now a folder structure ala iPhone/Android? I seldom use it, writing is way faster, and still I'm annoyed.

Yes, I can use Linux as daily if I want to and Mac if i absolutely must. Pretty tool agnostic, but Windows has been with me since 3.1. Old habits and safe havens. WSL has been my goto when Windows is not enough.

Prediction for the future of 2036: 1. Having one or more humanoid robot in the house is normal. 2. Zero to ten hours work week is normal. Many people work because they want to, not from need. 3. UBI is normal 4. The richest does not exist as today. 5. Stronger democracy.

And I do mean for all countries if the world. It will take time to get to everyone. Also there will exist communities, counties, countries that does not want all this, and hold on to "the old ways" for their reasons. Still i strongly believe this will all make the world better. I hope.

If I were you, I'd start watching all the Sci-Fis about robots becoming mainstream. After, it's not Sci-Fi. I'm very optimistic, but still afraid how the world will look in the future.

Trying to build my own "PA" or "Jarvis", using AI. I'm stuck at calling. I want to call and enable the PA to call med with voice. So far i think running my own virtual PBX and use VoIP to a supported iOS app is a possible way. Anyone?

I use #Claude, mostly #Opus5, for coding. I actually believe it writes safer code than me. Even says things like "I also found this error, should I patch it?". It finds errors along the way and surfaces them to me, unprompted. I think that is amazing.

I need a clean desk to focus. First I'll clean the desk. Then sort every cable i own and sort my old HDDs and SSDs. While I'm at it I'll just write about in on #bluesky. What were I doing, again?

When people said dumb things like "I have x Claude Pro subscriptions", I did not know yet. When I have a heavy project that needs a lot of context #Opus5 blows though tokens like the Cookie Monster from Sesame Street. :)

My #frontend #AI (#Claude in #WSL #Ubuntu) found an unused endpoint I made in #backend (Claude in #powershell) and autonomously decided i wanted a search field on the page it belonged and just made it. It was correct. It works. Comments?

I find myself creating most frontend code with #AI. Then I go in and enjoy the craft of perfecting it. Moving buttons, renaming, reorganizing, colors, meaning, UX. Something nice about not needing to manually do the grunt-work and focus on the details that matter most.

Snap-videos from me can't be viewed by my friends. So they stay forever as unread. They are not happy having a permanent unread-symbol... I think it's kinda funny, though. People calling me saying "I can't se your snaps", like I'm famous... :-)

The two weeks of included #Claude #Fable did help me finish a personal project from 2020. It is kind of dumb though, that we have to stop working until the next 5 hour reset, and then the weekly reset.

I quickly used my share of Fable in Claude Pro. It's good, but I think it's better suited for larger projects. Opus 4.8 for my projects seems more the good enough. Still, Fable is amazing. Per token prizing will probably be a hard sell internally.

I feel like every time a new #AI-model comes out it is so much better i forget how good the previous model were. #Fable is a game changes, but so was #Opus and #Sonnet. I remember when Sonnet 3.5 blew my mind, and took my on a path to being an architect more than a coder...

My problem solving workflow now is to get a good graps at what's wrong, make sure i get all the context for a prompt and then prompt #Claude for one shot a solution. Better prompts with all relevant context and steering really are important to get to the solution faster.

I made an AI thing I'm kinda proud of, inspired by the movie HER. Not what you think! It's a virtual private assitant that auto-checks mail, calendar, CRM, tasklist, and promts me before meetings, summarize my day, can write drafts, can invite to Teams-meetings, reminds me of things.

How? Built with #Opus 4.8. A lot of pyhon, APIs and MS graph, local transcription and local voice gen on GPU. Lot's of moving parts. Usable and net positive day one! #Claude #Sonnet API for some prompts that need reasoning. Basically just said yes, when it looked like Opus had a good idea.

I just repurposed an old server, a then beefy i7-3770k. Started out as a VM-server 10+ years ago at work. Then my work computer for a while. Then my home server for a while. Now a VR-rig. Not 4K, but Beat Sabre works perfectly. All hardware borderline supported, but it works! Todays HW-prices sigh..

Day 3 of using #Fedora: I'll never go back to Windows full time. At least for what I do day to day at work. Supports #mssql with #podman or #docker and even native if you are frivolous. #dotnet is native as well. Thank you, Linux community :)

Day 2 with #Fedora #Linux - now booting into my Windows drive for a second. It works. Some quirks, yes, but less than on Windows. Getting to a fully working work environment has gone so fast. Installing apps and tools as i need them. Claude helps with suggestions and CLI-commands.

I just finished my first day with #Fedora. A workday. A full workday. I'm a Microsoft fanboy, and wanted to try the switch to #Linux. Had to do it all-in to make sure I did not relapse. As a full 365 users, I did not touch Windows today! Normal work got done, with some pauses to install packages...

Installed Fedora on a separate drive so I can go back and find things and use Windows specific apps I may need. I think this could work out! Linux with Windows VM instead of Windows with WSL (which has also been great!). Still some work tasks I've not tried yet... New is fun! :)