India and yeah still living here

As far as i remember, Google lens can be used for recognising and copying text from images in photos. Apart from that, OCR would be the best bet to do it cross platform. Something like tesseract-ocr should work

For neovim, there is an excellent plugin called auto-dark-mode (github.com/f-person/aut...), which essentially does the same thing. On detecting the system theme, it sets appropriate vim background. That is then picked up by the solarized theme (github.com/maxmx03/sola...).

For my shell, since I use fish, I created a function which queries the GlobalPreferences for the current system theme (interestingly on MacOS querying the GlobalPreferences when in light theme throws an error but returns "Dark" for dark theme). This function is called on every new shell creation.C

This is so true. A lot of the times individuals who do this force taking short term decisions, which lead to a a lot of headaches down the road.

That’s why a only posts feed exists as well. From what I’ve seen though, the interaction on posts on this platform seems quite genuine. If there’s not much of a value add, folks might not reply much.

Shameless plug, I wrote an article on our reason to shift away from docker compose as well : insiders.dashtoon.com/kubernetes-a...

This accurately captures the reason we moved out of docker compose to kubernetes as well. Towards the end of it we had watchtower running to handle deployment of services, nginx with multiple subdomains and so on. Another problem with the setup was RBAC, since anyone could ssh into the vm.

India

ps: it was in fun and we reset after the first time they faced this.

ps: ps: okay maybe we let it go on for two to three times before resetting it 😅

We actually pranked one of our colleagues with this. They had a habit of constantly leaving their system unlocked and unattended. So we went ahead aliased `ssh` to `sl`. It was quite funny to see them perplexed as to why there was a train running across the screen instead of logging into a vm.

I’ve been wondering about the same thing, and one of the core reason seems to be that has incentivised users for engagement while earlier there was no incentive for users to post or take part. This has lead to more controversial takes and random bot posting on threads.

Going through the entire Assassin’s Creed franchise slowly. Currently playing Black Flag

It could turn out that I'm unnecessarily pessimistic about the future and everything turns out okay with developers actually having their cake and eating it too. But my current stance is that it seems highly unlikely.

The pain of discovering solutions and actually going deeper is now abstracted away, and one just needs to now "prompt better" to get stuff done.

One caveat is that the codebase that I've been working on was mostly written by junior engineers. They are the ones which would be the most affected due to the ease that an LLM now affords them.

Quite frequently I've had to spend multiple hours trying to understand the flow of control and in the end re-doing the logic to add functionality due to rigidity of the logic that was written.

I've myself faced a lot of performance and extensibility issues while working on pieces of code that were architected and created primarily by prompting.

And this would manifest in multiple ways like increased security vulnerabilities, obscure performance issues, logical bugs for edge cases not thought about and so on. This in turn would make the job of those who inherit these systems much harder due to the amount of unstructured and disjoint logic.