Anything I know the best practices on how to architect. I don’t do “agentic coding”. I treat AI as a junior developer I have to tell everything and check the work at every phase.
I work in cloud consulting specializing in app dev. I know AWS well (trust me on this). I don’t know Terraform or the CDK. I use CloudFormation. I was very comfortable not knowing either on projects because I knew I could troubleshoot issues and be prescriptive and detailed.
I wouldn’t try to be an “Azure Architect” and not know the first thing about Azure and depend on AI.
In web dev, writing load tests and user journey scripts has become a lot easier and something where the end result is more important than the code flavor, compared to maintainable site feature code.
Turn on persist logs in the dev tools network tab; go through as much as the flow as possible; filter out domains that aren’t your site: google, facebook, external api calls; download the .har for all of whats left; convert har into a k6 script using a library, or dump that into an llm to convert it to their newer browser script and point the llm at their docs; edit it to be dynamic so you can test paths from different product pages/types, and scrape and enter proper guids, etc.
On the flip side we’re seeing a lot more bot traffic likely due to bad actors doing this too, but by writing these tests you begin to see what calls cause load and be proactive.
I Manage IT for a small/medium company. We were paying quite a bit for helpdesk software. Last year they moved their modules around so the features we wanted (IT Inventory) cost extra per item. With the help of AI I was able to vibe code my own helpdesk with Python/Flask. It does everything we need to do (and then some lol) and works great for us. We keep it on the inside of our network. It has been a great and fun project for me in my spare time and it is all possible because of AI!
We’ve managed to put quite a dent in the feature backlog we had for our admin tooling. We knew what features we wanted, they didn’t require marketing or comms but did require time to spend doing them.
I did want it to improve our e2e testing but it didn’t make it as easy as I expected.
My personal opinion is that if a team couldn't write something before AI then they should be very careful about writing it with the help of AI.
For example a team that couldn't write a new encrypted messaging app without AI, gets an AI to write them one. How do they check that the code is actually secure? Writing encryption code is very hard to get correct, in fact most humans can't get it right, and if you don't understand the intricacies of cryptography then you'll never pick up the mistakes the AI makes.
Anything I know the best practices on how to architect. I don’t do “agentic coding”. I treat AI as a junior developer I have to tell everything and check the work at every phase.
I work in cloud consulting specializing in app dev. I know AWS well (trust me on this). I don’t know Terraform or the CDK. I use CloudFormation. I was very comfortable not knowing either on projects because I knew I could troubleshoot issues and be prescriptive and detailed.
I wouldn’t try to be an “Azure Architect” and not know the first thing about Azure and depend on AI.
In web dev, writing load tests and user journey scripts has become a lot easier and something where the end result is more important than the code flavor, compared to maintainable site feature code.
Turn on persist logs in the dev tools network tab; go through as much as the flow as possible; filter out domains that aren’t your site: google, facebook, external api calls; download the .har for all of whats left; convert har into a k6 script using a library, or dump that into an llm to convert it to their newer browser script and point the llm at their docs; edit it to be dynamic so you can test paths from different product pages/types, and scrape and enter proper guids, etc.
On the flip side we’re seeing a lot more bot traffic likely due to bad actors doing this too, but by writing these tests you begin to see what calls cause load and be proactive.
I Manage IT for a small/medium company. We were paying quite a bit for helpdesk software. Last year they moved their modules around so the features we wanted (IT Inventory) cost extra per item. With the help of AI I was able to vibe code my own helpdesk with Python/Flask. It does everything we need to do (and then some lol) and works great for us. We keep it on the inside of our network. It has been a great and fun project for me in my spare time and it is all possible because of AI!
https://helpfuldjinn.com/ https://github.com/DjinnRutger/HelpDesk-Public
The exact same things I could build before. There was never anything stopping me in the first place.
We’ve managed to put quite a dent in the feature backlog we had for our admin tooling. We knew what features we wanted, they didn’t require marketing or comms but did require time to spend doing them.
I did want it to improve our e2e testing but it didn’t make it as easy as I expected.
My personal opinion is that if a team couldn't write something before AI then they should be very careful about writing it with the help of AI.
For example a team that couldn't write a new encrypted messaging app without AI, gets an AI to write them one. How do they check that the code is actually secure? Writing encryption code is very hard to get correct, in fact most humans can't get it right, and if you don't understand the intricacies of cryptography then you'll never pick up the mistakes the AI makes.
fair point
Web sites that look professional. I have not been able to do that until now. And in pure HTML/CSS so no need for bloat and slowness.
doom on the prompt