Jaime

[tunnie]

Claude is our choice at work, it's been quite impressive. So much so they are laying off 50% of my teams 

How many bodies is 50%?

Android is going from 10 to 3
iOS is going from 7 to 3

But I'll gain a few from more bodies from some cheaper countries, making it over-all ~50%, although that is just developers. That does not include my other Mangers (5) that I have reporting into me or my QA (App Testing) Team.



I have a sneaky feeling I'll be for the chop this time next year, I'm fully aware I'm middle-management bloat and there are way to many of us now they are doing this down sizing.

AI is powerful, even this useless student is making some fairly decent production level code changes. 

Next step        Influencer

[Viral_Jim]

Tech is a pretty tough field to be in just now, we are taking an absolute hatchet to our teams, ostensibly off the back of a merger at the start of the year, but in reality due to the overwhelming need to cost-cut. I've seen Claude mentioned in a few meeting packs for us too, not sure where we're deploying it however.

[TheBoy]

At the moment, we aren't deploying much AI generated code into production, as we are finding too many issues, particularly with security rather than functionality, that takes too much time to fix.  We are using it a fair bit in other areas though.

Individuals are using it extensively to write reports and presentations, and annual reviews.  Being a company that predominately uses Copilot, it's very obvious when someone has been lazy....

[tunnie]

I found Co-Pilots default engine (can't remember what it was) would just get into a loop all the time, with me correcting it and then it going off on the same wrong route again then getting stuck and saying I was right to correct it.

I've found VS-Code with Claude much more powerful.

We now have tools installed to monitor AI vs manual code going into production, the expectation here is you need to use AI daily to write your code. If you are not using it, why are you not using it.

One team is even trying an entire sprint going 100% AI and having no manually written code. 

[Doctor Gollum]

How is that even considered as an option?

Ok, any code/software/programming is only as good as the person that wrote it, but allowing a stupid system potentially programmed by idiots to write itself is only going to end badly...

It's going to be the automated version of letting a five year old do the family shop... You're going to be eating lucky charms until the lights go out.

And for all those involved you do know that for work/life balance to be a thing, you need to be employed...

[TheBoy]

We now have tools installed to monitor AI vs manual code going into production, the expectation here is you need to use AI daily to write your code. If you are not using it, why are you not using it.

I think MS has led the way on using AI for coding - I think it was last year Nadella proudly claimed that 30% of code in their repos were AI generated....

...and I think it shows, given the dramatic increase in failures in their software over the last 3 or 4 years. MS code, be that Windows, Office or backend server systems, has become so bad, even an Apple hater like me tends to use a Mac for personal use!


The general consensus amongst those doing mission critical is that all AI code needs to be fully human checked, and this needs at least the same staffing levels as pre AI.  Some companies are looking for a quick win, and generative AI appears to give that, with the hope that it will also fix the flaws before the company gets bitten.

Top tier hacking groups - prob the state level type - are already showing (more PoC at this stage) ways to get generative AI to implement exploitable weaknesses....