teejaypeeps

The article reads as though an LLM wrote it

I don't believe they are, in the main. I think many folks are in a bubble and believe this to be the case, but the large employers don't have the majority of engineers working today. Think of all the small cap shops, less than 100 employees, with just a couple of folks looking after the on-prem dB. They've not migrated to the cloud. Will they? It's hard to tell.

But engineering is shifting my friend, that's for sure. The syntax master is dead, those who tied their worth to recalling obscure functions and producing the most beautiful code are gone. Engineers need to sit more with the business. We will become a BA/PO engineer hybrid, or the POs will take our jobs and just get agents to build it all.

Architecture is going to CALM. Ontology provides the business definitions. Data contracts form the semantic layer.

All of that can be automated via a GUI into a repo with bots merging, after CI passes.

The apocalypse won't be a revolution, it'll be an evolution. A slow war of attrition.

If you wanna stay ahead, I reckon you're going to have to get used to leaving your agents doing stuff whilst you're talking to the business.

1) Claude Pro. Just the £18 a month one, as that's more than enough for my personal needs. Just the one sub.

2) I've an OpenAI API endpoint to route things to free models, too. Just general query things. odysseus sits over that, plus a few tiny local models that sit on 2x4GB Nvidia T600 cards. Text embedding, classification, email summaries to myself etc.

https://git.thomaspeoples.com/user/login?redirect_to=%2Fthom...

3) I started off with aider and still mildly use it, but CC just hooks you in, like crack. Its addictive and fun. I use Github CoPilot at work, which is drier than Ghandi's sandal and not as nice of an experience as Claude.

4) Yes. Just a mini-orchestration skill, as part of a plug-in, which fits mine (and colleagues') use cases nicely:

https://github.com/wraith-proxy/wraith-skills

5) Nope! Tried ClawdBot and didnt like it. Not enough RAM on my NAS for Hermes and all the other crud I've got going on

Yes, I wpuld be surprised if it was not. The type of information people are putting into these models, with blind faith people aren't reading them, is mind boggling really. One reason locally hosted models "win" is things like "sort out my monthly finances " and stuff you do not want to be saying to Big Tech.

I use Groq APIs for the free tokens to oss-gpt :-)