With the recently released models and various open source tooling like harnesses or even solutions such as Graphify (is anyone actually using this?) or the death of OpenClaw I feel rather behind with regards to how I apply AI during my day to day engineering tasks.
The goal for this post is then therefore to act as a probe into what the public considers as standard/best practices when working with AI and maybe act as a sort of sharing space for articles / ideas you have found interesting or apply while working with AI.
I feel like this could be of great value (although maybe the approach I'm taking right now is not the most efficient) to see what the public thinks is working/is bogus since all of the recent hype trains and their subsequent crashes make me (and potentially you too) feel out of touch.
Clarifications:
- by saying AI I refer to AI assistants such as claude, codex or similar.
- I have a central folder I use with context, project folders, tasks.md files etc.
- I have a skill like /plan-my-day that makes a nice html file with tasks list, status of my work and projects etc.
- when I'm starting up a new idea I use matt pocock's `grill-me` skill, I dictate a lot with https://handy.computer/ text going straight into claude code window.
- once I have a decent brain dump I figure out what to do next, whether it's code up a prototype or find some one to bounce ideas off and ask some questions (an internal person or a customer)
I'm always looking for some kind of novel way that people have found to use, and the highest ROI thing I've found is the concept of the grill-me skill.
Please share other workflows, ideas, skills etc.
Some advice:
Don't install a whole plugin of skills, they use way too many tokens at rest in your context, instead work in your own .claude folder often, and vendor copies of specific skills or write your own.
I'll have to check out grill-me. It seems to slightly address a problem I have where for one off tasks it seems no workflow at all is necessary, but planning out projects before implementing them is rather unwieldy with claude especially when I move towards POC stages where the I feel like refactoring the design based on my PoC results is just an extremely detatched process.
Feeling behind is natural, there's far, far more tools out there than accounted for. The main thing is to ask yourself how do you work and what's needed for your goals, projects, etc. I'm personally not quite there is letting the ai go autonomously, like if I'm working on paper I have it connect to zotero so I have a list of what's being cited. I think a lot of tools are bogus, you can go far with just model file + llama.cpp + tmux is just getting something running, obviously making a skill can automate some aspects of what the model does but remember models are not deterministic by nature.
My advice is find what you'll actually use and ignore the rest, I don't use a lot of skills or plugins really, maybe I could sit down and make 1-2 just thinking on it but really can't be asked. If someone tells you you're behind always remember in 2026 there are far more people globally who's biggest concern is if they'll have food to eat, running water, or working lights not "should I just the latest openai or anthropic model"
The goal for this post is then therefore to act as a probe into what the public considers as standard/best practices when working with AI and maybe act as a sort of sharing space for articles / ideas you have found interesting or apply while working with AI.
I feel like this could be of great value (although maybe the approach I'm taking right now is not the most efficient) to see what the public thinks is working/is bogus since all of the recent hype trains and their subsequent crashes make me (and potentially you too) feel out of touch.
Clarifications: - by saying AI I refer to AI assistants such as claude, codex or similar.
- I have a central folder I use with context, project folders, tasks.md files etc. - I have a skill like /plan-my-day that makes a nice html file with tasks list, status of my work and projects etc. - when I'm starting up a new idea I use matt pocock's `grill-me` skill, I dictate a lot with https://handy.computer/ text going straight into claude code window. - once I have a decent brain dump I figure out what to do next, whether it's code up a prototype or find some one to bounce ideas off and ask some questions (an internal person or a customer)
I'm always looking for some kind of novel way that people have found to use, and the highest ROI thing I've found is the concept of the grill-me skill.
Please share other workflows, ideas, skills etc.
Some advice:
Don't install a whole plugin of skills, they use way too many tokens at rest in your context, instead work in your own .claude folder often, and vendor copies of specific skills or write your own.
I'll have to check out grill-me. It seems to slightly address a problem I have where for one off tasks it seems no workflow at all is necessary, but planning out projects before implementing them is rather unwieldy with claude especially when I move towards POC stages where the I feel like refactoring the design based on my PoC results is just an extremely detatched process.
Feeling behind is natural, there's far, far more tools out there than accounted for. The main thing is to ask yourself how do you work and what's needed for your goals, projects, etc. I'm personally not quite there is letting the ai go autonomously, like if I'm working on paper I have it connect to zotero so I have a list of what's being cited. I think a lot of tools are bogus, you can go far with just model file + llama.cpp + tmux is just getting something running, obviously making a skill can automate some aspects of what the model does but remember models are not deterministic by nature.
My advice is find what you'll actually use and ignore the rest, I don't use a lot of skills or plugins really, maybe I could sit down and make 1-2 just thinking on it but really can't be asked. If someone tells you you're behind always remember in 2026 there are far more people globally who's biggest concern is if they'll have food to eat, running water, or working lights not "should I just the latest openai or anthropic model"