Which Harness Are You Using Today?

xReminisce

Elite Member
Joined
Dec 29, 2012
Messages
7,335
Reaction score
36,666
With coding agents evolving so fast, I’ve been paying a lot more attention lately to the harness driving the model rather than just the model itself. The way these runtimes manage context, handle tool permissions, and execute code in the background actually makes a massive difference in how smooth the daily workflow actually feels.

So what AI Harness are you using currently?
Obviously everyone has their own different sweet spot for their setup.

What drives your day-to-day work, and how well is it fitting into your actual routine?

Mine is currently using CC, AGY and GH Copilot in Cursor IDE (along with Cursor).
Although I am experimenting with Orca right now to switch away from Cursor.
While work is asking us to try Omnigent but lol maybe we'll see
 
VSCode + Codex on the side, I feel thayt they make the models work better for their own harness, as in if you work with openai models use codex, claude then claude code, etc...

I will be trying deepseek new harness which they say it is positive!
 
TBH this post is the first i learned of the technical AI word "harness"? lol
 
I will be trying deepseek new harness which they say it is positive!
Yeah, i was going to but that's only API calls only.
So probably when I run API-based models like kimmy and DS
TBH this post is the first i learned of the technical AI word "harness"? lol
sorry :(
 
I feel like there is a whole massive eco system of automation building up with things like what your doing, and a whole new language and word a minute! lol . I'm defiantly a dinosaur, i only trust the web and copy paste when it comes to AI.

1787062436308.png
 
I feel like there is a whole massive eco system of automation building up with things like what your doing, and a whole new language and word a minute! lol . I'm defiantly a dinosaur, i only trust the web and copy paste when it comes to AI.
You are missing out then! With a harness, you have access to tools in agentic mode and I feel it is much more powerful than normal chats.
 
You are missing out then! With a harness, you have access to tools in agentic mode and I feel it is much more powerful than normal chats.
I feel like there is a whole massive eco system of automation building up with things like what your doing, and a whole new language and word a minute! lol . I'm defiantly a dinosaur, i only trust the web and copy paste when it comes to AI.

View attachment 541582

yeah pigging back on what @SEOCodeLab is saying
Chat mode is heavily regulated and instructed to reply back a certain way.
You hear all the noise currently about "Claude Watermarking", well Chat is worse.
Because it forces an agent to respond in a certain matter, and when the agent truly has no idea or hasn't done research yet, it's forced to reply back hoping it is correct.
Because it's cheaper for the company to give you a couple of wrong answer back and forth, than to actually have the agent research and give you the correct answer. In the long run anyways.
Ontop of Chat-AI Bots barely retains memory. Even the "Memory mode" is full of shit, its basic Regex, and some providers go even further and cheap out (talking about you Perplexity when you aint using its Computer mode) that it's truncates and only regex the PREVIEW of the file you gave it.
 
With coding agents evolving so fast, I’ve been paying a lot more attention lately to the harness driving the model rather than just the model itself. The way these runtimes manage context, handle tool permissions, and execute code in the background actually makes a massive difference in how smooth the daily workflow actually feels.

So what AI Harness are you using currently?
Obviously everyone has their own different sweet spot for their setup.

What drives your day-to-day work, and how well is it fitting into your actual routine?

Mine is currently using CC, AGY and GH Copilot in Cursor IDE (along with Cursor).
Although I am experimenting with Orca right now to switch away from Cursor.
While work is asking us to try Omnigent but lol maybe we'll see
Mostly sticking with GitHub Copilot inside VS Code for my day-to-day PHP/WordPress work, it's simple enough that it doesn't get in the way. Tried a couple of the newer agent-based setups but for smaller freelance projects the overhead of managing permissions/context isn't really worth it yet. Curious to hear how Orca compares though, might give it a shot if the workflow's smoother.
 
Mostly sticking with GitHub Copilot inside VS Code for my day-to-day PHP/WordPress work, it's simple enough that it doesn't get in the way. Tried a couple of the newer agent-based setups but for smaller freelance projects the overhead of managing permissions/context isn't really worth it yet. Curious to hear how Orca compares though, might give it a shot if the workflow's smoother.
Yeah, I been hearing praises of Orca, i installed it just about an hour or so ago.
I'll give it a few days to see how it handles my workflow
 
Yeah, I been hearing praises of Orca, i installed it just about an hour or so ago.
I'll give it a few days to see how it handles my workflow
so far
orca not playing nice with my cursor CLI, non stop conhost JSON error
pissing me off.
i'll look more into it tomorrow i guess
 
Back
Top