There is a fine tuned version of Quinn three, eight 27 B on hugging face for coding with the pie agent harness.
If you are unfamiliar with pie, it is literally like the most lightweight coding wrapper.
Uh, And they have done fine tuning on the model to make it better for writing code with Pi.
And I think that's really cool because I run a 27 gigabyte quantized version of that.
And how many people have computers that could hold that and do inference on it? It's most company machines, most company developer machines.
come with at least 32 gigabyte, if not 48 or 64, which means you could host that and be coding with it without having to pay for tokens.
So what kind of machine are you running it on? Is this the one running on your, on your, this one? I have
it'll be faster yes but if you want to have an agent running if for the situation right where you're You've hit your code Claude limit because you're on the team plan and it has a limit every five hours and you want to still keep working.
This would keep, let you keep working while you do, uh, you could manually do some code or you could do some reviews or whatever it like, is it as fast? No, but it's going to be humans read at about 15 tokens a second.
And so anything faster than that is going to be as fast, like it's going to be writing back to you as fast as you can process.
Can you run 15 asynchronous ones? No, but you can like, it's still very usable.
There is a fine tuned version of Quinn three, eight 27 B on hugging face for coding with the pie agent harness.
If you are unfamiliar with pie, it is literally like the most lightweight coding wrapper.
Uh, And they have done fine tuning on the model to make it better for writing code with Pi.
And I think that's really cool because I run a 27 gigabyte quantized version of that.
And how many people have computers that could hold that and do inference on it? It's most company machines, most company developer machines.
come with at least 32 gigabyte, if not 48 or 64, which means you could host that and be coding with it without having to pay for tokens.
So what kind of machine are you running it on? Is this the one running on your, on your, this one? I have
it'll be faster yes but if you want to have an agent running if for the situation right where you're You've hit your code Claude limit because you're on the team plan and it has a limit every five hours and you want to still keep working.
This would keep, let you keep working while you do, uh, you could manually do some code or you could do some reviews or whatever it like, is it as fast? No, but it's going to be humans read at about 15 tokens a second.
And so anything faster than that is going to be as fast, like it's going to be writing back to you as fast as you can process.
Can you run 15 asynchronous ones? No, but you can like, it's still very usable.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.