> I'm getting old and I value my remaining time on the planet.
It's an interesting sentiment. I, too, am getting old and value my remaining time on the planet, and so I code by hand every chance I get. :) Luckily I'm in a position to be able to do that.
Speaking for myself, I feel a difference if I stop using AI for a week and just rely on regular web searches. And I have a fair amount of professional experience
Also, speaking for academia, AI is basically all we talk about now when it comes to curriculum and instruction. That's not to say that we only rely on AI, or something like that, but we talk a lot about how to get basically anything done now. It's the biggest learning experience we've ever had as instructors, and I suspect we'll be trying to figure it out for a long time to come.
> I've never liked or trusted Canvas's gradebook, and so although I do upload grades to Canvas so students can see them, my primary gradebook is always a spreadsheet I maintain locally.
That makes you one better than me. :( One thing's for sure--I'm never trusting it again.
I already had almost all my materials outside of Canvas and just used their API to upload it. So at least that's safe. But the grades... dang. Luckily we're only halfway through our quarter and it's not finals week.
Our instance is still down, but your update gives me hope.
We're in the same boat. I would rather have NO llm, than an llm that collects my data (which you should assume is all of them, unless you've been asleep for the last 20 years).
Fortunately, I don't have to pick one or the other - instead I run Qwen 3.6 35B A3B. It's a bit slow with my 8gb GPU (I'm in the process of getting a bigger one) but again, to me the choice isn't "what's the best I can get", it's "what's the best local I can get".
I let Qwen3.6-27B chew on a bug all last night. It choked at some point and stopped responding (probably a context overflow before pi-coding-agent could compact it). Claude Sonnet 4.6 found and fixed the bug in under 10 minutes.
Qwen3.6 is pretty amazing for a 27B model, but it's not hard to run into its limits. With a Radeon R9700 and unsloth's 6-bit quantization, I get ~20 TPS and 110k context, so it can do a fair bit quickly.
You definitely need to watch it more than a model 100 times larger. But the fact that it runs one 1 GPU and does what it does is insane. Imagine what a 30b model looks like in 6 months or 1 year?
Inference speed is still slow in a meaningfully different way. The models are good, but not great, and much slower, which for coding means a 2-3 minute task with claude code and opus takes an hour and has a higher chance of being wrong.
reply