I watched my PCP read the Google AI summary to me in office when I asked a question about my visit about a year ago... I'd rather just use my Claude sub at this point than pay $180 for 10 minutes of interaction. AI should be disruptive in this space, but it's a walled garden and I feel like they will do everything to avoid disruption.
So a breach occurs and they recommend people rotate keys? But they claim fraud and break everyone's access with existing keys?
Why anyone uses their services, or even the harness product at this point is beyond me. They just shill cheap and crappy models for training and make boneheaded decisions and "mistakes".
What other open source harnesses are there which have a good, albeit opinionated, experience out-of-the-box?
The $10/40 monthly subscriptions are a great deal. Don't proxy your subs through any one vendor, own your keys, because this is just one of many such events.
Too heavy handed and opinionated, not interested in the "agent team" setup found therein. I have a set of plugins I prefer, building blocks assembled to create my own take, with some custom ones on the horizon.
I'm generally disappointed with the Pi team and their dismissal of a major security gap
The magic is a tropical tasting fruit! I have a tree and it's tough to actually get one of the fruits when they are ripe. This year I got lucky with a few that feel into my fenced garden section and I got them before the critters.
I think Durian smells more like natural gas than vomit, but I know what you mean about the vomit smell. Especially when you have a huge harvest and the fragrance is overwhelming
I can't get 3.8 to exit thinking loops. It will just think and think and think on the most trivial topics. I wanted it to port a speed test powershell script to c#. Claude opus 5 completes it under 60 seconds. I let 3.8 churn about 6 different times for 30+ minutes and it never wrote a single line of code to disk. It wrote lots of lines in thinking.
I've found it tends toward long thinking loops even for simple tasks (and any quantization seems to increase their length), but those do exit eventually, unlike with Qwen 3.6.
I use the Unsloth UD_Q2_K_XL GGUF with default parameters, along with that custom template linked elsewhere in the thread, and no K/V cache quantization.
For smaller models, you'll probably find anything below Q4 will need handholding. Check Unsloth's graphs at the different quantisations VS error rates and you'll see why.
My experience has been that anything less than a 4-bit quant has a tendency to go off the rails. There’s a threshold of coherency that is being crossed somewhere internal to the model I guess.
Try the same prompt with a larger quant (even if it runs very slowly because the model no longer fits in VRAM) & see if Qwen does better - if so, there’s your answer.
That sounds like something is off - I'm using UD-Q4_K_XL on pi with xhigh thinking, and unless I'm vastly underestimating the complexity of the script that's the kind of task I would expect to take a couple of minutes (getting ~30t/s decode). What server are you running, and are you using the recommended parameters from qwen/unsloth?
