I assume it was too expensive, because it's really not a bad tool. I used it recently to make my twitter pfp :)

Theres tons of ai apps. They're all general use chatbots or coding agents. Manus, Cursor, ChatGPT. Almost every app that has a robust search uses a reranker llm. AI is everywhere.

As far as totally new products - I built one (Habit.am - wordless journaling for mental health) and new products require new habits, people trying new things, its not that easy to change people's behavior. It would be much easier for me to sell my little app if it was a literal plain old journal.

Even if it runs, this will run slowly, and heat up.

I think local will always have a place, but the infrastructure is going to be used in my humble opinion.

Today yes, but between the improved performance of smaller on device models and the hardware itself getting better this issue is short lived.

I don't want to put information into a black box of mystery that can then be used for other monetization purposes. I am still waiting for a realistic local solution.

Have you tried qwen3.5 running locally? It’s quite “good enough”.

Compute evolved from batch systems with time sharing to responsive systems in your pocket. Why wouldn’t that happen here?

there was a time when mainframes was the main thing.. we’ll look back and say data centers was a thing.. (hopefully if we lucky)

is it just me or could every set of claude skills just be a blog? I don't install skills without reading them first anyway.

I ran into it while building - I should have tested different temps too - I was just trying to get cli style tool calls to be more reliable

yeah temperature is probably worth a run, we noticed even small adjustments changed how the model interpreted formatting expectations quite a bit.

I ran tests of 100 attempts with different prompt/scenario combinations. Each "attempt"/theory had 3 different system prompts wordings. Most of the prompts did not mention a colon, but it kept appearing. When I added negative instructions against using a colon, the quality went down (most of the tool calls were malformed, one common issue was markdown ticks in front) It was only when my system prompt acted like colons were normal that I kept getting 100/100 perfect expected tool calls. I ranked my system prompts by which returned the most consistent commands.

I also expected this. Please run some experiments and maybe other models are different

I tried over 20 variations of different system prompts. Once I changed my tool to expect the colon, it also felt like it was running/calling tools faster, but I need to do a larger test to be sure.

Try this prompt "Hi Claude, please make my site accessible. Pay particular attention to contrast ratios for text and the backgrounds they are on.

Seemed a bit off, but I’m no designer. Thanks for the tip.

I dislike that the source of the skill is not linked directly.

Let me get that fixed and listed prominently

updated! You should find it right above Description