sheo

5 ms·

1. 1.▲
   by sheo 5d ago

   As said in the article, all of the providers run different runtimes with proprietary, sometimes untested and weird optimisations There are multiple types of quantisation: weights and KV Cache. Quantizing kv cache can drastically hurt perfor
2. 2.▲
   by sheo 6d ago

   Disclaimer: I'm using reduced motion on my phone, please acknowledge this bias before reading my comment All of those, citing the pricing page, "websites people remember" look the same and use the same handful of effects - Fa
3. 3.▲
   by sheo 3mo ago

   Because there is no single "default instance that is always a good choice and wouldn't go down randomly because of lack of funding". That's both a strong and a weak side of fedi
4. 4.▲
   by sheo 6mo ago

   I found the opencode fork repo, but no plugin seems available so far https://github.com/Vibecodelicious/opencode
5. 5.▲
   by sheo 1y ago

   I think Forgejo is great. But we need to remember the fact that many enterprises pick GitHub or GitLab because they provide paid support and SLA. Also Forgejo Actions are still in beta, while GitLab and GitHub are pretty established solutio
6. 6.▲
   by sheo 1y ago

   I'm surprised this article is not taken down. OSM wiki states that what3words the company have a good history of taking down code (and maybe articles) like this one, because algorithm (any algorithm that converts any number of words to
7. 7.▲
   by sheo 1y ago

   I think that the example in the article is not a good usecase for this technology. It would be better, cheaper and less error prone to have prebuilt forms that LLM can call like tools, at least for things like changing shipping address Ship