Daniel Suarez-Mash

@kilocode I have to say, this is one of the most insightful posts on AI models I’ve seen in a while. Fantastic analysis @kilocode 🔥 This is what we should be reading about instead of obsessing over benchmarks. Open models aren’t the best, but they are the best value for money.

@opencode Can you use capital letters when you start a sentence?

we've added unique user rankings some models are token heavy so they skew upwards in rankings - unique people using the model is a more accurate ranking we'll orient more of our data around this metric

@AdinaYakup @PaddlePaddle Any idea whether this will arrive in the Docling python package?

Unlimited-OCR 🔥New OCR from @PaddlePaddle It can parse hundreds of pages in a single pass while maintaining stable speed. The key idea is R-SWA (Reference Sliding Window Attention), which keeps KV cache constant during decoding. 🏆 93% on OmniDocBench 📈 +6% over DeepSeek-OCR

@kilocode @_inception_ai Love this. Is mercury edit 2 a model we can run locally? Or is the inference time too slow for that? I’m currently enjoying the free codestral autocomplete, but assuming this new autocomplete will only be available with the kilo pass?

Next-Edit is live in Kilo, powered by Mercury Edit 2 from @_inception_ai. Autocomplete predicts the next few tokens ahead of your cursor. Next-Edit predicts your next actual edit anywhere in the file. Hit Tab to accept. And it's free for everyone for 30 days!

@kilocode Feels like there should be a method to automatically compare the results and synthesise work done across all 3 work trees into one final worktree?

Same prompt. Three models. Three answers. You keep the one you like. <<< Agent Manager is live in Kilo Code. >>>

@thsottiaux I use @kilocode 👀

Do you use codex /goal occasionally or as your main way to get things done?

@mattpocockuk Using your /teach skill to be proficient with @huggingface Transformer's library, something I've always struggled with! It's going great, lesson 0001 done 👍

I poured my 10 years of teaching experience into a skill. It's called /teach, and it can teach you anything. Here's how it taught me to solve a Rubik's cube:

@kilocode I have to say, this is one of the most insightful posts on AI models I’ve seen in a while. Fantastic analysis @kilocode 🔥 This is what we should be reading about instead of obsessing over benchmarks. Open models aren’t the best, but they are the best value for money.

@ostensiblyneil How does this compare to running models on apple silicon via llama.cpp? Is it worth benchmarking these metrics against other runtimes like llama.cpp, ollama or oMLX?