A lab notebook on AI running on my own hardware: quantization, throughput, VRAM and the real cost of running LLMs locally.https://blog.jonathanschenker.com.brenhttps://blog.jonathanschenker.com.br/en/posts/coloring-drawings-with-local-ai-from-prompt-to-print-ready-filehttps://blog.jonathanschenker.com.br/en/posts/coloring-drawings-with-local-ai-from-prompt-to-print-ready-fileGenerating the image is the easy part, and that's where most tutorials end. What determines whether the child can paint is what comes next: binarizing, vectorizing, and printing without jagged edges.Tue, 30 Jun 2026 00:00:00 GMTimageFluxlocal AIprintingMLXhttps://blog.jonathanschenker.com.br/en/posts/qwen3-6-27b-how-much-vram-it-really-needshttps://blog.jonathanschenker.com.br/en/posts/qwen3-6-27b-how-much-vram-it-really-needsA 27B model that uses half the KV cache of a Llama 3 8B. The hybrid architecture breaks the usual calculation — and decides, for less than 1 GB, which cards are left out.Thu, 25 Jun 2026 00:00:00 GMTLocal LLMVRAMQwenHybrid AttentionGPUhttps://blog.jonathanschenker.com.br/en/posts/how-much-vram-an-llm-really-useshttps://blog.jonathanschenker.com.br/en/posts/how-much-vram-an-llm-really-usesThe back-of-the-envelope math — parameters times bits — is off by several gigabytes, because it ignores the KV cache. Here is the full calculation, with the formula, the numbers and how to verify it on your own GPU.Wed, 10 Jun 2026 00:00:00 GMTlocal LLMVRAMquantizationGPUhttps://blog.jonathanschenker.com.br/en/posts/spec-file-doesnt-make-model-accuratehttps://blog.jonathanschenker.com.br/en/posts/spec-file-doesnt-make-model-accurateIn July I argued that structured spec changes the outcome. I measured in October: 360 generations, 3 local models, 3 ways of asking for the same thing. What pays off is the written information, not the file format.Tue, 14 Oct 2025 00:00:00 GMTspec-drivencode agentsAiderlocal LLMbenchmarkhttps://blog.jonathanschenker.com.br/en/posts/prd-comes-first-spec-prevents-agent-errorshttps://blog.jonathanschenker.com.br/en/posts/prd-comes-first-spec-prevents-agent-errorsThe model doesn't make mistakes due to ignorance: it makes mistakes because it filled the gap the prompt left, and filled it in a plausible way. What changes in the code when the contract comes before the request.Tue, 15 Jul 2025 00:00:00 GMTcode agentsspec-drivenClaude CodePRD