Special Topic

Open Weights, Distillation & Local Runs

9 notesOpen reading · no login

Special Topic09notes

THE QUESTION THIS PAGE ANSWERS

ANSWER FIRST

What will the “Open Weights, Distillation & Local Runs” AI learning path help you do?

Separate open weights from open source, read licenses before deployment, and reason about distillation, hardware, and local serving. Finish with a grounded choice of what your own computer can actually run. The path contains 9 free notes, each centered on one question you can understand and test.

DECISION RULE

Core themes include What Open Source Actually Opens, How Large Models Get Smaller, Running It on Your Own Machine.

TRY NEXT

Begin with “What Are Weights? Everything a Model Knows How to Do,” then choose the next note by the task in front of you.

WATCH FOR

Do not optimize for finishing the list. Explaining one trade-off with your own example matters more than opening more titles.

What this route helps you practice

Open the first note

Each chapter follows a class of real decisions. Follow the sequence, or enter at the problem you are solving today.

9notes

01What Are Weights? Everything a Model Knows How to DoThe file that months of training finally condenses into: what it looks like, how big it is, and why holding the weights means holding controlConcept6 min↗02Real vs. Fake Open Source: How to Read a LicenseThree questions that locate how open a model is; the same yardstick applied to Qwen, Mistral, DeepSeek, Llama, and API-only modelsSelection6 min↗03Open Source Is a Business: What Each Vendor Is AfterSix vendors' open-source strategies and paths to revenue; why the number of derivative models says more than download countsCase Study7 min↗04Emergence: Why Capabilities Appear SuddenlyCapabilities jump in steps once a model crosses a scale threshold — plus the academic dispute this phenomenon is still underConcept8 min↗05Why Make Models SmallerThree practical motives — cost, speed, on-premise deployment — and the things small models cannot doMethodology9 min↗06How Distillation Works: From Teacher to StudentThe five-step pipeline, soft labels, and temperature; using the six distilled models DeepSeek open-sourced alongside R1 as the sampleCase Study7 min↗07The Cost of Distillation: Models Are Getting More AlikeVerbal tics, formatting quirks, and identity confusion inherited wholesale; why multi-model cross-validation may be fakeDeep Dive7 min↗08How Large a Model Can Your Computer RunPick a GPU or Mac model for a real-time answer; the VRAM formula, quantization levels, and the MoE mismatch between memory and speedInteractive9 min↗09Getting Started with Ollama and LM StudioThe full set of commands from install to running, how to read model tags, how to choose a quantization level, and the three most common trapsHands-on9 min↗