Category: AI
You are here if AI showed up in your job and the names on the screen do not match the buttons. This category is practical. It is not a model-release calendar, and it is not one lesson pasted under a new product name. The series split the pile so a Claude post and a ChatGPT post can teach different work.
Everyday paths are for a person with no account yet, or a person whose chat forgot a file: setup, prompting, memory, agents, safety, and which product fits a job. Product paths stay on one tool: Claude, ChatGPT, Gemini, or Grok, from a first login to the surface you should not open. Open-weight paths are Llama, DeepSeek, Qwen, Kimi, and GLM: what the name refers to, and when a normal paid chat is the easier tool. Engineering paths are for running a model, retrieving from your own files, and fine-tuning. Several of those hubs are still waiting on a first live part, and the hub page says so.
Where to start: no account yet, use AI setup from zero. You already pay for one product, open that product’s series and use the first part. You want the idea before the brand, use Practical AI. You want every series in order, use Learn and the AI tracks.
When Claude or ChatGPT is easier than running a Llama model yourself
Choose Llama when local or open weights earn the ops cost. Choose a closed $20 seat when the job is writing and the GPU project has no owner.…
Community versions of Llama: how to pick a fine-tuned model you can trust
Avoid the Llama variant zoo. Keep one default Instruct stack, match the model card to the file, and remember license and Acceptable Use still follow the weights.
First useful tasks to try with a Llama model
Start Llama with bounded tasks and a verify step. Quote-check summaries, run the transform yourself, and never paste unchecked SLA numbers into a steerco slide.
Should you use hosted Llama or run it yourself?
Choose hosted Llama or self-host on purpose. Local files stay local only if you do not flip a cloud toggle or paste into a logged-in host tab.
Which Llama model size fits your laptop, and which needs a server?
Match Llama sizes to real RAM and VRAM. Scout-class fits many laptops; huge MoE names need servers. Write the hardware ticket from a load test, not from a…
