Waa Maxay DeepSeek? Oo OPENAI Iyo Maraykanba ka Qaylisiiyay

Videogan waxaan ku qeexayaa waa maxay DeepSeek iyo waxa uu kaga duwan yahay models-kii hore, waxaana loogu talagalay qofka rabo inuu AI-ga computer-kiisa ku run-gareeyo. Farqiga ugu weyn waa reasoning-ka: models-kii hore su'aal markaad weydiiso jawaabta ayay iska soo tufi jireen, kan se intuusan jawaabta ku siin ayuu ka fikirayaa, isla markaana is saxayaa.

Waxaan ku tusayaa sida Ollama loogu run-gareeyo, waxaanan isticmaalayaa 8 billion-ka. Billions-ku waa parameters-ka lagu tababaray model-ka, waxayna go'aaminayaan awoodda uu kombuyuutarkaaga ka rabo. Haddii VRAM-kaagu yar yahay, 1.5 billion-ka ayaa kuu habboon. Waxaan weydiinayaa xisaab fudud, waxaanan ku tusayaa thinking-ka iyo solution-ka oo kala go'an. Kadib waxaan ka codsanayaa code React ah ee GitHub API-ga wacaya, wuxuuna ii soo celiyaa component, loading, error handling iyo sharaxaad.

Ugu dambayn waxaan ka hadlayaa buuqa uu DeepSeek dhaliyay. Shirkado waaweyn ayaa sheegi jiray in models-ka noocan ah ee lagu tababaro kharash aan la koobi karin, kadibna waxaa soo baxay koox Shiineys ah ee isla shaqada side project ka dhigtay, taasoo suuqa gilgishay ilaa cinwaannada adduunka. Waxaan sidoo kale idin tusayaa Chatbox haddaadan doonayn inaad interface gaar ah dhisato, iyo sida developer-ku uu u dooran karo inuu isagu interface-kiisa dhisto.

A 12-minute Somali (Af-Soomaali) explainer on DeepSeek and running AI locally. It covers what reasoning or thinking models do differently from earlier models, what the parameter counts (1.5B, 7B, 8B) mean, and how VRAM and RAM decide which model your computer can run. The host installs Ollama, runs DeepSeek locally with no internet connection, shows the visible thinking step on a simple maths question, then asks the model for React code that calls the GitHub API and reviews the generated component, loading state and error handling. It also explains the market shock DeepSeek caused and mentions Chatbox as a ready-made client. From Dugsiiye, free on YouTube.