The story of DeepSeek is often told from R1.
But it has taken Liang Wenfeng and his team more than 20 years to get here. Machine vision, programmed trading and magic square quantification are the first half of this road; later, they brought the team into basic AI research and went to DeepSeek. This video explains this path.
Turned Deepseek Harness into a APP on the computer.
Recipe -
Top-5 Best Value #LLM Models of the Day at UTC-04
| Model | #AAII | Price | | Gemini 3.7 Flash | 56 | $0.29 | | DeepSeek V4 Flash 0731 | 52 | $0.06 | | GPT-5.6 Luna | 52 | $0.18 | | Gemini 3.6 Flash | 52 | $0.57 | | GLM 5.2 | 53 | $0.71 |
Deepseek harness just beat hermes at the one thing i didn’t expect
I ran both agents side by side inside the same Agentic OS — and the winner depends entirely on the job. Here’s where the gap showed up: → DeepSeek Harness hit 136,000 GitHub stars in 2 days → Its entire
You have reached the end of the archive
All of deepseek