@10xmylife: Cui Tianyi has joined DeepSeek! When I saw his blog during undergrad, I was amazed—how could someone be so talented? Looking forward to DeepSeek's products in the future.
Summary
Cui Tianyi (former quantitative specialist at Jane Street) joined DeepSeek in March 2025 as the head of the newly formed Harness team, developing a desktop agent product that competes with Anthropic's Claude Code.
View Cached Full Text
Cached at: 05/22/26, 05:51 AM
Cui Tianyi has joined DeepSeek. I remember reading his blog during my undergraduate years and being absolutely amazed—how could someone be this brilliant? Looking forward to DeepSeek’s products in the coming months.
Weird Thoughts (@0xLogicrw): Cui Tianyi, former co-founder of quantitative investment firm TSY Capital and a 9-year veteran at Wall Street quant giant Jane Street, officially joined DeepSeek in March this year. He now leads the newly established Harness team, spearheading the development of a desktop Agent product that directly competes with Anthropic’s Claude Code.
During his undergraduate studies, Cui Tianyi won the ACM competition six times.
Similar Articles
Antirez Deepseek 4.1 flash gguf on HF
Antirez has uploaded the quantized gguf version of Deepseek 4.1 flash model to Hugging Face, with partial availability and questions on usage.
@xiaomovps: Found a way to use Deepseek V4.1 Flash for free Workbuddy international version has synced and launched Deepseek V4.1 F…
Workbuddy international version has launched Deepseek V4.1 Flash, allowing free access without queuing and with fast speed.
@0x0SojalSec: You don’t need 256GB+ RAM to run a DeepSeek-V4.1-Flash 763B model locally, - Only need 64GB system RAM. - offloaded a 2…
The article explains how to run the DeepSeek-V4.1-Flash 763B model locally with only 64GB RAM by offloading a 200GB Engram to NVMe and using DSpark, achieving 200 TPS on 4 Max-Q cards.
Livebench added Deepseek v4.1 flash
Livebench has added Deepseek v4.1 flash to its evaluations, showing it performs on par with 5.6 sol at less than one-tenth the cost.
@joey00072fp4: deepseek v4.1 flash, two stage decoder arch, 20+20=40 layers first 20 layers build global kv of csa2 and swa, (called e…
DeepSeek-V4.1-Flash introduces a two-stage decoder architecture with 40 layers, activating only 8B parameters during prefill and 16B during decode, and includes 196B Engram memory for significant efficiency gains over previous versions.