Built in the Blind Spot

👀 You're watching someone else's cut. Paste your own URL to direct one you can tune.

Press play to listen. Free to change, as often as you like.

Early access · normally $10

One payment covers your export: the watermark-free 1920×1080 MP4 of this cut. Re-downloads are free forever. Watching stays free.

Researched from the web. Every claim traces to a source you can open:

1. DeepSeek was founded in 2023 by Liang Wenfeng, who co-founded the quantitative hedge fund High-Flyer in 2015 with college classmates and funded DeepSeek in part with the fund's proceeds.Forbes · 2025
2. Liang began buying thousands of Nvidia graphics processors in 2021 for an unnamed AI side project, before the Biden administration restricted AI chip exports to China, and acquaintances at the time regarded it as a quirky hobby.Fortune (citing the Financial Times) · 2025-01-27
3. High-Flyer, the parent hedge fund incorporated in February 2016 and headquartered in Hangzhou, listed roughly US$10 billion in assets under management as of December 2025.Wikipedia (High-Flyer) · 2025-12 · medium confidence
4. DeepSeek released its free open-source V3 model in late 2024, saying it was trained with less powerful chips at a fraction of the cost of OpenAI or Google models, then released the R1 reasoning model in January 2025.CNBC · 2026-01-06
5. R1 was released on 20 January 2025, the same day as President Trump's second inauguration, and by 27 January DeepSeek's iPhone app had overtaken ChatGPT as the most-downloaded free app on the US Apple App Store.CSIS · 2025
6. Nvidia's stock fell about 17% from Friday's close to Monday's close (from $142.62 to $118.58 a share), wiping nearly $600 billion off its market capitalisation.TechCrunch · 2025-01-27
7. The sell-off extended beyond Nvidia: Broadcom fell 11% and Microsoft 7.5% between the 24 January and 3 February 2025 closes.IoT Analytics · 2025-02 · medium confidence
8. On 27 January 2025 DeepSeek said it was temporarily limiting new user registrations because of what it called large-scale malicious attacks on its services.CNBC · 2025-01-27
9. App analytics cited by Bloomberg showed the DeepSeek assistant topped download charts across 140 markets, with 16 million downloads in its first 18 days, roughly double the 9 million ChatGPT recorded at its launch.Gulf News / Bloomberg · 2025-02 · medium confidence
10. Nvidia CEO Jensen Huang publicly argued the market misread DeepSeek, calling the R1 open-source reasoning model "incredibly exciting" and saying inference and post-training still require heavy compute.TechCrunch · 2025-02-21
11. In a peer-reviewed Nature paper published on 17 September 2025, DeepSeek disclosed for the first time that R1's training cost about $294,000 using 512 Nvidia H800 chips, on top of roughly $6 million spent on the underlying base model.Investing.com (reporting on the Nature paper) · 2025-09-18
12. R1 is regarded as the first major large language model to go through formal academic peer review, a step Nature framed as a counterweight to AI industry hype.HyperAI (summarising Nature) · 2025-09 · medium confidence
13. In supplementary material to the Nature paper, DeepSeek acknowledged for the first time that it owns Nvidia A100 chips, used in preparatory experiments with a smaller model.Investing.com · 2025-09-18
14. Italy's data protection authority, the Garante, ordered DeepSeek to block its chatbot in Italy in late January 2025 after finding its responses on privacy inadequate, and South Korea's PIPC suspended new app downloads in mid-February 2025 after the company acknowledged partially neglecting Korean daReuters (via Yahoo) · 2025-02-17
15. Australia barred DeepSeek from government devices and systems on 4 February 2025, and Taiwan banned government departments from using the service over security and data-transfer concerns.Malay Mail / Reuters roundup · 2025-06-29 · medium confidence
16. DeepSeek's South Korean service became available again at the end of April 2025 after privacy-related changes.Malay Mail / Reuters roundup · 2025-06-29 · medium confidence
17. Unlike January 2025, DeepSeek's later releases did not repeat the market shock: Nvidia became the first company to reach a $5 trillion valuation in October 2025, Broadcom's shares rose 49% across 2025 and ASML's 36%.CNBC · 2026-01-06
18. In mid-June 2026 DeepSeek closed its first-ever external funding round, raising more than 50 billion yuan (about $7.4 billion) at a valuation above $50 billion, having been self-funded until then.Asia Financial (citing WSJ and The Information) · 2026-06-17 · medium confidence
19. By late August 2026 DeepSeek was reported to be closing a further raise of about 50 billion yuan at a roughly $74 billion pre-money valuation, with participation from existing backers Monolith, Shixiang Capital and battery maker CATL, ahead of a possible Shanghai STAR Market listing filed as early aTech Startups (citing SCMP and WSJ) · 2026-08-28 · medium confidence

Paste this where you want the video: it autoplays silently, loops, and plays sound on tap:

Or share the link; it opens this exact cut, fullscreen: