# deepseek: the idea

> the whale-funds-the-lab model — the lineage deepseek inherited, what it proved, and what it left unresolved.

## lineage

- 1982–2021 — the renaissance template: renaissance technologies is the ancestor liang read himself — he wrote the preface to the chinese edition of zuckerman's simons biography in 2021: simons 'went through countless sleepless nights' before his model worked, and 'every time i face a difficulty at work, i think of simons's words.' quant profits funding fundamental research, on the record. (sources: [liangs preface to zuckermans simons book (2021)](https://www.businesstimes.com.sg/wealth/deepseek-just-latest-hedge-fund-innovation))
- 2002–2015 — the zju network: the zhejiang-university pipeline — the whole founding cohort is zju engineering: liang's meng cohort, xu jin's chu kochen mixed-honors class, the classmates who started automating trades in 2008. an institutional network, not a talent market. (sources: [zju page: xu jin bio — chu kochen mixed class, signal-processing phd, yutu rover nav](https://www.zju.edu.cn/2017/0608/c33485a514822/pagem.htm), [the jacobi-to-high-flyer history](https://www.bianews.com/news/details?id=205377), [2008 stock-dabbling trio](https://techcentral.co.za/liang-wenfeng-built-deepseek-hedge-fund/258390/), [thesis title, advisor xiang zhiyu, chengdu years](https://interestingengineering.com/engineers-directory/liang-wenfeng))
- 2015–2021 — the fund as patron: the fund as patron — a quant fund that had already gone all-ai by 2017, passed rmb 10b aum by 2019, and peaked near rmb 100b in 2021 generated the profits that bought the fleet. deepseek never needed a venture round because the whale was already winning. (sources: [official timeline 2008–2022](https://www.high-flyer.cn/history/), [one of nine speakers](https://www.reuters.com/technology/deepseek-founder-liang-wenfeng-puts-focus-chinese-innovation-2025-01-28/), [may-8 round launch](https://www.coinlive.com/news/deepseek-s-liang-wenfeng-we-don-t-want-alibaba-s-money-we-ll-contribute))
- 2020–2022 — infrastructure first: the infrastructure first — fire-flyer i (2020) and ii (2021) predate the lab by years. the cluster, the 3fs filesystem, hfreduce, and the time-sharing scheduler were built for trading; they became the lab's substrate. compute preceded the research org. (sources: [the hfai blog: the 10k-a100 cluster](https://www.high-flyer.cn/blog/hfai/), [fire-flyer ai-hpc paper](https://arxiv.org/abs/2408.14158), [fire-flyer ii stats (8tb/s read, feb-2022 stats)](https://www.high-flyer.cn/en/), [official timeline 2008–2022](https://www.high-flyer.cn/history/))
- 2023–2025 — the open-weights movement: the open-weights precedent — meta's llama line proved that open weights could shape an ecosystem; deepseek's mit-licensed r1 radicalized the position, and meta's own retreat to closed frontier models left the lab as the movement's exemplar. (sources: [the chinese open-weights landscape](https://www.turingpost.com/p/chinesemodels), [american standard open-source quote](https://www.constellationr.com/insights/news/meta-deepseek-custom-silicon-ai-optimizing-engineering-and-business), [second deepseek moment — kimi k2](https://xpert.digital/en/sputnik-the-moment-of-artificial-intelligence/))
- 2024 — the techno-idealist doctrine: the techno-idealist doctrine — the july 2024 interview supplies the theory in the founder's words: originality over imitation, open source as 'a cultural act more than a commercial one,' the organization as the moat. the ideas were published before the market tested them. (sources: [the canonical techno-idealism interview (于丽丽/暗涌)](https://www.36kr.com/p/2872793466982535), [full-text mirror used for verbatim quotes](https://wallstreetcn.com/articles/3719982), [annotated mirror](https://www.laojinchuhai.com/zh/classics/liang-wenfeng-anchorage-2024))
- 2022– — the embargo as engine: the embargo as engine — us export controls (october 2022, tightened october 2023) postdate the fleet's purchase and predate the lab's scaling. liang: 'our problem has never been money, it's the embargo.' constraint became the design brief. (sources: [allen's house testimony: the $5.576m math](https://republicans-science.house.gov/_cache/files/0/d/0d8e8a6c-2a09-413d-9d78-e21df078bc9e/EAB890C6D96E33647BC244A4EFCC742660C0EE09981A7B46E6BC1EC8C80F5386.gregory-c.-allen-testimony.pdf), [state dept official allegation](https://asiatimes.com/2025/06/deepseek-gets-nvidias-high-end-gpus-via-singapore-us-official/), [full-text mirror used for verbatim quotes](https://wallstreetcn.com/articles/3719982), [how deepseek trained on crippled hardware](https://www.nextplatform.com/ai/2025/01/27/how-did-deepseek-train-its-ai-model-on-a-lot-less-and-crippled-hardware/1656116))

## what deepseek proved

- the whale can fund the lab: a quant fund can finance a frontier lab — no venture round, no cap-table dilution, no product mandate. fire-flyer ii was bought with fund profits; the lab ran for three years on the whale's cash flow before its first external raise in 2026. (sources: [official timeline 2008–2022](https://www.high-flyer.cn/history/), [may-8 round launch](https://www.coinlive.com/news/deepseek-s-liang-wenfeng-we-don-t-want-alibaba-s-money-we-ll-contribute), [law-firm equity-structure analysis](https://dehehantong.com/en/news/articles/156), [first-ever liang interview](https://m.36kr.com/p/2272896094586500))
- the embargo engineered efficiency: constraint produces efficiency — 2.788m h800 gpu-hours for v3's final run, mla cutting memory, moe activating 37b of 671b, the fire-flyer paper matching dgx-a100 throughput at ~half cost. sanctioned hardware forced the engineering that became the headline. (sources: [allen's house testimony: the $5.576m math](https://republicans-science.house.gov/_cache/files/0/d/0d8e8a6c-2a09-413d-9d78-e21df078bc9e/EAB890C6D96E33647BC244A4EFCC742660C0EE09981A7B46E6BC1EC8C80F5386.gregory-c.-allen-testimony.pdf), [fire-flyer ai-hpc paper](https://arxiv.org/abs/2408.14158), [how deepseek trained on crippled hardware](https://www.nextplatform.com/ai/2025/01/27/how-did-deepseek-train-its-ai-model-on-a-lot-less-and-crippled-hardware/1656116), [the teardown: $5.576m real but narrow](https://www.theteardown.co/deepseek))
- open weights as distribution: open weights move markets and ecosystems — r1 under mit hit #1 in 52 countries, landed in azure ai foundry within days, and reset the competitive frame so completely that 'the second deepseek moment' became the label for everyone else's open releases. (sources: [#1 us app store jan 26](https://techcrunch.com/2025/01/27/deepseek-displaces-chatgpt-as-the-app-stores-top-app/), [r1 in azure ai foundry + github catalog](https://azure.microsoft.com/en-us/blog/deepseek-r1-is-now-available-on-azure-ai-foundry-and-github/), [second deepseek moment — kimi k2](https://xpert.digital/en/sputnik-the-moment-of-artificial-intelligence/), [the chinese open-weights landscape](https://www.turingpost.com/p/chinesemodels))
- pricing as a weapon: cost-plus pricing works as strategy — v2 at ~rmb 1/m tokens started a price war that cut chinese llm prices 80–97%; the theoretical-545%-margin disclosure showed the headroom was real. the lab priced like the fund: costs plus margin, not scarcity pricing. (sources: [full-text mirror used for verbatim quotes](https://wallstreetcn.com/articles/3719982), [doubao at ¥0.0008/1k tokens; 99.3% cut](https://techflowpost.com/en-US/article/17993), [545% theoretical](https://www.scmp.com/tech/big-tech/article/3300734/deepseek-shows-power-v3-r1-models-theoretical-545-profit-margin), [english mirror of the disclosure](https://gigazine.net/gsc_news/en/20250303-deepseek-ai-cost-profit/))
- the homegrown bench: young homegrown teams can do frontier work — 'no unfathomable geniuses,' no overseas returnees on v2, bottom-up staffing and unbounded cluster access; the result was grpo, the v3 architecture, and the first major llm through nature peer review. (sources: [full-text mirror used for verbatim quotes](https://wallstreetcn.com/articles/3719982), [~150 staff; the gold-medal hiring rep](https://www.donga.com/news/article/all/20250211/131006164/1), [r1 peer-reviewed](https://www.nature.com/articles/s41586-025-09422-z), [guos own page: ai researcher in deepseek, july 2023–april 2026 + core-contributor lists](https://guoday.github.io/))
- the papers are the product: papers and infrastructure are the product — the v3/r1 reports, the fire-flyer paper, open source week's six repos, and nature peer review built the lab's credibility without a marketing org. 'letting the research reports speak' was the whole comms strategy. (sources: [letting research reports speak](https://tthmx.com/hangye/90274.html), [open source week index](https://github.com/deepseek-ai/open-infra-index), [r1 peer-reviewed](https://www.nature.com/articles/s41586-025-09422-z), [fire-flyer ai-hpc paper](https://arxiv.org/abs/2408.14158))

## what stayed unfinished

- r2 never shipped: r2 never shipped — reported stalled in june 2025 (liang unsatisfied), then reportedly failed on huawei ascend hardware in august. the r-line that produced the shock never produced a successor; the v-line absorbed the flagship role. (sources: [deepseek-r2 launch stalled; ceo balks](https://www.reuters.com/world/china/deepseek-r2-launch-stalled-ceo-balks-progress-information-reports-2025-06-26/), [ft: ascend training failure delayed r2](https://www.reuters.com/world/china/deepseeks-launch-new-ai-model-delayed-by-huawei-chip-issues-ft-reports-2025-08-14/), [r2 status: unannounced, unreleased as of jul-2026](https://www.layer3labs.io/guides/deepseek-r2-explained))
- the ascend wall: the domestic-chip training path stalled — the ascend run failed even with huawei engineers onsite; the lab settled into nvidia-for-training, ascend-for-inference, and months of v4 delay for ascend optimization. the hedge exists; the substitution doesn't. (sources: [ft: ascend training failure delayed r2](https://www.reuters.com/world/china/deepseeks-launch-new-ai-model-delayed-by-huawei-chip-issues-ft-reports-2025-08-14/), [huawei onsite team](https://www.theregister.com/software/2025/08/14/dodgy_huawei_chips_nearly_sunk_deepseeks_next_gen_r2_mode/753616), [v3.2-exp ships day-1 ascend/cann support](https://www.tomshardware.com/tech-industry/deepseek-new-model-supports-huawei-cann), [v4 delayed for ascend-tuned stack (yuyuantantian)](https://www.bloomberg.com/news/articles/2026-04-26/deepseek-v4-delay-shows-shift-to-china-chips-cctv-account-says))
- research met revenue: 'research, not products' met the market — the app that beat chatgpt sits #2 in china behind doubao on questmobile's august 2025 mau, and the august 2026 api price hikes (up to ~12x on cache hits) read like incumbent margin management. (sources: [questmobile aug 2025: doubao 157m vs deepseek 143m mau](https://www.scmp.com/tech/big-tech/article/3325858/bytedance-chatbot-doubao-still-chinas-most-popular-ai-app-rival-deepseek-loses-users), [up-to-12x cache-hit hike](https://en.sedaily.com/international/2026/08/14/deepseek-raises-api-prices-up-to-12-fold-as-chinas-ai-firms), [the significant price-rise warning](https://thenextweb.com/news/deepseek-significant-api-price-increase), [the api changelog: v4-pro ga 2026-08-13](https://api-docs.deepseek.com/updates))
- the end of pure patronage: the self-funding purity ended — the lab that 'never had a financing problem' opened its first external round in may 2026, liang injecting up to rmb 20b personally, state capital reported the likely lead. the whale still funds; it no longer funds alone. (sources: [may-8 round launch](https://www.coinlive.com/news/deepseek-s-liang-wenfeng-we-don-t-want-alibaba-s-money-we-ll-contribute), [alibaba collapse](https://www.gate.com/news/detail/whale-lab-deepseek-and-alibabas-fundraising-negotiations-failed-to-reach-an-20949731), [the may 2026 round; the fewest-conditions stance](https://www.chaincatcher.com/en/article/2263550), [the first cfo: yan wentao](https://m.163.com/dy/article/L6PV3CAP05568W0A.html))
- the bench got poached: the flat org got poached — no titles and bottom-up staffing built the bench, then bytedance seed, tencent hunyuan, xiaomi, and deeproute hired it: ~10 of the v4 report's ~300 authors are now former employees, including grpo co-creator guo daya. (sources: [v4 report ~300 authors](https://www.newsglobenow.com/new346166.html), [departures list incl](https://substack.com/@sinocism/note/c-245366529), [corroboration of 10 departures](https://www.gogoai.xin/article/deepseek-loses-10-key-researchers-as-chinas-ai-talent-war-intensifies), [guos own page: ai researcher in deepseek, july 2023–april 2026 + core-contributor lists](https://guoday.github.io/))

## the arguments

- (for) the deepseek pattern is the new lab model: a profitable patron (quant fund, cloud giant, ad business) finances frontier research, open weights buy the ecosystem, and papers carry the marketing. kimi, qwen, and zhipu all read as variations on it. (sources: [may-8 round launch](https://www.coinlive.com/news/deepseek-s-liang-wenfeng-we-don-t-want-alibaba-s-money-we-ll-contribute), [the chinese open-weights landscape](https://www.turingpost.com/p/chinesemodels), [second deepseek moment — kimi k2](https://xpert.digital/en/sputnik-the-moment-of-artificial-intelligence/), [first-ever liang interview](https://m.36kr.com/p/2272896094586500))
- (against) the pattern is not portable — it required a fund that peaked at ~rmb 100b, a 10k-gpu purchase timed before controls nobody could have planned for, and a founder who codes. copying the org chart copies the least load-bearing part. (sources: [one of nine speakers](https://www.reuters.com/technology/deepseek-founder-liang-wenfeng-puts-focus-chinese-innovation-2025-01-28/), [full-text mirror used for verbatim quotes](https://wallstreetcn.com/articles/3719982), [fire-flyer ai-hpc paper](https://arxiv.org/abs/2408.14158), [optimization not revolution counter-take](https://www.ifri.org/en/memos/deepseek-moment))
- (nuance) open weights were both the product and the vulnerability — the ecosystem made deepseek inevitable, and the same openness made distillation accusations, uncensored rehosts, and self-hosted deployments outside its control. the strategy and the exposure are the same artifact. (sources: [#1 us app store jan 26](https://techcrunch.com/2025/01/27/deepseek-displaces-chatgpt-as-the-app-stores-top-app/), [srinivas: us/eu-hosted, uncensored](https://mspoweruser.com/you-can-now-try-uncensored-deepseek-r1-via-perplexity-ps-its-us-hosted/), [openai evidence claim](https://www.theverge.com/news/601195/openai-evidence-deepseek-distillation-ai-data), [anthropics own account: 24k accts/16m exchanges](https://www.anthropic.com/news/detecting-and-preventing-distillation-attacks))
- (nuance) the $5.576m number did the damage and the distortion — real and correctly scoped in the report, then flattened into 'r1 cost $6m' and a $589b nvidia repricing. the lesson is that markets priced a narrative about cost, not the cost. (sources: [allen's house testimony: the $5.576m math](https://republicans-science.house.gov/_cache/files/0/d/0d8e8a6c-2a09-413d-9d78-e21df078bc9e/EAB890C6D96E33647BC244A4EFCC742660C0EE09981A7B46E6BC1EC8C80F5386.gregory-c.-allen-testimony.pdf), [the teardown: $5.576m real but narrow](https://www.theteardown.co/deepseek), [nvidia −17%/−$600b single-biggest one-day loss](https://web.archive.org/web/20250129002522/https:/www.cbsnews.com/news/what-is-deepseek-ai-china-stock-nvidia-nvda-asml/), [apples-to-oranges cost accounting](https://creativestrategies.com/dispelling-deepseek-myths-studying-v3/))
- (against) the state adjacency is load-bearing now — premier's symposium, xi's symposium, zhejiang investor screening, state capital reported leading the raise. the lab that wanted to be judged on papers is becoming national infrastructure, which protects it and binds it. (sources: [jan-20 li qiang symposium](https://www.scmp.com/tech/policy/article/3295662/beijing-meeting-puts-spotlight-chinas-new-face-ai-deepseek-founder-liang-wenfeng), [feb-17 xi symposium confirmed via yuyuantantian](https://www.thestandard.com.hk/finance/article/227878/DeepSeek-founder-confirmed-in-the-meeting-with-Xi), [travel curbs + zhejiang investor screening](https://techcrunch.com/2025/03/14/china-is-reportedly-keeping-deepseek-under-close-watch/), [the may 2026 round; the fewest-conditions stance](https://www.chaincatcher.com/en/article/2263550), [deepseek unmasked](https://chinaselectcommittee.house.gov/sites/evo-subsites/selectcommitteeontheccp.house.gov/files/evo-media-document/DeepSeek%20Final.pdf))
- (for) the efficiency story is also a demand story — nadella's jevons counter held: cheaper inference expanded usage, and the lab's own price hikes and capacity strain followed. efficiency didn't shrink the compute race; it re-priced it. (sources: [nadellas jevons post verbatim](https://fortune.com/2025/01/27/microsoft-ceo-satya-nadella-deepseek-optimism-jevons-paradox/), [nadella context](https://www.geekwire.com/2025/microsoft-ceo-says-ai-use-will-skyrocket-with-more-efficiency-amid-craze-over-deepseek/), [energy-efficiency paradox](https://theconversation.com/deepseek-claims-to-have-cured-ais-environmental-headache-the-jevons-paradox-suggests-it-might-make-things-worse-248720), [up-to-12x cache-hit hike](https://en.sedaily.com/international/2026/08/14/deepseek-raises-api-prices-up-to-12-fold-as-chinas-ai-firms))

## successors

- [moonshot / kimi](https://www.moonshot.ai): the 'second deepseek moment' — kimi k2 thinking beating gpt-5 on hle and browsecomp made the open-weights frontier a chinese category, not a deepseek monopoly. (sources: [second deepseek moment — kimi k2](https://xpert.digital/en/sputnik-the-moment-of-artificial-intelligence/), [k2 thinking beats gpt-5 on hle/browsecomp](https://www.linkedin.com/posts/conorgrennan_breaking-kimi-k2-may-be-chinas-second-deepseek-activity-7392592898703278080-E8tl))
- [alibaba / qwen](https://qwen.ai): the platform-scale variant — qwen's open-weights line pairs the lab strategy with a cloud business, and its ecosystem terms reportedly killed its deepseek investment talks. (sources: [the chinese open-weights landscape](https://www.turingpost.com/p/chinesemodels), [alibaba collapse](https://www.gate.com/news/detail/whale-lab-deepseek-and-alibabas-fundraising-negotiations-failed-to-reach-an-20949731))
- [zhipu / glm](https://www.z.ai): the academic-lab survivor — glm-4 keeps a third chinese open-weights lineage in the frontier set the landscape coverage tracks. (sources: [the chinese open-weights landscape](https://www.turingpost.com/p/chinesemodels), [top-16 chinese open-source list](https://eu.36kr.com/en/p/3404902313217417))
- [minimax](https://www.minimaxi.com): the fellow accused — anthropic's february 2026 disclosure names minimax beside deepseek in the distillation campaigns; the cohort shares both the strategy and the scrutiny. (sources: [anthropics own account: 24k accts/16m exchanges](https://www.anthropic.com/news/detecting-and-preventing-distillation-attacks), [the anthropic distillation figures](https://www.cnbc.com/2026/02/24/anthropic-openai-china-firms-distillation-deepseek.html))
- [mistral](https://mistral.ai): the european open-weights control case — same license-first strategy without the whale behind it; venture-funded where deepseek was fund-funded. (sources: [the chinese open-weights landscape](https://www.turingpost.com/p/chinesemodels))
- [meta / llama](https://ai.meta.com): the patron who retreated — llama defined the open-weights position deepseek radicalized; the war-room reporting and capex pledge trace the incumbent's answer. (sources: [american standard open-source quote](https://www.constellationr.com/insights/news/meta-deepseek-custom-silicon-ai-optimizing-engineering-and-business), [strengthened our conviction](https://www.theverge.com/news/602233/meta-deepseek-zuckerberg-earings-q4), [hundreds of billions capex pledge](https://techcrunch.com/2025/01/29/zuck-shrugs-off-deepseek-vows-to-spend-hundreds-of-billions-on-ai/))

## open questions

- what is the true compute inventory — the documented 10k a100s and self-reported 2,048 h800s, or semianalysis's ~50k hopper-class estimate, or something the commerce probe hasn't reached?
- what is the lab's total spend — r&d budget, cumulative training cost, and the fund's losses or profits funding it are all undisclosed; the $5.576m and $294k figures cover scoped runs only.
- did r2 die or get folded — was the reasoning line merged into the v-series, or does a stalled successor still exist internally?
- what did the 2026 round actually close at — ~rmb 300b open, ~$52b per forbes, higher headline figures elsewhere — and on whose terms, given alibaba's collapse and tencent's declined ~20%?
- does the ascend hedge ever become the primary path — v4's months of delay for domestic-chip optimization is reported; whether the next flagship can train without nvidia is the open question the embargo keeps asking.
- will the distillation accusations ever be evidenced — openai's ft claim was never publicly substantiated; anthropic's 2026 disclosure gave numbers but no technical artifact, and deepseek's nature-exchange denial stands on the record.
- can the flat org survive the raise — external capital, a cfo, and ipo prep against a culture built on 'no titles, anyone calls the cluster,' with ~10 v4 authors already gone.
- does the whale's fund keep compounding — +56.55% in 2025 on >rmb 70b, after the 2021 drawdown apology and the 2024 neutral-strategy exit; the patron's durability is the lab's runway.

AI-drafted at Ben Guo's direct request and credited to Hraness; every claim links to its cataloged source.

Canonical page: https://hraness.com/deepseek/the-idea


Published by @ben in profile revision 1.
Original document: https://hraness.com/deepseek/the-idea
Original credited author: Hraness
AI-drafted at Ben Guo's direct request and credited to Hraness; every claim links to its cataloged source.

## Structured source and record data

````json
{"anchors":[{"id":"lineage","originalUrl":"https://hraness.com/deepseek/the-idea#lineage","title":"lineage"},{"id":"proved","originalUrl":"https://hraness.com/deepseek/the-idea#proved","title":"what deepseek proved"},{"id":"stalled","originalUrl":"https://hraness.com/deepseek/the-idea#stalled","title":"what stayed unfinished"},{"id":"arguments","originalUrl":"https://hraness.com/deepseek/the-idea#arguments","title":"the arguments"},{"id":"successors","originalUrl":"https://hraness.com/deepseek/the-idea#successors","title":"successors"},{"id":"open-questions","originalUrl":"https://hraness.com/deepseek/the-idea#open-questions","title":"open questions"}],"provenance":{"draftingDisclosure":"AI-drafted at Ben Guo's direct request and credited to Hraness; every claim links to its cataloged source.","originalAuthor":"Hraness","originalRevision":"67a31601fbb25e7d6587faae5bc883e2ddebf7571588a3642fe0216994378ece","originalUrl":"https://hraness.com/deepseek/the-idea","source":"hraness"},"records":[{"id":"record-72513e022df6b76003b9","kind":"deepseek-agent-era-arguments","label":"the deepseek pattern is the new lab model: a profitable patron (quant fund, cloud giant, ad business) finances frontier research, open weights buy the ecosystem, and papers carry the marketing. kimi, qwen, and zhipu all read as variations on it.","original":{"claim":"the deepseek pattern is the new lab model: a profitable patron (quant fund, cloud giant, ad business) finances frontier research, open weights buy the ecosystem, and papers carry the marketing. kimi, qwen, and zhipu all read as variations on it.","id":"arg-patron-model","side":"for","sourceIds":["src-hp-coinlive-funding","src-ci-turingpost-chinesemodels","src-ci-xpert-kimi","src-hp-36kr-interview1"]},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-db362ca467309d55b804","source-b151d813e9817329df15","source-7929874beb28e643e609","source-07eaa7564c5f63b78585"]},{"id":"record-126e64ecab3b34ea0fac","kind":"deepseek-agent-era-arguments","label":"the pattern is not portable — it required a fund that peaked at ~rmb 100b, a 10k-gpu purchase timed before controls nobody could have planned for, and a founder who codes. copying the org chart copies the least load-bearing part.","original":{"claim":"the pattern is not portable — it required a fund that peaked at ~rmb 100b, a 10k-gpu purchase timed before controls nobody could have planned for, and a founder who codes. copying the org chart copies the least load-bearing part.","id":"arg-not-portable","side":"against","sourceIds":["src-hp-reuters-liang","src-hp-wallstreetcn-interview2","src-hp-fireflyer-paper","src-ci-ifri-deepseek-moment"]},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-5bb15e05d2d051f06e4f","source-c9bfc1101f4c8e3a80cf","source-ee048b67b69c4d38c620","source-0f036aee238510dc3dd5"]},{"id":"record-86a176713ef61e93c8d7","kind":"deepseek-agent-era-arguments","label":"open weights were both the product and the vulnerability — the ecosystem made deepseek inevitable, and the same openness made distillation accusations, uncensored rehosts, and self-hosted deployments outside its control. the strategy and the exposure are the same artifact.","original":{"claim":"open weights were both the product and the vulnerability — the ecosystem made deepseek inevitable, and the same openness made distillation accusations, uncensored rehosts, and self-hosted deployments outside its control. the strategy and the exposure are the same artifact.","id":"arg-openness-both","side":"nuance","sourceIds":["src-hp-techcrunch-appstore","src-ci-mspoweruser-perplexity","src-ci-verge-distillation","src-hp-anthropic"]},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-acebedb59dd691984f43","source-a9178d68fc3e339ab065","source-1a377fcc410e5e97f373","source-0211196e9ddb845746f9"]},{"id":"record-e19a0c2b71876f7a8598","kind":"deepseek-agent-era-arguments","label":"the $5.576m number did the damage and the distortion — real and correctly scoped in the report, then flattened into 'r1 cost $6m' and a $589b nvidia repricing. the lesson is that markets priced a narrative about cost, not the cost.","original":{"claim":"the $5.576m number did the damage and the distortion — real and correctly scoped in the report, then flattened into 'r1 cost $6m' and a $589b nvidia repricing. the lesson is that markets priced a narrative about cost, not the cost.","id":"arg-narrative-pricing","side":"nuance","sourceIds":["src-hp-allen-testimony","src-ci-teardown","src-hp-cbs-nvidia","src-ci-creativestrategies-myths"]},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-807def74933e03156976","source-acdf9ae0c597d9d7c8fe","source-79c1aa742d9e9d5a21d6","source-b73bdd79c86c1e2c64de"]},{"id":"record-68e0b41aac3aef6596ad","kind":"deepseek-agent-era-arguments","label":"the state adjacency is load-bearing now — premier's symposium, xi's symposium, zhejiang investor screening, state capital reported leading the raise. the lab that wanted to be judged on papers is becoming national infrastructure, which protects it and binds it.","original":{"claim":"the state adjacency is load-bearing now — premier's symposium, xi's symposium, zhejiang investor screening, state capital reported leading the raise. the lab that wanted to be judged on papers is becoming national infrastructure, which protects it and binds it.","id":"arg-state-adjacency","side":"against","sourceIds":["src-hp-scmp-liqiang","src-hp-standard-xi","src-hp-techcrunch-watch","src-hp-chaincatcher-funding","src-ci-house-select-report"]},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-c83479b1406635460658","source-aa6aee6a5efbf005dd77","source-537d6289b1b6c1dbe6d9","source-77d1012fefcd2e81ad02","source-4723c56b6f02585ff949"]},{"id":"record-4b72b31a850d1fe7da7a","kind":"deepseek-agent-era-arguments","label":"the efficiency story is also a demand story — nadella's jevons counter held: cheaper inference expanded usage, and the lab's own price hikes and capacity strain followed. efficiency didn't shrink the compute race; it re-priced it.","original":{"claim":"the efficiency story is also a demand story — nadella's jevons counter held: cheaper inference expanded usage, and the lab's own price hikes and capacity strain followed. efficiency didn't shrink the compute race; it re-priced it.","id":"arg-jevons","side":"for","sourceIds":["src-ci-fortune-jevons","src-ci-geekwire-jevons","src-ci-conversation-jevons","src-ci-sedaily-price-hike"]},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-596ae44386face52f345","source-03e27b5b0da205b6a127","source-9f31fbb840e790bc8564","source-7df40d3de7d54d968e41"]},{"id":"record-1ad8eda6397807e61724","kind":"deepseek-idea-lineage","label":"the renaissance template","original":{"contribution":"renaissance technologies is the ancestor liang read himself — he wrote the preface to the chinese edition of zuckerman's simons biography in 2021: simons 'went through countless sleepless nights' before his model worked, and 'every time i face a difficulty at work, i think of simons's words.' quant profits funding fundamental research, on the record.","era":"1982–2021","id":"lin-renaissance","name":"the renaissance template","sourceIds":["src-ci-businesstimes-hedgefund"]},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-b01af09f8668309df25a"]},{"id":"record-42752e6e46a2bad3ab04","kind":"deepseek-idea-lineage","label":"the zju network","original":{"contribution":"the zhejiang-university pipeline — the whole founding cohort is zju engineering: liang's meng cohort, xu jin's chu kochen mixed-honors class, the classmates who started automating trades in 2008. an institutional network, not a talent market.","era":"2002–2015","id":"lin-zju","name":"the zju network","sourceIds":["src-hp-zju-xujin","src-hp-bianews-history","src-hp-techcentral-liang","src-hp-ie-liang"]},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-e06948c401a48f463465","source-e85ca01f7c1f488b5018","source-777866aa5f7202981ebf","source-e642177ea7ed77782280"]},{"id":"record-f823738e6e09894396d3","kind":"deepseek-idea-lineage","label":"the fund as patron","original":{"contribution":"the fund as patron — a quant fund that had already gone all-ai by 2017, passed rmb 10b aum by 2019, and peaked near rmb 100b in 2021 generated the profits that bought the fleet. deepseek never needed a venture round because the whale was already winning.","era":"2015–2021","id":"lin-fund-as-patron","name":"the fund as patron","sourceIds":["src-hp-highflyer-history","src-hp-reuters-liang","src-hp-coinlive-funding"]},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-5eb143aeec914b6d4c1e","source-5bb15e05d2d051f06e4f","source-db362ca467309d55b804"]},{"id":"record-d65b1ad354b2c756f37f","kind":"deepseek-idea-lineage","label":"infrastructure first","original":{"contribution":"the infrastructure first — fire-flyer i (2020) and ii (2021) predate the lab by years. the cluster, the 3fs filesystem, hfreduce, and the time-sharing scheduler were built for trading; they became the lab's substrate. compute preceded the research org.","era":"2020–2022","id":"lin-infra-first","name":"infrastructure first","sourceIds":["src-hp-hfai-blog","src-hp-fireflyer-paper","src-hp-highflyer-home","src-hp-highflyer-history"]},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-e24e7de33a56f8ac2121","source-ee048b67b69c4d38c620","source-ff1ee7c6d614cc863b4d","source-5eb143aeec914b6d4c1e"]},{"id":"record-f990aa8a39d7ca1383c5","kind":"deepseek-idea-lineage","label":"the open-weights movement","original":{"contribution":"the open-weights precedent — meta's llama line proved that open weights could shape an ecosystem; deepseek's mit-licensed r1 radicalized the position, and meta's own retreat to closed frontier models left the lab as the movement's exemplar.","era":"2023–2025","id":"lin-open-weights","name":"the open-weights movement","sourceIds":["src-ci-turingpost-chinesemodels","src-ci-constellation-meta","src-ci-xpert-kimi"]},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-b151d813e9817329df15","source-987d33d02fef5b47a9de","source-7929874beb28e643e609"]},{"id":"record-24e2fc79d550b2880633","kind":"deepseek-idea-lineage","label":"the techno-idealist doctrine","original":{"contribution":"the techno-idealist doctrine — the july 2024 interview supplies the theory in the founder's words: originality over imitation, open source as 'a cultural act more than a commercial one,' the organization as the moat. the ideas were published before the market tested them.","era":"2024","id":"lin-idealism","name":"the techno-idealist doctrine","sourceIds":["src-hp-36kr-interview2","src-hp-wallstreetcn-interview2","src-hp-laojinchuhai"]},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-e2c7feee45a597771920","source-c9bfc1101f4c8e3a80cf","source-e41d366fa1f22b558c6c"]},{"id":"record-33e5396d404ca41ebc65","kind":"deepseek-idea-lineage","label":"the embargo as engine","original":{"contribution":"the embargo as engine — us export controls (october 2022, tightened october 2023) postdate the fleet's purchase and predate the lab's scaling. liang: 'our problem has never been money, it's the embargo.' constraint became the design brief.","era":"2022–","id":"lin-embargo","name":"the embargo as engine","sourceIds":["src-hp-allen-testimony","src-hp-asiatimes","src-hp-wallstreetcn-interview2","src-hp-nextplatform"]},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-807def74933e03156976","source-562cf7588945d2a05816","source-c9bfc1101f4c8e3a80cf","source-b43c57bf1bf2c459d887"]},{"id":"record-dc7a0fbe2e5cdd3e2c96","kind":"deepseek-idea-open-questions","label":"what is the true compute inventory — the documented 10k a100s and self-reported 2,048 h800s, or semianalysis's ~50k hopper-class estimate, or something the commerce probe hasn't reached?","original":{"text":"what is the true compute inventory — the documented 10k a100s and self-reported 2,048 h800s, or semianalysis's ~50k hopper-class estimate, or something the commerce probe hasn't reached?"},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":[]},{"id":"record-9fb2d95cd8f487dff911","kind":"deepseek-idea-open-questions","label":"what is the lab's total spend — r&d budget, cumulative training cost, and the fund's losses or profits funding it are all undisclosed; the $5.576m and $294k figures cover scoped runs only.","original":{"text":"what is the lab's total spend — r&d budget, cumulative training cost, and the fund's losses or profits funding it are all undisclosed; the $5.576m and $294k figures cover scoped runs only."},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":[]},{"id":"record-2b1a4c401613baf4b215","kind":"deepseek-idea-open-questions","label":"did r2 die or get folded — was the reasoning line merged into the v-series, or does a stalled successor still exist internally?","original":{"text":"did r2 die or get folded — was the reasoning line merged into the v-series, or does a stalled successor still exist internally?"},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":[]},{"id":"record-19f69bf99ce6e138bf4a","kind":"deepseek-idea-open-questions","label":"what did the 2026 round actually close at — ~rmb 300b open, ~$52b per forbes, higher headline figures elsewhere — and on whose terms, given alibaba's collapse and tencent's declined ~20%?","original":{"text":"what did the 2026 round actually close at — ~rmb 300b open, ~$52b per forbes, higher headline figures elsewhere — and on whose terms, given alibaba's collapse and tencent's declined ~20%?"},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":[]},{"id":"record-934d46de0141ec7b7821","kind":"deepseek-idea-open-questions","label":"does the ascend hedge ever become the primary path — v4's months of delay for domestic-chip optimization is reported; whether the next flagship can train without nvidia is the open question the embargo keeps asking.","original":{"text":"does the ascend hedge ever become the primary path — v4's months of delay for domestic-chip optimization is reported; whether the next flagship can train without nvidia is the open question the embargo keeps asking."},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":[]},{"id":"record-14712b7fa7b0485fb2a0","kind":"deepseek-idea-open-questions","label":"will the distillation accusations ever be evidenced — openai's ft claim was never publicly substantiated; anthropic's 2026 disclosure gave numbers but no technical artifact, and deepseek's nature-exchange denial stands on the record.","original":{"text":"will the distillation accusations ever be evidenced — openai's ft claim was never publicly substantiated; anthropic's 2026 disclosure gave numbers but no technical artifact, and deepseek's nature-exchange denial stands on the record."},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":[]},{"id":"record-75f5b158da1e780c5ff9","kind":"deepseek-idea-open-questions","label":"can the flat org survive the raise — external capital, a cfo, and ipo prep against a culture built on 'no titles, anyone calls the cluster,' with ~10 v4 authors already gone.","original":{"text":"can the flat org survive the raise — external capital, a cfo, and ipo prep against a culture built on 'no titles, anyone calls the cluster,' with ~10 v4 authors already gone."},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":[]},{"id":"record-e41ac06eca5936e17d22","kind":"deepseek-idea-open-questions","label":"does the whale's fund keep compounding — +56.55% in 2025 on >rmb 70b, after the 2021 drawdown apology and the 2024 neutral-strategy exit; the patron's durability is the lab's runway.","original":{"text":"does the whale's fund keep compounding — +56.55% in 2025 on >rmb 70b, after the 2021 drawdown apology and the 2024 neutral-strategy exit; the patron's durability is the lab's runway."},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":[]},{"id":"record-29cbdf3e308a1e846dbc","kind":"deepseek-idea-proved","label":"the whale can fund the lab","original":{"detail":"a quant fund can finance a frontier lab — no venture round, no cap-table dilution, no product mandate. fire-flyer ii was bought with fund profits; the lab ran for three years on the whale's cash flow before its first external raise in 2026.","id":"clm-proved-patron","sourceIds":["src-hp-highflyer-history","src-hp-coinlive-funding","src-hp-dehehantong-equity","src-hp-36kr-interview1"],"title":"the whale can fund the lab"},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-5eb143aeec914b6d4c1e","source-db362ca467309d55b804","source-20ac7e2113fd2e7046b5","source-07eaa7564c5f63b78585"]},{"id":"record-45a44f086c189ae53a91","kind":"deepseek-idea-proved","label":"the embargo engineered efficiency","original":{"detail":"constraint produces efficiency — 2.788m h800 gpu-hours for v3's final run, mla cutting memory, moe activating 37b of 671b, the fire-flyer paper matching dgx-a100 throughput at ~half cost. sanctioned hardware forced the engineering that became the headline.","id":"clm-proved-constraint","sourceIds":["src-hp-allen-testimony","src-hp-fireflyer-paper","src-hp-nextplatform","src-ci-teardown"],"title":"the embargo engineered efficiency"},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-807def74933e03156976","source-ee048b67b69c4d38c620","source-b43c57bf1bf2c459d887","source-acdf9ae0c597d9d7c8fe"]},{"id":"record-8fc39d017833b616e783","kind":"deepseek-idea-proved","label":"open weights as distribution","original":{"detail":"open weights move markets and ecosystems — r1 under mit hit #1 in 52 countries, landed in azure ai foundry within days, and reset the competitive frame so completely that 'the second deepseek moment' became the label for everyone else's open releases.","id":"clm-proved-open-weights","sourceIds":["src-hp-techcrunch-appstore","src-ci-azure-r1","src-ci-xpert-kimi","src-ci-turingpost-chinesemodels"],"title":"open weights as distribution"},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-acebedb59dd691984f43","source-848b616a2251fdfff2ea","source-7929874beb28e643e609","source-b151d813e9817329df15"]},{"id":"record-e7e3339e4b7a45c0862e","kind":"deepseek-idea-proved","label":"pricing as a weapon","original":{"detail":"cost-plus pricing works as strategy — v2 at ~rmb 1/m tokens started a price war that cut chinese llm prices 80–97%; the theoretical-545%-margin disclosure showed the headroom was real. the lab priced like the fund: costs plus margin, not scarcity pricing.","id":"clm-proved-pricing","sourceIds":["src-hp-wallstreetcn-interview2","src-ci-techflow-doubao","src-ci-scmp-545-margin","src-ci-gigazine-545"],"title":"pricing as a weapon"},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-c9bfc1101f4c8e3a80cf","source-19357a942001743d6668","source-b98e2e12f7e159ef11b0","source-03181c9972b7680c28a9"]},{"id":"record-7e5875843a4f6c071d17","kind":"deepseek-idea-proved","label":"the homegrown bench","original":{"detail":"young homegrown teams can do frontier work — 'no unfathomable geniuses,' no overseas returnees on v2, bottom-up staffing and unbounded cluster access; the result was grpo, the v3 architecture, and the first major llm through nature peer review.","id":"clm-proved-homegrown","sourceIds":["src-hp-wallstreetcn-interview2","src-hp-donga","src-hp-nature-r1","src-hp-guoday"],"title":"the homegrown bench"},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-c9bfc1101f4c8e3a80cf","source-f6ee4abd62d2f8da7bcc","source-7f0f229eb58f8d7e84fc","source-da3a681e38b99fe423fe"]},{"id":"record-5d07ab94539624a2cb6c","kind":"deepseek-idea-proved","label":"the papers are the product","original":{"detail":"papers and infrastructure are the product — the v3/r1 reports, the fire-flyer paper, open source week's six repos, and nature peer review built the lab's credibility without a marketing org. 'letting the research reports speak' was the whole comms strategy.","id":"clm-proved-papers","sourceIds":["src-hp-xinhua-office","src-hp-openinfra","src-hp-nature-r1","src-hp-fireflyer-paper"],"title":"the papers are the product"},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-9c18637a18012ad21bfe","source-f3a7aa842ef643e17128","source-7f0f229eb58f8d7e84fc","source-ee048b67b69c4d38c620"]},{"id":"record-e18a69bed3daa3f2f345","kind":"deepseek-idea-stalled","label":"r2 never shipped","original":{"detail":"r2 never shipped — reported stalled in june 2025 (liang unsatisfied), then reportedly failed on huawei ascend hardware in august. the r-line that produced the shock never produced a successor; the v-line absorbed the flagship role.","id":"clm-stalled-r2","sourceIds":["src-hp-reuters-r2june","src-hp-reuters-r2ft","src-hp-layer3-r2"],"title":"r2 never shipped"},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-e44c73d6384b196b35b1","source-3e57c32cccac9391cd8d","source-754d23aad0ebe499cfb3"]},{"id":"record-94553e7a3f285c052e0d","kind":"deepseek-idea-stalled","label":"the ascend wall","original":{"detail":"the domestic-chip training path stalled — the ascend run failed even with huawei engineers onsite; the lab settled into nvidia-for-training, ascend-for-inference, and months of v4 delay for ascend optimization. the hedge exists; the substitution doesn't.","id":"clm-stalled-ascend","sourceIds":["src-hp-reuters-r2ft","src-hp-register-huawei","src-ci-tomshardware-v32-cann","src-hp-bloomberg-v4delay"],"title":"the ascend wall"},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-3e57c32cccac9391cd8d","source-94edfd719a41e2504487","source-e1a3129a6a4e1d643aa2","source-632e2ce0695ef973e337"]},{"id":"record-87b2e56a4b801c24690d","kind":"deepseek-idea-stalled","label":"research met revenue","original":{"detail":"'research, not products' met the market — the app that beat chatgpt sits #2 in china behind doubao on questmobile's august 2025 mau, and the august 2026 api price hikes (up to ~12x on cache hits) read like incumbent margin management.","id":"clm-stalled-products","sourceIds":["src-ci-scmp-doubao-users","src-ci-sedaily-price-hike","src-ci-tnw-price-rise","src-hp-api-changelog"],"title":"research met revenue"},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-26c1b3789feb89c5a197","source-7df40d3de7d54d968e41","source-5fdab4f20d9aed78ae18","source-08e7ad42ba6d460b8b9c"]},{"id":"record-7074c1263fb35d3101c6","kind":"deepseek-idea-stalled","label":"the end of pure patronage","original":{"detail":"the self-funding purity ended — the lab that 'never had a financing problem' opened its first external round in may 2026, liang injecting up to rmb 20b personally, state capital reported the likely lead. the whale still funds; it no longer funds alone.","id":"clm-stalled-purity","sourceIds":["src-hp-coinlive-funding","src-hp-gate-funding","src-hp-chaincatcher-funding","src-hp-163-cfo"],"title":"the end of pure patronage"},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-db362ca467309d55b804","source-ad6edf1a24679cc18fc3","source-77d1012fefcd2e81ad02","source-c8ea6a03a1163037ad9e"]},{"id":"record-9e4724b92edcdb5f925d","kind":"deepseek-idea-stalled","label":"the bench got poached","original":{"detail":"the flat org got poached — no titles and bottom-up staffing built the bench, then bytedance seed, tencent hunyuan, xiaomi, and deeproute hired it: ~10 of the v4 report's ~300 authors are now former employees, including grpo co-creator guo daya.","id":"clm-stalled-org","sourceIds":["src-hp-newsglobe-departures","src-hp-sinocism-departures","src-hp-gogoai-departures","src-hp-guoday"],"title":"the bench got poached"},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-78d59c1fd11aca3e19b3","source-3340252f2b93a8d7f1a0","source-5d5a3a02db337ef887e9","source-da3a681e38b99fe423fe"]},{"id":"record-2fec3793431cd8565cae","kind":"deepseek-idea-successors","label":"moonshot / kimi","original":{"href":"https://www.moonshot.ai","id":"suc-kimi","name":"moonshot / kimi","note":"the 'second deepseek moment' — kimi k2 thinking beating gpt-5 on hle and browsecomp made the open-weights frontier a chinese category, not a deepseek monopoly.","sourceIds":["src-ci-xpert-kimi","src-ci-linkedin-kimi"]},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-7929874beb28e643e609","source-3e6d37b985ba4fe2a449"]},{"id":"record-8d51f87c770824679d12","kind":"deepseek-idea-successors","label":"alibaba / qwen","original":{"href":"https://qwen.ai","id":"suc-qwen","name":"alibaba / qwen","note":"the platform-scale variant — qwen's open-weights line pairs the lab strategy with a cloud business, and its ecosystem terms reportedly killed its deepseek investment talks.","sourceIds":["src-ci-turingpost-chinesemodels","src-hp-gate-funding"]},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-b151d813e9817329df15","source-ad6edf1a24679cc18fc3"]},{"id":"record-8befe82e83725797e066","kind":"deepseek-idea-successors","label":"zhipu / glm","original":{"href":"https://www.z.ai","id":"suc-zhipu","name":"zhipu / glm","note":"the academic-lab survivor — glm-4 keeps a third chinese open-weights lineage in the frontier set the landscape coverage tracks.","sourceIds":["src-ci-turingpost-chinesemodels","src-ci-36kr-chinese-players"]},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-b151d813e9817329df15","source-e15dd5bd656a3956d2d6"]},{"id":"record-2eecd1b8b420a6f601fd","kind":"deepseek-idea-successors","label":"minimax","original":{"href":"https://www.minimaxi.com","id":"suc-minimax","name":"minimax","note":"the fellow accused — anthropic's february 2026 disclosure names minimax beside deepseek in the distillation campaigns; the cohort shares both the strategy and the scrutiny.","sourceIds":["src-hp-anthropic","src-hp-cnbc-anthropic"]},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-0211196e9ddb845746f9","source-cbd96a7ad11433393eca"]},{"id":"record-3410d729f4e310c26cdb","kind":"deepseek-idea-successors","label":"mistral","original":{"href":"https://mistral.ai","id":"suc-mistral","name":"mistral","note":"the european open-weights control case — same license-first strategy without the whale behind it; venture-funded where deepseek was fund-funded.","sourceIds":["src-ci-turingpost-chinesemodels"]},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-b151d813e9817329df15"]},{"id":"record-3e10ae817b5d5d0ffad2","kind":"deepseek-idea-successors","label":"meta / llama","original":{"href":"https://ai.meta.com","id":"suc-llama","name":"meta / llama","note":"the patron who retreated — llama defined the open-weights position deepseek radicalized; the war-room reporting and capex pledge trace the incumbent's answer.","sourceIds":["src-ci-constellation-meta","src-ci-verge-zuckerberg","src-ci-techcrunch-zuck"]},"originalUrl":"https://hraness.com/deepseek/the-idea","sourceIds":["source-987d33d02fef5b47a9de","source-cddb62e713f17fab2d89","source-15af28a373cbfdf1bdab"]}],"sources":[{"author":null,"id":"source-e15dd5bd656a3956d2d6","note":"top-16 chinese open-source list; arena standings","originalId":"src-ci-36kr-chinese-players","published":{"precision":"month","text":"2025-07"},"publisher":"eu.36kr.com","title":"top-16 chinese open-source list","type":"reporting","url":"https://eu.36kr.com/en/p/3404902313217417"},{"author":null,"id":"source-848b616a2251fdfff2ea","note":"r1 in azure ai foundry + github catalog","originalId":"src-ci-azure-r1","published":{"precision":"day","text":"2025-01-29"},"publisher":"azure.microsoft.com","title":"r1 in azure ai foundry + github catalog","type":"reporting","url":"https://azure.microsoft.com/en-us/blog/deepseek-r1-is-now-available-on-azure-ai-foundry-and-github/"},{"author":null,"id":"source-b01af09f8668309df25a","note":"liang's preface to zuckerman's simons book (2021)","originalId":"src-ci-businesstimes-hedgefund","published":{"precision":"year","text":"2025"},"publisher":"businesstimes.com.sg","title":"liangs preface to zuckermans simons book (2021)","type":"reference","url":"https://www.businesstimes.com.sg/wealth/deepseek-just-latest-hedge-fund-innovation"},{"author":null,"id":"source-987d33d02fef5b47a9de","note":"\"american standard\" open-source quote","originalId":"src-ci-constellation-meta","published":{"precision":"day","text":"2025-01-29"},"publisher":"constellationr.com","title":"american standard open-source quote","type":"reference","url":"https://www.constellationr.com/insights/news/meta-deepseek-custom-silicon-ai-optimizing-engineering-and-business"},{"author":null,"id":"source-9f31fbb840e790bc8564","note":"energy-efficiency paradox","originalId":"src-ci-conversation-jevons","published":{"precision":"month","text":"2025-02"},"publisher":"theconversation.com","title":"energy-efficiency paradox","type":"reference","url":"https://theconversation.com/deepseek-claims-to-have-cured-ais-environmental-headache-the-jevons-paradox-suggests-it-might-make-things-worse-248720"},{"author":null,"id":"source-b73bdd79c86c1e2c64de","note":"apples-to-oranges cost accounting; altman/amodei figure scoping","originalId":"src-ci-creativestrategies-myths","published":{"precision":"month","text":"2025-02"},"publisher":"creativestrategies.com","title":"apples-to-oranges cost accounting","type":"reference","url":"https://creativestrategies.com/dispelling-deepseek-myths-studying-v3/"},{"author":null,"id":"source-596ae44386face52f345","note":"nadella's jevons post verbatim","originalId":"src-ci-fortune-jevons","published":{"precision":"day","text":"2025-01-27"},"publisher":"fortune","title":"nadellas jevons post verbatim","type":"reporting","url":"https://fortune.com/2025/01/27/microsoft-ceo-satya-nadella-deepseek-optimism-jevons-paradox/"},{"author":null,"id":"source-03e27b5b0da205b6a127","note":"nadella context; ives pushback","originalId":"src-ci-geekwire-jevons","published":{"precision":"day","text":"2025-01-27"},"publisher":"geekwire.com","title":"nadella context","type":"reporting","url":"https://www.geekwire.com/2025/microsoft-ceo-says-ai-use-will-skyrocket-with-more-efficiency-amid-craze-over-deepseek/"},{"author":null,"id":"source-03181c9972b7680c28a9","note":"english mirror of the disclosure","originalId":"src-ci-gigazine-545","published":{"precision":"day","text":"2025-03-03"},"publisher":"gigazine.net","title":"english mirror of the disclosure","type":"reporting","url":"https://gigazine.net/gsc_news/en/20250303-deepseek-ai-cost-profit/"},{"author":null,"id":"source-4723c56b6f02585ff949","note":"\"deepseek unmasked\"; 4 findings incl. china mobile backend","originalId":"src-ci-house-select-report","published":{"precision":"day","text":"2025-04-16"},"publisher":"chinaselectcommittee.house.gov","title":"deepseek unmasked","type":"reporting","url":"https://chinaselectcommittee.house.gov/sites/evo-subsites/selectcommitteeontheccp.house.gov/files/evo-media-document/DeepSeek%20Final.pdf"},{"author":null,"id":"source-0f036aee238510dc3dd5","note":"\"optimization not revolution\" counter-take; 143 employees","originalId":"src-ci-ifri-deepseek-moment","published":{"precision":"year","text":"2025"},"publisher":"ifri.org","title":"optimization not revolution counter-take","type":"reference","url":"https://www.ifri.org/en/memos/deepseek-moment"},{"author":null,"id":"source-3e6d37b985ba4fe2a449","note":"k2 thinking beats gpt-5 on hle/browsecomp; huang \"nanoseconds behind\"","originalId":"src-ci-linkedin-kimi","published":{"precision":"day","text":"2025-11-07"},"publisher":"linkedin","title":"k2 thinking beats gpt-5 on hle/browsecomp","type":"reporting","url":"https://www.linkedin.com/posts/conorgrennan_breaking-kimi-k2-may-be-chinas-second-deepseek-activity-7392592898703278080-E8tl"},{"author":null,"id":"source-a9178d68fc3e339ab065","note":"srinivas: us/eu-hosted, uncensored","originalId":"src-ci-mspoweruser-perplexity","published":{"precision":"day","text":"2025-01-27"},"publisher":"mspoweruser.com","title":"srinivas: us/eu-hosted, uncensored","type":"reporting","url":"https://mspoweruser.com/you-can-now-try-uncensored-deepseek-r1-via-perplexity-ps-its-us-hosted/"},{"author":null,"id":"source-b98e2e12f7e159ef11b0","note":"545% theoretical; 226.75 nodes×8×h800; $87k→$562k","originalId":"src-ci-scmp-545-margin","published":{"precision":"day","text":"2025-03-01"},"publisher":"scmp.com","title":"545% theoretical","type":"reporting","url":"https://www.scmp.com/tech/big-tech/article/3300734/deepseek-shows-power-v3-r1-models-theoretical-545-profit-margin"},{"author":null,"id":"source-26c1b3789feb89c5a197","note":"questmobile aug 2025: doubao 157m vs deepseek 143m mau","originalId":"src-ci-scmp-doubao-users","published":{"precision":"month","text":"2025-09"},"publisher":"scmp.com","title":"questmobile aug 2025: doubao 157m vs deepseek 143m mau","type":"reporting","url":"https://www.scmp.com/tech/big-tech/article/3325858/bytedance-chatbot-doubao-still-chinas-most-popular-ai-app-rival-deepseek-loses-users"},{"author":null,"id":"source-7df40d3de7d54d968e41","note":"up-to-12x cache-hit hike; peak/off-peak split; ipo context","originalId":"src-ci-sedaily-price-hike","published":{"precision":"day","text":"2026-08-14"},"publisher":"en.sedaily.com","title":"up-to-12x cache-hit hike","type":"reporting","url":"https://en.sedaily.com/international/2026/08/14/deepseek-raises-api-prices-up-to-12-fold-as-chinas-ai-firms"},{"author":null,"id":"source-acdf9ae0c597d9d7c8fe","note":"teardown: $5.576m real-but-narrow; doubao overtook; r2 slipped; v4 trails 3–6 mo","originalId":"src-ci-teardown","published":{"precision":"year","text":"2026"},"publisher":"theteardown.co","title":"the teardown: $5.576m real but narrow","type":"reference","url":"https://www.theteardown.co/deepseek"},{"author":null,"id":"source-15af28a373cbfdf1bdab","note":"\"hundreds of billions\" capex pledge","originalId":"src-ci-techcrunch-zuck","published":{"precision":"day","text":"2025-01-29"},"publisher":"techcrunch","title":"hundreds of billions capex pledge","type":"reporting","url":"https://techcrunch.com/2025/01/29/zuck-shrugs-off-deepseek-vows-to-spend-hundreds-of-billions-on-ai/"},{"author":null,"id":"source-19357a942001743d6668","note":"doubao ¥0.0008/1k tokens; 99.3% cut; sector-wide price war","originalId":"src-ci-techflow-doubao","published":{"precision":"month","text":"2024-05"},"publisher":"techflowpost.com","title":"doubao at ¥0.0008/1k tokens; 99.3% cut","type":"reporting","url":"https://techflowpost.com/en-US/article/17993"},{"author":null,"id":"source-5fdab4f20d9aed78ae18","note":"the \"significant\" price-rise warning","originalId":"src-ci-tnw-price-rise","published":{"precision":"year","text":"2026"},"publisher":"the next web","title":"the significant price-rise warning","type":"reporting","url":"https://thenextweb.com/news/deepseek-significant-api-price-increase"},{"author":null,"id":"source-e1a3129a6a4e1d643aa2","note":"v3.2-exp day-1 ascend/cann/vllm support","originalId":"src-ci-tomshardware-v32-cann","published":{"precision":"day","text":"2025-09-29"},"publisher":"tomshardware.com","title":"v3.2-exp ships day-1 ascend/cann support","type":"reporting","url":"https://www.tomshardware.com/tech-industry/deepseek-new-model-supports-huawei-cann"},{"author":null,"id":"source-b151d813e9817329df15","note":"kimi k2/r1/qwen3/glm-4.5 comparative landscape","originalId":"src-ci-turingpost-chinesemodels","published":{"precision":"year","text":"2025"},"publisher":"turingpost.com","title":"the chinese open-weights landscape","type":"reference","url":"https://www.turingpost.com/p/chinesemodels"},{"author":null,"id":"source-1a377fcc410e5e97f373","note":"openai evidence claim; msft api exfiltration; sacks quote","originalId":"src-ci-verge-distillation","published":{"precision":"day","text":"2025-01-29"},"publisher":"the verge","title":"openai evidence claim","type":"reporting","url":"https://www.theverge.com/news/601195/openai-evidence-deepseek-distillation-ai-data"},{"author":null,"id":"source-cddb62e713f17fab2d89","note":"\"strengthened our conviction\"; inference compute","originalId":"src-ci-verge-zuckerberg","published":{"precision":"day","text":"2025-01-29"},"publisher":"the verge","title":"strengthened our conviction","type":"reporting","url":"https://www.theverge.com/news/602233/meta-deepseek-zuckerberg-earings-q4"},{"author":null,"id":"source-7929874beb28e643e609","note":"\"second deepseek moment\" — kimi k2","originalId":"src-ci-xpert-kimi","published":{"precision":"day","text":"2025-07-21"},"publisher":"xpert.digital","title":"second deepseek moment — kimi k2","type":"reference","url":"https://xpert.digital/en/sputnik-the-moment-of-artificial-intelligence/"},{"author":null,"id":"source-c8ea6a03a1163037ad9e","note":"cfo yan wentao (b.1991, hillhouse); 2009 internship anecdote","originalId":"src-hp-163-cfo","published":{"precision":"month","text":"2025-09"},"publisher":"m.163.com","title":"the first cfo: yan wentao","type":"reporting","url":"https://m.163.com/dy/article/L6PV3CAP05568W0A.html"},{"author":null,"id":"source-07eaa7564c5f63b78585","note":"first-ever liang interview; naming of 深度求索; gpu ladder quote; funding logic","originalId":"src-hp-36kr-interview1","published":{"precision":"day","text":"2023-05-24"},"publisher":"m.36kr.com","title":"first-ever liang interview","type":"interview","url":"https://m.36kr.com/p/2272896094586500"},{"author":null,"id":"source-e2c7feee45a597771920","note":"the canonical \"techno-idealism\" interview (于丽丽/暗涌)","originalId":"src-hp-36kr-interview2","published":{"precision":"day","text":"2024-07-17"},"publisher":"36kr.com","title":"the canonical techno-idealism interview (于丽丽/暗涌)","type":"interview","url":"https://www.36kr.com/p/2872793466982535"},{"author":null,"id":"source-807def74933e03156976","note":"csis's allen house testimony: $5.576m math; oct-7-2022 controls; h800 workaround history","originalId":"src-hp-allen-testimony","published":{"precision":"year","text":"2025"},"publisher":"republicans-science.house.gov","title":"allen's house testimony: the $5.576m math","type":"reference","url":"https://republicans-science.house.gov/_cache/files/0/d/0d8e8a6c-2a09-413d-9d78-e21df078bc9e/EAB890C6D96E33647BC244A4EFCC742660C0EE09981A7B46E6BC1EC8C80F5386.gregory-c.-allen-testimony.pdf"},{"author":null,"id":"source-0211196e9ddb845746f9","note":"anthropic's own account: 24k accts/16m exchanges; ds 150k+","originalId":"src-hp-anthropic","published":{"precision":"day","text":"2026-02-23"},"publisher":"anthropic","title":"anthropics own account: 24k accts/16m exchanges","type":"primary","url":"https://www.anthropic.com/news/detecting-and-preventing-distillation-attacks"},{"author":null,"id":"source-08e7ad42ba6d460b8b9c","note":"v4-pro ga 8/13/26; vision-exp 8/21/26; v4.1-flash 9/10/26","originalId":"src-hp-api-changelog","published":{"precision":"month","text":"2026-09"},"publisher":"api-docs.deepseek.com","title":"the api changelog: v4-pro ga 2026-08-13","type":"primary","url":"https://api-docs.deepseek.com/updates"},{"author":null,"id":"source-562cf7588945d2a05816","note":"state dept official allegation; wang 50k-h100 claim; nvidia \"lawfully acquired h800\"; v3=2048 h800s","originalId":"src-hp-asiatimes","published":{"precision":"day","text":"2025-06-25"},"publisher":"asiatimes.com","title":"state dept official allegation","type":"reporting","url":"https://asiatimes.com/2025/06/deepseek-gets-nvidias-high-end-gpus-via-singapore-us-official/"},{"author":null,"id":"source-e85ca01f7c1f488b5018","note":"jacobi 2013; name origins; 80k-capital story; dji's wang tao; aum ladder; 4 partners; donations","originalId":"src-hp-bianews-history","published":{"precision":"year","text":"2025"},"publisher":"bianews.com","title":"the jacobi-to-high-flyer history","type":"reporting","url":"https://www.bianews.com/news/details?id=205377"},{"author":null,"id":"source-632e2ce0695ef973e337","note":"v4 delayed for ascend-tuned stack (yuyuantantian)","originalId":"src-hp-bloomberg-v4delay","published":{"precision":"day","text":"2026-04-26"},"publisher":"bloomberg","title":"v4 delayed for ascend-tuned stack (yuyuantantian)","type":"reporting","url":"https://www.bloomberg.com/news/articles/2026-04-26/deepseek-v4-delay-shows-shift-to-china-chips-cctv-account-says"},{"author":null,"id":"source-79c1aa742d9e9d5a21d6","note":"nvidia −17%/−$600b \"single-biggest one-day loss\"; ives $8b liang assets","originalId":"src-hp-cbs-nvidia","published":{"precision":"day","text":"2025-01-27"},"publisher":"wayback machine","title":"nvidia −17%/−$600b single-biggest one-day loss","type":"archive","url":"https://web.archive.org/web/20250129002522/https:/www.cbsnews.com/news/what-is-deepseek-ai-china-stock-nvidia-nvda-asml/"},{"author":null,"id":"source-77d1012fefcd2e81ad02","note":"same round; liang's \"fewest conditions\" stance","originalId":"src-hp-chaincatcher-funding","published":{"precision":"day","text":"2026-05-09"},"publisher":"chaincatcher.com","title":"the may 2026 round; the fewest-conditions stance","type":"reporting","url":"https://www.chaincatcher.com/en/article/2263550"},{"author":null,"id":"source-cbd96a7ad11433393eca","note":"minimax 13m / moonshot 3.4m / ds 150k; openai letter earlier in feb","originalId":"src-hp-cnbc-anthropic","published":{"precision":"day","text":"2026-02-24"},"publisher":"cnbc","title":"the anthropic distillation figures","type":"reporting","url":"https://www.cnbc.com/2026/02/24/anthropic-openai-china-firms-distillation-deepseek.html"},{"author":null,"id":"source-db362ca467309d55b804","note":"may-8 round launch; liang's rmb 20b; ff1/ff2 cost figures; ~84% control","originalId":"src-hp-coinlive-funding","published":{"precision":"month","text":"2026-05"},"publisher":"coinlive.com","title":"may-8 round launch","type":"reporting","url":"https://www.coinlive.com/news/deepseek-s-liang-wenfeng-we-don-t-want-alibaba-s-money-we-ll-contribute"},{"author":null,"id":"source-20ac7e2113fd2e7046b5","note":"law-firm equity-structure analysis; 17 jul 2023, rmb 10m, liang >80% via chengpu/cheng'en","originalId":"src-hp-dehehantong-equity","published":{"precision":"year","text":"2025"},"publisher":"dehehantong.com","title":"law-firm equity-structure analysis","type":"reporting","url":"https://dehehantong.com/en/news/articles/156"},{"author":null,"id":"source-f6ee4abd62d2f8da7bcc","note":"~150 staff; gold-medal hiring rep; no kpis/titles; >bytedance pay","originalId":"src-hp-donga","published":{"precision":"day","text":"2025-02-11"},"publisher":"donga.com","title":"~150 staff; the gold-medal hiring rep","type":"reporting","url":"https://www.donga.com/news/article/all/20250211/131006164/1"},{"author":null,"id":"source-ee048b67b69c4d38c620","note":"fire-flyer ai-hpc paper; 10k pcie a100; authors \"deepseek ai\" incl. wenfeng liang","originalId":"src-hp-fireflyer-paper","published":{"precision":"month","text":"2024-08"},"publisher":"arxiv.org","title":"fire-flyer ai-hpc paper","type":"primary","url":"https://arxiv.org/abs/2408.14158"},{"author":null,"id":"source-ad6edf1a24679cc18fc3","note":"alibaba collapse; ~$45b; tencent 20% declined; ic-fund lead possibility; hf 56.55% on 70b","originalId":"src-hp-gate-funding","published":{"precision":"day","text":"2026-05-09"},"publisher":"gate.com","title":"alibaba collapse","type":"reporting","url":"https://www.gate.com/news/detail/whale-lab-deepseek-and-alibabas-fundraising-negotiations-failed-to-reach-an-20949731"},{"author":null,"id":"source-5d5a3a02db337ef887e9","note":"corroboration of 10 departures","originalId":"src-hp-gogoai-departures","published":{"precision":"year","text":"2026"},"publisher":"gogoai.xin","title":"corroboration of 10 departures","type":"reporting","url":"https://www.gogoai.xin/article/deepseek-loses-10-key-researchers-as-chinas-ai-talent-war-intensifies"},{"author":null,"id":"source-da3a681e38b99fe423fe","note":"guo's own page: \"ai researcher in deepseek, july 2023–april 2026\" + core-contributor lists","originalId":"src-hp-guoday","published":{"precision":"month","text":"2026-09"},"publisher":"guoday.github.io","title":"guos own page: ai researcher in deepseek, july 2023–april 2026 + core-contributor lists","type":"primary","url":"https://guoday.github.io/"},{"author":null,"id":"source-e24e7de33a56f8ac2121","note":"gtc-2022 blog: 10k-a100 cluster, ≥90% util, 3fs/hfreduce/hfai.nn, 50+ labs, 2.7m gpu-hrs","originalId":"src-hp-hfai-blog","published":{"precision":"year","text":"2022"},"publisher":"high-flyer.cn","title":"the hfai blog: the 10k-a100 cluster","type":"primary","url":"https://www.high-flyer.cn/blog/hfai/"},{"author":null,"id":"source-5eb143aeec914b6d4c1e","note":"official timeline 2008–2022; ff1/ff2 specs+costs; donations; hq address 汇金国际大厦a座14层","originalId":"src-hp-highflyer-history","published":{"precision":"month","text":"2026-09"},"publisher":"high-flyer.cn","title":"official timeline 2008–2022","type":"primary","url":"https://www.high-flyer.cn/history/"},{"author":null,"id":"source-ff1ee7c6d614cc863b4d","note":"fire-flyer ii stats (8tb/s read, feb-2022 stats)","originalId":"src-hp-highflyer-home","published":{"precision":"month","text":"2026-09"},"publisher":"high-flyer.cn","title":"fire-flyer ii stats (8tb/s read, feb-2022 stats)","type":"primary","url":"https://www.high-flyer.cn/en/"},{"author":null,"id":"source-e642177ea7ed77782280","note":"thesis title, advisor xiang zhiyu, chengdu years; gives 22-may-1985 (single-source)","originalId":"src-hp-ie-liang","published":{"precision":"year","text":"2025"},"publisher":"interestingengineering.com","title":"thesis title, advisor xiang zhiyu, chengdu years","type":"reference","url":"https://interestingengineering.com/engineers-directory/liang-wenfeng"},{"author":null,"id":"source-e41d366fa1f22b558c6c","note":"annotated mirror; calls it his only deep text interview","originalId":"src-hp-laojinchuhai","published":{"precision":"day","text":"2024-07-17"},"publisher":"laojinchuhai.com","title":"annotated mirror","type":"archive","url":"https://www.laojinchuhai.com/zh/classics/liang-wenfeng-anchorage-2024"},{"author":null,"id":"source-754d23aad0ebe499cfb3","note":"r2 status: unannounced, unreleased as of jul-2026; delay reporting recap","originalId":"src-hp-layer3-r2","published":{"precision":"month","text":"2026-07"},"publisher":"layer3labs.io","title":"r2 status: unannounced, unreleased as of jul-2026","type":"reporting","url":"https://www.layer3labs.io/guides/deepseek-r2-explained"},{"author":null,"id":"source-7f0f229eb58f8d7e84fc","note":"r1 peer-reviewed; nature 645:633–638; received feb-14, accepted jul-17","originalId":"src-hp-nature-r1","published":{"precision":"day","text":"2025-09-17"},"publisher":"nature.com","title":"r1 peer-reviewed","type":"primary","url":"https://www.nature.com/articles/s41586-025-09422-z"},{"author":null,"id":"source-78d59c1fd11aca3e19b3","note":"v4 report ~300 authors; 10 flagged departed; destinations","originalId":"src-hp-newsglobe-departures","published":{"precision":"year","text":"2026"},"publisher":"newsglobenow.com","title":"v4 report ~300 authors","type":"reporting","url":"https://www.newsglobenow.com/new346166.html"},{"author":null,"id":"source-b43c57bf1bf2c459d887","note":"v3 671b/37b, 14.8t tokens; h800 analysis","originalId":"src-hp-nextplatform","published":{"precision":"day","text":"2025-01-27"},"publisher":"nextplatform.com","title":"how deepseek trained on crippled hardware","type":"reporting","url":"https://www.nextplatform.com/ai/2025/01/27/how-did-deepseek-train-its-ai-model-on-a-lot-less-and-crippled-hardware/1656116"},{"author":null,"id":"source-f3a7aa842ef643e17128","note":"open source week index; daily drops list","originalId":"src-hp-openinfra","published":{"precision":"day","text":"2025-02-24"},"publisher":"github","title":"open source week index","type":"repo","url":"https://github.com/deepseek-ai/open-infra-index"},{"author":null,"id":"source-94edfd719a41e2504487","note":"huawei onsite team; ascend→inference-only; no fp8","originalId":"src-hp-register-huawei","published":{"precision":"day","text":"2025-08-14"},"publisher":"theregister.com","title":"huawei onsite team","type":"reporting","url":"https://www.theregister.com/software/2025/08/14/dodgy_huawei_chips_nearly_sunk_deepseeks_next_gen_r2_mode/753616"},{"author":null,"id":"source-5bb15e05d2d051f06e4f","note":"one of nine speakers; >rmb 100b end-2021 aum","originalId":"src-hp-reuters-liang","published":{"precision":"day","text":"2025-01-28"},"publisher":"reuters","title":"one of nine speakers","type":"reporting","url":"https://www.reuters.com/technology/deepseek-founder-liang-wenfeng-puts-focus-chinese-innovation-2025-01-28/"},{"author":null,"id":"source-3e57c32cccac9391cd8d","note":"ft: ascend training failure delayed r2","originalId":"src-hp-reuters-r2ft","published":{"precision":"day","text":"2025-08-14"},"publisher":"reuters","title":"ft: ascend training failure delayed r2","type":"reporting","url":"https://www.reuters.com/world/china/deepseeks-launch-new-ai-model-delayed-by-huawei-chip-issues-ft-reports-2025-08-14/"},{"author":null,"id":"source-e44c73d6384b196b35b1","note":"r2 stalled; liang unsatisfied; planned may","originalId":"src-hp-reuters-r2june","published":{"precision":"day","text":"2025-06-26"},"publisher":"reuters","title":"deepseek-r2 launch stalled; ceo balks","type":"reporting","url":"https://www.reuters.com/world/china/deepseek-r2-launch-stalled-ceo-balks-progress-information-reports-2025-06-26/"},{"author":null,"id":"source-c83479b1406635460658","note":"jan-20 li qiang symposium","originalId":"src-hp-scmp-liqiang","published":{"precision":"day","text":"2025-01-21"},"publisher":"scmp.com","title":"jan-20 li qiang symposium","type":"reporting","url":"https://www.scmp.com/tech/policy/article/3295662/beijing-meeting-puts-spotlight-chinas-new-face-ai-deepseek-founder-liang-wenfeng"},{"author":null,"id":"source-3340252f2b93a8d7f1a0","note":"departures list incl. guo daya (apr 2026), grpo, 37k citations","originalId":"src-hp-sinocism-departures","published":{"precision":"year","text":"2026"},"publisher":"substack","title":"departures list incl","type":"community","url":"https://substack.com/@sinocism/note/c-245366529"},{"author":null,"id":"source-aa6aee6a5efbf005dd77","note":"feb-17 xi symposium confirmed via yuyuantantian","originalId":"src-hp-standard-xi","published":{"precision":"day","text":"2025-02-17"},"publisher":"thestandard.com.hk","title":"feb-17 xi symposium confirmed via yuyuantantian","type":"reporting","url":"https://www.thestandard.com.hk/finance/article/227878/DeepSeek-founder-confirmed-in-the-meeting-with-Xi"},{"author":null,"id":"source-777866aa5f7202981ebf","note":"2008 stock-dabbling trio; no overseas experience; systematic turn 2015","originalId":"src-hp-techcentral-liang","published":{"precision":"month","text":"2025-02"},"publisher":"techcentral.co.za","title":"2008 stock-dabbling trio","type":"reporting","url":"https://techcentral.co.za/liang-wenfeng-built-deepseek-hedge-fund/258390/"},{"author":null,"id":"source-acebedb59dd691984f43","note":"#1 us app store jan 26; 51 countries (appfigures)","originalId":"src-hp-techcrunch-appstore","published":{"precision":"day","text":"2025-01-27"},"publisher":"techcrunch","title":"#1 us app store jan 26","type":"reporting","url":"https://techcrunch.com/2025/01/27/deepseek-displaces-chatgpt-as-the-app-stores-top-app/"},{"author":null,"id":"source-537d6289b1b6c1dbe6d9","note":"travel curbs + zhejiang investor screening","originalId":"src-hp-techcrunch-watch","published":{"precision":"day","text":"2025-03-14"},"publisher":"techcrunch","title":"travel curbs + zhejiang investor screening","type":"reporting","url":"https://techcrunch.com/2025/03/14/china-is-reportedly-keeping-deepseek-under-close-watch/"},{"author":null,"id":"source-c9bfc1101f4c8e3a80cf","note":"full-text mirror used for verbatim quotes","originalId":"src-hp-wallstreetcn-interview2","published":{"precision":"day","text":"2024-07-18"},"publisher":"wallstreetcn.com","title":"full-text mirror used for verbatim quotes","type":"interview","url":"https://wallstreetcn.com/articles/3719982"},{"author":null,"id":"source-9c18637a18012ad21bfe","note":"\"letting research reports speak\"; api pricing change note","originalId":"src-hp-xinhua-office","published":{"precision":"day","text":"2025-02-25"},"publisher":"tthmx.com","title":"letting research reports speak","type":"reporting","url":"https://tthmx.com/hangye/90274.html"},{"author":null,"id":"source-e06948c401a48f463465","note":"zju page: xu jin bio — chu kochen mixed class, signal-processing phd, yutu rover nav","originalId":"src-hp-zju-xujin","published":{"precision":"day","text":"2017-06-08"},"publisher":"zju.edu.cn","title":"zju page: xu jin bio — chu kochen mixed class, signal-processing phd, yutu rover nav","type":"primary","url":"https://www.zju.edu.cn/2017/0608/c33485a514822/pagem.htm"}]}
````