Notes on DeepSeek

44 comments · 2,378 words

Complete Created: Jun 10, 09:19 AM (00:03:57)

Models: Claude Opus 4.6 (analyze) · Gemini 3 Flash (tag) · Gemini 3 Flash (summarize)

Article URL: https://twitter.com/NikoMcCarty/status/2064686557400100884

Article could not be fetched: Failed to fetch https://twitter.com/NikoMcCarty/status/2064686557400100884: 404 Not Found

Comment Summary

Topics

• Distillation and Competence (9 comments) (Whether reliance on distillation indicates lack of innovation, argument that distillers can never reach frontier, counterargument that DeepSeek produces innovative papers and excellent models like MiMo, parallel to 'China only copies' narrative)
• Model Distillation Ethics (7 comments) (Debate over whether distillation constitutes theft or legitimate technique, comparison to how GPT trained on internet data, noting API fees were paid, irony of Anthropic criticizing distillation while pirating books for training data)
• China vs US AI Philosophy (6 comments) (China treats AI as just another technology rather than a singularity moment, focused on basic needs and infrastructure, contrasted with US religious/cult-like AI hype from Altman and Amodei, dreams of singularity seen as luxury)
• US AI Hype and Sales Culture (5 comments) (Criticism of US AI companies for religious rhetoric, constant release cycles, gatekeeping, FUD, Altman talking about creating religion, Dario meeting Vatican, claims of building God, overzealous statements about AI capabilities)
• Deleted Post Dynamics (5 comments) (Original tweet was removed, preserved via Pastebin, discussion about whether to wait for OP to repost with edits, community effort to preserve and share the content)
• AI Accessibility and Cost (4 comments) (DeepSeek enabling $10/month AI access versus $100/month western alternatives, concern about income disparity, desire for SoTA AI as universal right, frustration with unlimited work tokens but expensive personal use)
• Skepticism of Visit Report (4 comments) (Some commenters view the notes as a puff piece, suggest the visitor was being managed, others find details like employee count and HQ description genuinely informative, debate over substance of the report)
• Anthropic Criticism (4 comments) (Multiple threads criticizing Anthropic for hypocrisy around distillation complaints while pirating books, settling with authors, gatekeeping access, Mythos claims questioned if distillation attacks work against it)
• DeepSeek Company Culture (3 comments) (300 employees, unmarked building in Hangzhou, no branding, deliberately low profile, young energetic team, anti-996 work culture, family-like atmosphere, content to remain behind US companies, proudest achievement is R1)
• China Copying Narrative (3 comments) (Historical dismissal of Chinese innovation as mere copying, now China leads in EVs batteries solar drones, argument that starting from others' work and improving is legitimate, comparison to generic products in US markets)
• Chinese AI Censorship (3 comments) (DeepSeek aligns with party line on political topics, government regulates AI usage in software and services rather than models directly, debate over whether this distinction is meaningful, Tiananmen-denialism in models)
• AI Regulation Concerns (3 comments) (Question about when illegal models will emerge, China's approach of regulating AI applications rather than models, open weights models and ability to strip censorship, centralization concerns with US companies)
• Wikipedia Citation Parallel (3 comments) (Tangent comparing AI in education to Wikipedia's acceptance journey, from being banned as source to being more accepted, kids initially pushed to use AI for homework now disallowed, mixed messaging)
• China Market Pricing (3 comments) (Observation that in every market China dominates products remain inexpensive including solar panels batteries EVs drones, counterargument that Chinese Big Tech would be as bad as US Big Tech if they dominate)
• Open Weights and Power Dynamics (2 comments) (Argument that open weights models from China counter centralization fears, ability to download and run models locally, strip out censorship, US-based Anthropic and OpenAI centralization seen as more concerning)
• Chinese AI Competition Landscape (1 comment) (Alibaba Qwen, ByteDance, Moonshot Kimi competing with DeepSeek, poaching between groups common, big labs mostly in Beijing near Tsinghua and Peking University, Hangzhou as exception)
• Red Teaming and Safety (1 comment) (DeepSeek reportedly does not do red teaming on models, contrasted with western AI safety focus, China's pragmatic approach to AI safety through usage regulation rather than model-level restrictions)
• AI Environmental Impact (1 comment) (Brief mention of environmental concerns relative to AI spending, questioning whether improvements outside software development justify the spend and environmental cost of current AI development)
• Western AI Writers Influence (0 comments) (DeepSeek team reads western AI writers, listens to Dwarkesh podcast, reads Gwern, had never met Anthropic employees, cross-cultural intellectual exchange in AI research community)
• AI Job Loss Concerns (0 comments) (DeepSeek team's main concern is job displacement rather than AGI takeover, already high youth unemployment in China, not concerned with hostile AGI scenarios, practical focus on economic impact)

Raw Files

Execution Log

[2026-06-10T16:19:53.224Z] Starting step: fetch_pages (attempt 1)
[2026-06-10T16:19:53.274Z] Fetching HN page: https://news.ycombinator.com/item?id=48476474
[2026-06-10T16:19:53.479Z] Error in step fetch_pages: Failed to fetch https://news.ycombinator.com/item?id=48476474: 429 Too Many Requests
Error: Failed to fetch https://news.ycombinator.com/item?id=48476474: 429 Too Many Requests
    at fetchWithCurlAgent (index.js:1164:11)
    at async fetchPages (index.js:1179:18)
    at async Object.executeStage (index.js:2338:9)
    at async processQueueMessage (index.js:5180:20)
    at async Object.queue (index.js:5331:9)
[2026-06-10T16:19:53.505Z] Scheduling retry in 10s (attempt 2)
[2026-06-10T16:20:08.778Z] Starting step: fetch_pages (attempt 2)
[2026-06-10T16:20:08.820Z] Fetching HN page: https://news.ycombinator.com/item?id=48476474
[2026-06-10T16:20:08.942Z] Fetched HN page: 70862 bytes
[2026-06-10T16:20:09.140Z] Extracted title: Notes on DeepSeek
[2026-06-10T16:20:09.166Z] Extracted linked URL: https://twitter.com/NikoMcCarty/status/2064686557400100884
[2026-06-10T16:20:09.194Z] Fetching linked article: https://twitter.com/NikoMcCarty/status/2064686557400100884
[2026-06-10T16:20:09.662Z] Warning: Failed to fetch linked article: Failed to fetch https://twitter.com/NikoMcCarty/status/2064686557400100884: 404 Not Found
[2026-06-10T16:20:09.723Z] Completed step: fetch_pages in 918ms
[2026-06-10T16:20:10.153Z] Starting step: extract_text (attempt 1)
[2026-06-10T16:20:10.356Z] Extracted HN text: 16606 chars
[2026-06-10T16:20:10.545Z] Extracted 44 comments
[2026-06-10T16:20:10.617Z] Comment word count: 2378
[2026-06-10T16:20:10.704Z] Completed step: extract_text in 526ms
[2026-06-10T16:20:11.134Z] Starting step: analyze_content (attempt 1)
[2026-06-10T16:20:11.192Z] Calling claude-opus-4-6 (article: 0 chars, 44 comments)
[2026-06-10T16:20:42.230Z] Analysis complete: 20 topics, 3713 input tokens, 1195 output tokens
[2026-06-10T16:20:42.293Z] Completed step: analyze_content in 31131ms
[2026-06-10T16:20:42.623Z] Starting step: tag_comments (attempt 1)
[2026-06-10T16:20:42.681Z] Tagging 44 comments with 20 topics (batch size: 50)
[2026-06-10T16:20:42.711Z] Processing batch 1/1 (44 comments)
[2026-06-10T16:21:01.331Z] Batch 1 complete: 68 tags assigned
[2026-06-10T16:21:01.359Z] Tagging complete: 68 total tags, 8736 input tokens, 1025 output tokens
[2026-06-10T16:21:01.391Z] Completed step: tag_comments in 18736ms
[2026-06-10T16:21:01.769Z] Starting step: summarize_topics (attempt 1)
[2026-06-10T16:21:01.816Z] Summarizing 20 topics
[2026-06-10T16:21:01.933Z] Summarizing topic 1/20: "DeepSeek Company Culture # 300 employees, unmarked building in Hangzhou, no branding, deliberately low profile, young energetic team, anti-996 work culture, family-like atmosphere, content to remain behind US companies, proudest achievement is R1" (3 comments)
[2026-06-10T16:21:08.369Z] Topic 1 summarized (1593 in, 156 out)
[2026-06-10T16:21:08.419Z] Summarizing topic 2/20: "China vs US AI Philosophy # China treats AI as just another technology rather than a singularity moment, focused on basic needs and infrastructure, contrasted with US religious/cult-like AI hype from Altman and Amodei, dreams of singularity seen as luxury" (6 comments)
[2026-06-10T16:21:14.594Z] Topic 2 summarized (1810 in, 139 out)
[2026-06-10T16:21:14.710Z] Summarizing topic 3/20: "US AI Hype and Sales Culture # Criticism of US AI companies for religious rhetoric, constant release cycles, gatekeeping, FUD, Altman talking about creating religion, Dario meeting Vatican, claims of building God, overzealous statements about AI capabilities" (5 comments)
[2026-06-10T16:21:20.517Z] Topic 3 summarized (1284 in, 142 out)
[2026-06-10T16:21:20.569Z] Summarizing topic 4/20: "Model Distillation Ethics # Debate over whether distillation constitutes theft or legitimate technique, comparison to how GPT trained on internet data, noting API fees were paid, irony of Anthropic criticizing distillation while pirating books for training data" (7 comments)
[2026-06-10T16:21:28.772Z] Topic 4 summarized (1373 in, 116 out)
[2026-06-10T16:21:28.832Z] Summarizing topic 5/20: "Distillation and Competence # Whether reliance on distillation indicates lack of innovation, argument that distillers can never reach frontier, counterargument that DeepSeek produces innovative papers and excellent models like MiMo, parallel to 'China only copies' narrative" (9 comments)
[2026-06-10T16:21:34.718Z] Topic 5 summarized (1393 in, 143 out)
[2026-06-10T16:21:34.763Z] Summarizing topic 6/20: "Chinese AI Competition Landscape # Alibaba Qwen, ByteDance, Moonshot Kimi competing with DeepSeek, poaching between groups common, big labs mostly in Beijing near Tsinghua and Peking University, Hangzhou as exception" (1 comments)
[2026-06-10T16:21:41.625Z] Topic 6 summarized (1482 in, 144 out)
[2026-06-10T16:21:41.668Z] Summarizing topic 7/20: "AI Accessibility and Cost # DeepSeek enabling $10/month AI access versus $100/month western alternatives, concern about income disparity, desire for SoTA AI as universal right, frustration with unlimited work tokens but expensive personal use" (4 comments)
[2026-06-10T16:21:49.724Z] Topic 7 summarized (1273 in, 154 out)
[2026-06-10T16:21:49.771Z] Summarizing topic 8/20: "China Copying Narrative # Historical dismissal of Chinese innovation as mere copying, now China leads in EVs batteries solar drones, argument that starting from others' work and improving is legitimate, comparison to generic products in US markets" (3 comments)
[2026-06-10T16:21:54.775Z] Topic 8 summarized (776 in, 131 out)
[2026-06-10T16:21:54.816Z] Summarizing topic 9/20: "Chinese AI Censorship # DeepSeek aligns with party line on political topics, government regulates AI usage in software and services rather than models directly, debate over whether this distinction is meaningful, Tiananmen-denialism in models" (3 comments)
[2026-06-10T16:22:01.790Z] Topic 9 summarized (960 in, 110 out)
[2026-06-10T16:22:01.840Z] Summarizing topic 10/20: "AI Regulation Concerns # Question about when illegal models will emerge, China's approach of regulating AI applications rather than models, open weights models and ability to strip censorship, centralization concerns with US companies" (3 comments)
[2026-06-10T16:22:06.931Z] Topic 10 summarized (851 in, 112 out)
[2026-06-10T16:22:06.971Z] Topic 11 "Western AI Writers Influence # DeepSeek team reads western AI writers, listens to Dwarkesh podcast, reads Gwern, had never met Anthropic employees, cross-cultural intellectual exchange in AI research community": no comments, skipping
[2026-06-10T16:22:07.007Z] Topic 12 "AI Job Loss Concerns # DeepSeek team's main concern is job displacement rather than AGI takeover, already high youth unemployment in China, not concerned with hostile AGI scenarios, practical focus on economic impact": no comments, skipping
[2026-06-10T16:22:07.051Z] Summarizing topic 13/20: "Skepticism of Visit Report # Some commenters view the notes as a puff piece, suggest the visitor was being managed, others find details like employee count and HQ description genuinely informative, debate over substance of the report" (4 comments)
[2026-06-10T16:22:40.092Z] Topic 13 summarized (865 in, 129 out)
[2026-06-10T16:22:40.136Z] Summarizing topic 14/20: "Open Weights and Power Dynamics # Argument that open weights models from China counter centralization fears, ability to download and run models locally, strip out censorship, US-based Anthropic and OpenAI centralization seen as more concerning" (2 comments)
[2026-06-10T16:22:45.244Z] Topic 14 summarized (847 in, 140 out)
[2026-06-10T16:22:45.286Z] Summarizing topic 15/20: "Wikipedia Citation Parallel # Tangent comparing AI in education to Wikipedia's acceptance journey, from being banned as source to being more accepted, kids initially pushed to use AI for homework now disallowed, mixed messaging" (3 comments)
[2026-06-10T16:22:52.278Z] Topic 15 summarized (1007 in, 108 out)
[2026-06-10T16:22:52.322Z] Summarizing topic 16/20: "Red Teaming and Safety # DeepSeek reportedly does not do red teaming on models, contrasted with western AI safety focus, China's pragmatic approach to AI safety through usage regulation rather than model-level restrictions" (1 comments)
[2026-06-10T16:23:18.604Z] Topic 16 summarized (666 in, 90 out)
[2026-06-10T16:23:18.655Z] Summarizing topic 17/20: "AI Environmental Impact # Brief mention of environmental concerns relative to AI spending, questioning whether improvements outside software development justify the spend and environmental cost of current AI development" (1 comments)
[2026-06-10T16:23:24.322Z] Topic 17 summarized (945 in, 97 out)
[2026-06-10T16:23:24.365Z] Summarizing topic 18/20: "Anthropic Criticism # Multiple threads criticizing Anthropic for hypocrisy around distillation complaints while pirating books, settling with authors, gatekeeping access, Mythos claims questioned if distillation attacks work against it" (4 comments)
[2026-06-10T16:23:30.032Z] Topic 18 summarized (788 in, 107 out)
[2026-06-10T16:23:30.077Z] Summarizing topic 19/20: "China Market Pricing # Observation that in every market China dominates products remain inexpensive including solar panels batteries EVs drones, counterargument that Chinese Big Tech would be as bad as US Big Tech if they dominate" (3 comments)
[2026-06-10T16:23:37.197Z] Topic 19 summarized (1223 in, 124 out)
[2026-06-10T16:23:37.255Z] Summarizing topic 20/20: "Deleted Post Dynamics # Original tweet was removed, preserved via Pastebin, discussion about whether to wait for OP to repost with edits, community effort to preserve and share the content" (5 comments)
[2026-06-10T16:23:49.801Z] Topic 20 summarized (720 in, 110 out)
[2026-06-10T16:23:49.830Z] Summarization complete: 18 topics, 19856 input tokens, 2252 output tokens
[2026-06-10T16:23:49.854Z] Completed step: summarize_topics in 168054ms
[2026-06-10T16:23:49.901Z] Job completed successfully

LLM Invocations (Total: $0.0726)

← Back to all jobs