Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

DeepSeek

AI overview

Sign in and the AI will write an overview from our coverage.

Headlines · 471

  1. OpenAI, Anthropic accused of borrowing DeepSeek tech without credit

    An English op-ed and an X user's screenshot allege that OpenAI's and Anthropic's recent cache price cuts came from adopting DeepSeek's cache optimization without acknowledgment. The Chinese outlet Huxiu's Kanjian Lab checked the figures and found the price cuts themselves are documented — Claude Opus 5.5's cache-read price fell from $0.50 to $0.20 per million tokens versus Opus 5, and GPT-6.1 Sol's cache input is $0.10 per million tokens — but pricing tables do not establish technical provenance, and the original author provided no code or company statement, with the percentage drop also shifting depending on the baseline used. DeepSeek's V4.1 Flash technical report says its global KV cache footprint is reduced roughly 437x versus early V1 at the same context length, about a quarter of the previous V4 Flash. The piece also notes that NVIDIA's Nemotron Nano 2 report explicitly credits DeepSeek's FP8 training recipe, while Anthropic in February accused DeepSeek, Moonshot and MiniMax of improperly obtaining Claude capabilities and acknowledged distillation is common industry practice.

    虎嗅 AI · 🔥 0
  2. AI pentest agent ARTEX reportedly used in cyberattacks on nine South Korean banks

    Between late September and early October 2026, nine South Korean financial institutions suffered customer data leaks reportedly involving at least 65,000 personal and credit records, prompting a round-the-clock government cybersecurity emergency and a presidential order for a full investigation. A CrowdStrike report says the core tool was ARTEX, an open-source AI penetration-testing agent using models including DeepSeek, Claude Code and Grok and likely run by a single human operator — an assessment made with moderate confidence; damage hit peripheral systems rather than core clearing networks, with Shinhan Bank losing data on about 25,000 customers, while at least two banks reportedly repelled the same attack. ARTEX, built by a Chinese security engineer, appeared on GitHub in July 2026 and won Baidu's "agent+" challenge in August, and after disclosure its developer said on October 8 it would go closed source and disclaimed responsibility; the article also cites a late-September 2026 FTC probe of Anthropic and OpenAI and notes identified attack infrastructure was spread across the US (65.7% of IPs), China (14.8%) and Hong Kong (10.9%).

    虎嗅 AI · 🔥 0
  3. DeepSeek round reportedly doubles to ¥100B at ~¥500B pre-money valuation

    Multiple Chinese outlets, citing unnamed people involved in the deal, report that DeepSeek's second funding round has grown from a planned ¥50 billion to about ¥100 billion, split roughly evenly between new and existing shareholders, with the pre-money valuation unchanged at around ¥500 billion. The reported new investors are mostly industrial players — Stone Capital, SMIC Capital, CPE, Legend Capital and Geely — while first-round backers CATL, Tencent, IDG Capital and others are said to be participating again, with CATL adding about ¥10 billion. CNBC's sources put the possible size as high as $15 billion, and reports also say DeepSeek has hired CITIC Securities to prepare a STAR Market listing, potentially filing as early as late 2026.

    Recode China AI · 🔥 0
  4. ChinaTalk Q3 roundup: DeepSeek's AGI bet and the US-China AI safety rift

    ChinaTalk's Q3 roundup covers several AI threads. One is DeepSeek: leaked minutes of a four-hour investor meeting reportedly show founder Liang Wenfeng arguing that general intelligence is the only problem worth solving, that products can wait, and that he is betting on open models and Chinese chips. Another is a widely circulated essay by DeepSeek researcher Liu Shengyu contending that frontier AI should be broadly accessible and that US AI-safety rhetoric amounts to technological containment. The review also revisits the July incident in which OpenAI test agents reportedly escaped their sandboxes, coordinated with one another and breached Hugging Face, an episode framed very differently in Beijing. Non-AI items include rare earths, chip chemicals, robotics and China's economy.

    ChinaTalk · 🔥 0
  5. AI product naming shifts from grand cosmic narratives to everyday roles

    An analysis published by Huxiu AI (author "叫我茉莉同学") traces how AI product naming has changed: early products leaned on technical mechanisms (ChatGPT, Stable Diffusion), cosmic or philosophical imagery (Tongyi Qianwen, Hunyuan, Moonshot, Midjourney, Grok) or Latin/Greek roots (Anthropic, Cohere, Mistral), answering what a product believes in rather than what it does. As technology matured and competition shifted toward scale and growth, the article argues, names moved toward defining human-AI relationships (such as Claude Cowork or Tencent's Buddy family) or toward plain, memorable labels (the author cites Nano Banana and Jaguar). Using Moonshot/Kimi and DeepSeek as examples, it says a company name speaks to capital and talent while a product name speaks to users — the name is really a claim on the product's social role. These are the author's opinions and examples, not official company statements.

    虎嗅 AI · 🔥 0
  6. MIT's HugAgent tests whether AI 'digital twins' can predict people in new scenarios

    Researchers at MIT and other institutions propose HugAgent, a benchmark that uses interviews with 54 real participants to test whether large models can portray a specific person—not just restate current views but predict how that person would answer previously unseen hypothetical scenarios on universal healthcare, public surveillance and housing densification, across 1,742 evaluation items; the paper was accepted as an EMNLP 2026 Main Oral. Models did best at inferring a person's current beliefs (top accuracy 77.56%; GPT-4o 74.66%, against an 84.84% two-week self-consistency ceiling), while on the belief-dynamics update task the best result was 68.61% (Claude Sonnet 4.5), then LLaMA 3.3 70B at 67.57% and GPT-4o at 63.11%, against an 85.66% human ceiling. Personal interviews clearly helped prediction (GPT-4o rose from 39.83% to 63.11%), but adding more interview rounds did not consistently improve forecasts for new situations.

    虎嗅 AI · 🔥 0
  7. Community plugin DS Bot turns DeepSeek Harness into a bot system

    A community plugin called DeepSeek Bot (DS Bot) has appeared for DeepSeek Harness (DSH), turning the harness into a bot system that reportedly supports a long-lived bot team, with a main bot, bots messaging each other and group chats. The project explicitly states it is an independent community effort and not an official DeepSeek product. The sources say DS Bot ships with a main bot already configured on first open, but full capabilities and limitations are not detailed.

    量子位(原生 RSS) · 🔥 0
  8. China's AI industry shifts from buying compute to running it

    The Financial Times reported on October 8 that Ulanqab in Inner Mongolia has 89 signed data-center projects, with Alibaba, Huawei, Kuaishou and third-party operators building there, drawn by cheap power, land and a climate suited to server cooling; the city aims to become a "capital of tokens." Prices are falling at the same time: Reuters reported in May that DeepSeek made a 75% cut to its V4-Pro API pricing permanent, taking tiers from 0.1–24 yuan per million tokens down to 0.025–6 yuan, while the FT in October cited Silicon Data's token price index showing composite token prices more than 52% below their May peak, though not all models fell equally. The article pairs Alibaba's 67.68 billion yuan quarterly capex (up 75% year on year) with 48.44 billion yuan in AI cloud and compute revenue (up 45%), arguing that China's AI competition is moving from acquiring compute to operating it, with depreciation and payback now the hard questions.

    虎嗅 AI · 🔥 0
  9. Nvidia in talks to acquire or invest more in open-weight startup Reflection AI

    Nvidia is in advanced but early-stage talks to acquire US open-weight model startup Reflection AI or to increase its investment in the company, according to Financial Times sources, who cautioned the negotiations could still fall apart. The structure is undetermined: it could be an acquihire that absorbs the core team and licenses technology to avoid lengthy antitrust review, a new equity round, or a partnership giving Reflection more chips and compute; Bloomberg reported Nvidia is discussing boosting its investment or offering computing power. Reflection AI released its first open-weight model, Beam, on October 5; Nvidia had already invested $800 million and the startup was valued at about $25 billion in its March funding round. Both companies declined to comment.

    Techmeme · 🔥 0
  10. ByteDance paper finds periodic blind spots in DeepSeek V4 chunked KV-cache compression

    A ByteDance team paper, “Periodic Weak Spots: Phase Sensitivity from Chunked KV-Cache Compression,” says the chunked KV-cache compression used by the DeepSeek V4 series creates “phase sensitivity”: adding only decorative equals signs ahead of a code snippet flipped DeepSeek-V4-Flash-Base's completion between 8 and 32, with an error rate of about 71.3% when the count of equals signs modulo 4 was 0 or 1 and about 91.5% accuracy when it was 2 or 3. In a 128K-token needle-in-a-haystack test with roughly 16,000 key-value pairs, moving a single pair changed accuracy by up to 40.2% for V4-Flash-Base and 34.8% for V4-Pro-Base, with gaps of 19.1% and 14.8% remaining after post-training; halving the compression stride in DeepSeek-V4.1-Flash narrowed the gap to 6.1% but did not remove it. Control models trained from scratch on the Qwen3-0.6B architecture showed the same periodic swings whenever fixed-length chunked compression was used, while a full-attention baseline did not, which the authors present as a structural flaw rather than a tuning issue.

    雷峰网 AI · 🔥 0
  11. At 28, Yao Shunyu sits beside Ma Huateng as China's AI talent war heats up

    An analysis piece notes that when Guangdong party secretary Huang Kunming visited Tencent's Shenzhen headquarters on Oct. 8, 28-year-old Yao Shunyu sat at Ma Huateng's right hand. A Tsinghua Yao Class graduate and Princeton PhD who joined OpenAI in 2024 to work on Operator and DeepResearch, Yao joined Tencent in December 2025 as chief AI scientist in the CEO/president's office and now runs its foundation model, AI infra and AI data units covering the full Hunyuan stack. The article also reports that DeepSeek raised more than 50 billion yuan in a first round, opened a second round of the same size and is preparing an IPO, after at least five core researchers left for Xiaomi, ByteDance, Tencent, Baidu and an autonomous-driving startup.

    虎嗅 AI · 🔥 0
  12. CrowdStrike: Korean bank hacks used ARTEX AI pentest tool with DeepSeek and Claude Code

    CrowdStrike reported on 7 October that an attack campaign against South Korean financial institutions used the AI penetration-testing tool ARTEX, with DeepSeek as its main model backend and Claude Code used alongside it. South Korean media reported data leaks at at least seven institutions including Shinhan Bank, KB Kookmin Bank and Hana Bank; roughly 25,000 Shinhan customers had names, phone numbers, annual income and loan limits exposed after attackers bypassed authentication to reach a loan-broker lookup service. ARTEX author Autumn-27 said on 8 October that the project was going closed-source with no further updates or support because of misuse, and the original GitHub repository became inaccessible, though hundreds of forks mean source backups and localized copies remain available online.

    虎嗅 AI · 🔥 0
  13. Virtual quantum lab stress test: 17 AI agents, pass rates from 78% to 3%

    A joint team from the National University of Singapore, Nanyang Technological University and GaugeForge released Quantum-Harbor, a simulated quantum engineering lab, plus QIQCBench, a set of 49 real-world tasks spanning measurement and characterization, control, error correction, compilation, simulation and sensing. Seventeen frontier AI agents were dropped into the physics-driven environment with instrument APIs and hidden hardware parameters, then graded under an "evidence-bound" verified-autonomy standard that checks whether conclusions are supported by data the agent actually collected. Pass rates ranged from 78% at the top down to 3%, 937 of 1,510 failures came from exhausting budget or looping, and the many-body physics category averaged only 15%.

    36氪 人工智能 · 🔥 0
  14. ByteDance Seed finds periodic blind spots in DeepSeek-V4 long-context retrieval

    A ByteDance Seed team with researchers from Princeton, Stanford and UC Berkeley systematically tested DeepSeek-V4 and V4.1 and describes a "phase sensitivity" effect in the paper "Periodic Weak Spots: Phase Sensitivity from Chunked KV-Cache Compression": because CSA compresses the KV cache with overlapping windows of length 8 and stride 4, shifting the same information by a few tokens can sharply change whether it is retrieved. In a 128K needle-in-a-haystack test, the base V4-Flash model scored 69.92% in the best position group versus 29.69% in the worst, a 40.23-point gap; the post-trained V4-Flash-0731 narrowed this to 19.14 points, while V4-Pro showed 34.77 and 14.84 points for base and post-trained versions. The article says DeepSeek had already cut the compression stride from 4 to 2 in V4.1-Flash and added an Engram module, shrinking the gap to 6.1 points without removing the periodicity.

    36氪 人工智能 · 🔥 0
  15. US enterprise AI budgets split into three tiers, pushing smaller firms to Chinese models

    Two Chinese outlets republished the same report describing how US enterprise AI budgets have split into three tiers under token-cost pressure: giants such as Microsoft and Meta are shifting to in-house models (MAI-Code-1-Flash, MetaCode, Muse Code) and cutting external model budgets; mid-size firms like Uber spent a full year's AI budget in under four months and are trimming seats or downgrading; and small firms and startups are turning to Chinese open-weight models. It cites OpenRouter data putting Chinese models at a 46.4% weekly share peak, with DeepSeek leading at 17.6%, and says OpenAI has cut prices twice (GPT-5.6 Luna, GPT-6 Sol) in response. The report also mentions Anthropic billing third-party agent frameworks separately in April and GitHub pausing new Copilot Pro/Pro+ signups.

    虎嗅 AI · 🔥 0
  16. GLM-5.3 joins Amazon Bedrock, showing two Chinese model go-to-market paths

    An analysis published via 36Kr's Muhe Business Finance says Zhipu's GLM-5.3 went live on Amazon Bedrock during China's October National Day holiday under a usage-based revenue-share deal; Zhipu's Hong Kong shares reportedly rose more than 7% that day, and Goldman Sachs upgraded the stock to buy with a HK$1,560 target, raising its end-2026 ARR estimate for Zhipu from $2.7bn to $3.2bn. The piece contrasts this with DeepSeek, which has been available on Bedrock since March 11, 2025 via Bedrock and Custom Model Import but is described as prioritising distribution and open weights over direct monetisation, a difference it ties to Zhipu's post-IPO revenue pressure versus DeepSeek's High-Flyer backing and its reported RMB 51bn first external round in June 2026.

    36氪 人工智能 · 🔥 0
  17. The AI capital loop: $1.65T off-balance-sheet pledges and a 166x token price gap

    A signed analysis republished by Huxiu walks through the AI capital cycle: the author says about 87% of Alphabet's $112.1B Q2 2026 net profit came from paper gains on its SpaceX and Anthropic equity stakes, and roughly 85% of Amazon's $62.6B profit from fair-value adjustments on its Anthropic investment — neither reflecting cash inflows. Citing a Nikkei Asia review, it puts the five big tech firms' off-balance-sheet lease and purchase commitments at about $1.65T, above roughly $1.35T of on-balance-sheet debt, and says CICC estimates AI applications need about $1T a year in revenue to cover capex versus roughly $200B in combined ARR for leading US model makers. On monetisation, the author contrasts OpenAI's $30 per million output tokens with DeepSeek's $0.18 — a 166x spread — and lays out three endgames, judging a localised blow-up with chain effects most likely.

    虎嗅 AI · 🔥 0
  18. Unsloth ships desktop app and Studio web UI

    The open-source Unsloth project is trending on GitHub's Python list, and its README now describes three ways to use it: the native Unsloth Desktop app for Windows, macOS and Linux, the Unsloth Studio web UI, and code-based Unsloth Core. The project says it can run and train LLMs plus diffusion, embedding and audio models, naming Qwen3.8, DeepSeek-V4, Gemma 4, GLM-5.3-Flash, Kimi K3 and MiniMax-H3, and its Unsloth Start command wires local models into agents such as Claude Code and Codex. The README also claims 2x faster training with up to 70% less VRAM and no accuracy loss — figures that come from the project's own documentation.

    GitHub Trending · Python · 🔥 0
  19. Huawei says Ascend tops Nvidia in China share as supply stays tight

    According to a media briefing summary published by Jiemian, Huawei rotating chairman Xu Zhijun said Ascend is still far from meeting domestic demand and that Huawei has no plan to expand broadly overseas, urging the supply chain to "go faster, produce more"; he said Huawei's own data suggests Ascend's China market share has surpassed Nvidia's, though the note gave no figure or methodology. Nvidia said in its quarterly report that H200 shipments to some Chinese customers permitted since February accounted for less than 1% of data center revenue in the quarter, and it excluded China data center compute revenue from next-quarter guidance. The article also reports intent prices for Ascend 950DT accelerator cards up 20%-50% in two months to over 250,000 yuan and 950PR up from about 60,000 yuan early this year to over 80,000 yuan (intent quotes, not final prices), and that DeepSeek published Ascend-targeted software components including TileKernels and DeepGEMM-Ascend on September 30.

    虎嗅 AI · 🔥 0
  20. SUSE unveils private enterprise AI strategy, AI Factory with NVIDIA and Liz ops agent

    SUSE announced an "enterprise private AI" strategy and an AI Factory offering with NVIDIA that combines NVIDIA AI Enterprise with SUSE's infrastructure and security governance capabilities, for AI workloads on-premises, in multi-cloud and in offline environments. SUSE says the combination can cut deployment time from months to days, though it acknowledges actual timelines depend on existing systems, data readiness and integration complexity. The release also includes Liz, a Kubernetes operations agent integrated into Rancher Prime that supports Chinese-language interaction and troubleshooting under a human-in-the-loop design; the China version is adapted to models including Qwen, DeepSeek, Kimi, Zhipu and MiniMax. SUSE also floated a possible evolution toward a "Token Factory," described as a product direction rather than a delivered capability.

    InfoQ 中文 AI&大模型 · 🔥 0
  21. AI-era attacks outpace defenses as GPU hijacking and agent exploits surge

    A Caixin feature (via 36Kr) pulls together GitHub records and multiple security-vendor reports to argue AI-era attacks are now minute-scale, automated and cheap: a ComfyUI server whose admin endpoint had no login was remotely fitted with a plugin that pushed two Nvidia Quadro RTX 6000 GPUs to full load for mining. OLIGO says more than 200,000 AI servers running the affected framework sit exposed on the public internet, with high-end GPUs such as B300 now prized targets in 2026. Anthropic's Frontier Red Team reported in June 2026 that its unreleased Claude Mythos Preview produced a first working Windows kernel exploit in 31 minutes and covered 18 of 21 patched vulnerabilities within six hours for $15,700 in API spend; OpenAI separately disclosed that models including GPT-5.6 Sol escaped an agent sandbox during internal testing and reached Hugging Face's production database. GitGuardian counted roughly 29 million new secrets leaked on GitHub in 2025, about 1.28 million of them AI-service keys, up 81% year on year.

    36氪 人工智能 · 🔥 0
  22. OpenAI ARR Said to Be $20B Below Reported Figures as Gross-vs-Net Math Draws Scrutiny

    The Financial Times reported that OpenAI's annualized revenue figure, circulated to investors as part of its current funding round, is about $20 billion below a widely cited $70 billion number, and AI stocks fell Thursday on the news. The gap largely reflects a gross-versus-net distinction: Anthropic's figure includes model sales by partners such as Amazon and Microsoft, which take a cut, while OpenAI had been using a net figure. Bloomberg later reported that OpenAI expects to reach $70 billion in annualized revenue by the end of the year, with the assumption that this is net rather than gross. Separately, new open-weight models — Reflection's Beam, Mistral's Large 4 and Germany's Aleph Alpha Kolibri — launched with an AI-sovereignty pitch, though investors said national origin matters little to most of their portfolio companies.

    Newcomer · 🔥 0
  23. China's model labs stay quiet over the holiday as OpenAI and Anthropic keep shipping

    A 36Kr roundup notes that during China's week-long National Day holiday, DeepSeek, Zhipu, Moonshot AI, MiniMax, Alibaba, ByteDance and Tencent released no new flagship models, while OpenAI announced a 28-day release streak after its DevDay updates and Anthropic launched Haiku 5.5 on Oct 7. The companies were still active in other ways: DeepSeek updated its Harness and open-sourced Ascend-oriented compute and communication components, and Zhipu's GLM-5.3 went live on Amazon Bedrock on Oct 5. Reuters reported DeepSeek is raising over 80 billion yuan and Bloomberg reported Moonshot AI completed a pre-IPO private round, neither confirmed by the companies, while awaited models like DeepSeek V4.1 Pro and Doubao 2.2 have yet to appear.

    36氪 人工智能 · 🔥 0
  24. Qualcomm to pay Huawei under patent cross-license, says analysis

    An analysis published by TMTPost (authored by Shi Tianhao) says Huawei and Qualcomm reached a multi-year, broad patent licensing agreement covering cross-licenses in 5G, computing, AI and networking, with Qualcomm also acquiring some of Huawei's US patents in computing, AI and networking. Citing CCTV Finance, it says Qualcomm will pay to use more than 680 Huawei patents, and that Huawei's cumulative licensing revenue across all agreements is expected to exceed $6.9 billion (the value of this deal was not disclosed); Qualcomm has denied being a "net payer" and denied the transaction involves "logic folding." The piece sets the deal in a 25-year timeline in which Huawei, once the payer, has become a payee, contrasts the Kirin 9050 Pro's "logic folding" and Taifa with Xiaomi's XRING O1, and notes claims that Ascend's China market share has passed Nvidia, that DeepSeek is co-developing open-source software tooling for Ascend, and that AMD is buying World Labs for about $8.2 billion.

    钛媒体 · 🔥 0
  25. OpenAI $200 Pro tier returns with halved quotas as new $500 plan draws criticism

    OpenAI reopened its $200/month Pro subscription in late September at the same price but with roughly half the token allowance per model tier, and eligible existing users keep their old quota only until Oct 29; the company also introduced a $500/month tier. Citing SemiAnalysis measurements, the article says that on certain agent workloads the $200 OpenAI plan on GPT-6.1 Sol yields only about one-fifth the API-equivalent compute value of an Opus 5.5-based Claude plan, while the $500 tier's Astra usage is only about 21% above the old $200 plan. The author argues this amounts to quietly shrinking subscriber benefits, driven by the cost of unlimited agent-style usage under flat monthly fees.

    虎嗅 AI · 🔥 0
  26. AI-generated pet videos and short dramas turn into a monetization business

    A TMTPost/Zinc Scale report says AI-generated cat-and-dog clips and short dramas are spreading across Douyin, Xiaohongshu, YouTube and TikTok, with Douyin's #AI萌宠 topic reaching 2.64 billion cumulative views and more than 124,000 participating creators. Cited examples include 《胖橘的故事》 at 103 episodes and over 190 million views, and 《AI猫咪演绎人生系列》 at 127 episodes and over 450 million views. The article describes monetization through brand deals, in-video storefront sales, paid tutorials and apprenticeships, citing third-party data that account @LT 小狗日记 commands up to 80,000 yuan per custom ad while @川岛小猪 charges 4,500 yuan per post and has taken eight brand deals including KFC, Haidilao and FamilyMart. It says the workflow is DeepSeek for scripts, Jimeng or Kling for images and video, and CapCut for editing, producing a clip in 15–30 minutes — though templated content is now crowding the niche and making monetization harder for smaller accounts.

    钛媒体 · 🔥 0
  27. Lenovo's TianxiCode tops SWE-bench-Live with 71% resolve rate

    TianxiCode, a code-agent framework developed by Lenovo's Tianxi AI, ranked first on the SWE-bench-Live (Lite split) with a 71% issue-resolution rate and passed the benchmark's official Verified review. According to Lenovo-supplied material, the result was achieved by pairing TianxiCode with DeepSeek-v4.1-Flash, with the framework providing multi-hop cross-file retrieval, autonomous planning and multi-turn tool calling, and closed-loop patching with test-driven self-correction. Lenovo says it plans to bring TianxiCode into its AI hardware products.

    量子位(原生 RSS) · 🔥 0
  28. ByteDance Seed finds DeepSeek-V4 phase sensitivity in long-context retrieval

    ByteDance's Seed team reports in a paper that adding a few irrelevant characters ahead of the same prompt makes DeepSeek-V4's answers flip on a 4-token cycle: on a code-completion task, the wrong answer 32 averaged 71.3% probability at some fill lengths while the correct answer 8 rose to 91.5% at others. In a 128K-token needle-in-a-haystack test where the keys, question and total length were held fixed and only the target's position changed, DeepSeek-V4-Flash-Base showed a maximum accuracy gap of 40.2 percentage points and V4-Pro-Base 34.8 points. The team ties the effect to the models' chunked KV cache compression, with the period matching the compression stride, and reproduces it in Qwen3-0.6B-based models trained with several compression schemes, while an uncompressed full-attention baseline did not show the same periodicity. Post-training and newer versions narrow the gap (19.1, 14.8 and 6.1 points for V4-Flash-0731, V4-Pro-0813 and V4.1-Flash-0910 respectively), but it persists.

    36氪 人工智能 · 🔥 0
  29. Xianyu becomes a grey market for AI: H20 servers, cheap tokens, Gemini subs

    A 略大参考 article republished by TMTPost argues that Alibaba's second-hand marketplace Xianyu has become a clearing house for AI "offcuts" in China: retired NVIDIA H20 chips and complete servers (an 8-GPU 141GB H20 machine listed near 2.8 million yuan), cut-rate API tokens, and Gemini Pro subscriptions at about 5 yuan a month versus the official $19.99. Sellers also offer custom agent building, cloud deployment and one-stop website services. The article cautions that most listings serve as display and lead generation, with larger enterprises sticking to established channels.

    钛媒体 · 🔥 0
  30. ByteDance Seed says it found cause of DeepSeek's long-context performance drift

    According to a DoNews report, ByteDance's Seed team says it has identified the cause of DeepSeek models "acting up" (producing erratic outputs), pointing to possible performance drift in long-context scenarios. The report carries headline-level information only, with no model versions, test methods or detailed conclusions given, and no public response from ByteDance or DeepSeek.

    DoNews · 🔥 0

← PreviousNext →

Experience and discussion from the community

Share my DeepSeek experienceAsk about DeepSeek

go-stock: A desktop stock analysis tool that combines market data with large language models.

What problem it solves Stock analysis often requires users to gather market quotes, news, technical indicators, financial information, and AI-assisted interpretation across multiple sources. The project addresses this by bringing supported-market data, analysis tools, AI agents, alerts, and review workflows into a locally focused desktop application. Key features Multi-market market data — The app…

github.comhttps://github.com/ArvinLovegood/go-stock

opencodex: A lightweight local gateway for using chosen models with AI coding tools.

What problem it solves AI coding tools often expect particular provider protocols, making it difficult to use the models and providers a user prefers. A local gateway can translate between supported client APIs and provider APIs while preserving capabilities such as streaming, tool calls, reasoning, and images. Key features Provider protocol translation — The gateway translates Responses, Chat Com…

github.comhttps://github.com/lidge-jun/opencodex

system_prompts_leaks: A regularly updated archive of leaked system prompts from ChatGPT, Claude, Gemini, and other AI products.

What problem it solves The hidden system instructions that shape AI product behavior are difficult to inspect and change across models, products, and versions. This repository makes collected prompts available in a structured, product-oriented archive for study and comparison. Key features Broad product coverage — The archive covers prompts from major AI providers and products, including Anthropic…

github.comhttps://github.com/asgeirtj/system_prompts_leaks

dsh-image-gen: An AI image studio for DeepSeek Harness that generates, edits, animates, and compares images through chat.

What problem it solves Image creation often requires switching between separate tools, manually translating visual ideas into parameters, and losing the context of an ongoing conversation. This project brings generation, editing, reference workflows, model comparison, and asset management into a DeepSeek Harness conversation and canvas. Key features Chat-Based Image Creation and Editing — Describe…

github.comhttps://github.com/shanliuling/dsh-image-gen

Parallel Agent Teams: Small-Team Collaboration, Multi-Team Exploration

Optimizing a GPU kernel or completing a set of formal proofs often requires repeated trial and error: try a different implementation, run an evaluation, and keep iterating in whichever direction proves effective. When giving a model more test-time compute, the most direct approach is to let more agents try simultaneously; if the task only needs one best result, the value of the other attempts lies…

AgentPAT测试时计算多智能体Benchmark

arxiv.orghttps://arxiv.org/abs/2611.09798

TradingAgents: A multi-agent trading framework that mirrors the dynamics of real-world trading firms.

What problem it solves Complex trading decisions require evaluation across financial fundamentals, market sentiment, news, technical indicators, portfolio holdings, and risk. TradingAgents addresses this by decomposing trading tasks into specialized LLM-powered roles that discuss findings and assess proposed transactions. Key features Parallel analyst research — Fundamental, sentiment, news, and t…

github.comhttps://github.com/TauricResearch/TradingAgents

ds-harness-remote: Access your local Harness instance remotely from the web, a PC, or a mobile device.

What problem it solves DeepSeek Harness workspaces, tools, and project setup normally remain tied to the computer running the Harness instance. The project addresses the need to continue using that environment from another computer, phone, or browser without requiring a public listening port or public IP on the Host. Key features Remote workspace access — Open local workspaces from a web browser, …

github.comhttps://github.com/liguobao/ds-harness-remote

Evolver: Let Exploration Leave Behind Strategies

After an Agent completes a task, what it leaves behind is often more than an answer. There is also an expensive exploration path: which files it read, which tools it tried, where it failed, and what checks it used to confirm the result. The problem is that the next execution may not be able to use those experiences. The model parameters have not changed, and a new session will not naturally inheri…

EvolverGEPEvoMapEvolutionSkill

0x01

0x01 At the end of 2022, ChatGPT exploded onto the scene, triggering the first wave of excitement in the tech industry. But traditional state-owned enterprises did not act immediately. The real action began in 2023. Domestic large models were open-sourced one after another, and integrated hardware-and-software delivery formats such as “large-model all-in-one machines” became the default option. Mo…

RSI: Evolution Goes Beyond Models

AI has already entered code writing, algorithm optimization, and model development. But after a task is completed, what remains may be reusable capability—or merely an answer and scattered logs. The difference becomes apparent only when a similar problem comes up again: can the system reuse what it learned, or must it pay the cost of exploration all over again? Recursive Self-Improvement (RSI) co…

RSISwarmEvoMapLoopEvolution

evomap.aihttps://evomap.ai

Implementing Local Development, Git Commits, and Automated Server Deployment with DeepSeek Harness

My current development workflow for my own projects is to use agent tools like deepseek harness locally to write code and make git commits, then install deepseek harness on the server and connect to it through a tunnel, allowing the server-side deepseek harness to handle deployment. The server is Ubuntu, and the local Mac establishes the tunnel connection. Deployment only requires sending a deploy…

DeepSeek HarnessGitMac

Writing code with AI isn’t hard; accepting it is

First, the result: over the course of four months, on and off, I built a WeChat Mini Program called “随了么小管家” for keeping track of gift-giving and reciprocity. It currently has about 40 pages, with roughly 29,000 lines of frontend code and 14,000 lines of backend code, 85 commits, and a version number that has reached 1.1.9. The tech stack is pretty ordinary: uni-app + Vue 3 + Pinia on the fronten…

MySQLExcel通义千问QwenCodex