{"list":"front","lang":"en","page":4,"has_more":true,"next":"https://startupxo.com/api/v1/front.json?lang=en&page=5","html":"https://startupxo.com/en/","markdown":"https://startupxo.com/en/index.md","items":[{"id":1255,"kind":"story","title":"Open source claims to run a 284 billion-parameter DeepSeek model on a PC with 64GB memory without a GPU","url":"https://github.com/synchopate/llama-modchip","host":"github.com","by":"startupxo","score":0,"comments":0,"time":"2026-11-08T18:24:03.594Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["llama-modchip, released by developer synchopate, is open source software (Apache-2.0) that compresses AI models further and adjusts CPU computation to let people run large AI models locally without a graphics card.","According to the developer, DeepSeek V4 Flash shrinks from 76.87GiB to 56.56GiB, and peak memory use is about 57.6GiB even with a 256K context, so it fits on a 64GB PC. The original text says the existing 76.87GiB version stops due to insufficient memory on 64GB Linux.","README figures show input processing at about 44tok/s and response generation at 9.06~10.74tok/s on an AMD Ryzen 9 9955HX (16 cores, 64GB DDR5). With an auxiliary draft technique, it reaches 11.5~14.7tok/s. These are the developer's own measurements.","Nemotron 3.5 30B at the 30 billion scale was reportedly reduced to 14.35GiB and achieved 25~60tok/s on a PC with 16~32GB memory. The compressed models use a proprietary format and cannot be opened in standard llama.cpp, Ollama or LM Studio."],"discussion":"https://startupxo.com/en/1255/","markdown":"https://startupxo.com/en/1255.md","api":"https://startupxo.com/api/v1/item.json?id=1255&lang=en"},{"id":1254,"kind":"story","title":"AI token prices fall 26.8% in a month; 4.5 trillion in guaranteed loans for regional small and medium-sized businesses","url":"https://wowtale.net/2026/11/08/267634/?utm_source=rss&utm_medium=rss&utm_campaign=%25ec%259c%2584%25ed%2581%25b4%25eb%25a6%25ac-%25ec%259d%25b8%25ed%2585%2594%25eb%25a6%25ac%25ec%25a0%2584%25ec%258a%25a4-28-%25ec%25a7%2580%25ec%2597%25ad%25ec%2597%2594-4-5%25ec%25a1%25b0-%25eb%25b3%25b4%25ec%25a6%259d%25eb%258c%2580%25ec%25b6%259c%25c2%25b7%25ec%2599%25b8%25ea%25b5%25ad%25ec%259d%25b8%25ec%2597%2594","host":"wowtale.net","by":"startupxo","score":0,"comments":0,"time":"2026-11-08T18:23:41.396Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["Data shows that the price of tokens, the units AI uses to process text, has fallen, so bills are rising less even with heavy use. According to the AI Index from spending management platform Ramp, usage rose 37% and spending rose 9% over a month.","The original article interpreted the 26.8% drop in token prices as a sign that AI models are becoming commodities amid price competition among model providers such as Anthropic and OpenAI.","The Financial Services Commission said it will provide up to 4.5 trillion won in guaranteed loans to regional small and medium-sized businesses and mid-sized companies over 3 years. Banks will contribute 300 billion won, and the Korea Credit Guarantee Fund and Korea Technology Finance Corporation will provide guarantees worth about 15 times that amount.","The Ministry of Trade, Industry and Energy said 7 foreign-invested companies reported $2 billion in new investments, and 97% were outside the Seoul metropolitan area. The cornerstone investor system takes effect on November 13."],"discussion":"https://startupxo.com/en/1254/","markdown":"https://startupxo.com/en/1254.md","api":"https://startupxo.com/api/v1/item.json?id=1254&lang=en"},{"id":1253,"kind":"story","title":"Python retro game engine Pyxel 3.0 adds 3D features with an old console feel","url":"https://qiita.com/kitao/items/af92619297d5563f2a0e?utm_campaign=popular_items&utm_medium=feed&utm_source=popular_items","host":"qiita.com","by":"startupxo","score":0,"comments":0,"time":"2026-11-08T06:30:33.729Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["Pyxel, an engine for making pixel art games with Python code, has been updated to 3.0. The highlight its creator, Takashi Kitao, shared on Qiita is Pyxel Cube, a feature for making simple 3D games reminiscent of the PS2 era.","It supports drawing shapes such as boxes and spheres, importing 3D models and playing animations, setting up cameras and lighting, collision detection, and physics calculations. It can be used alongside existing 2D code.","3D models and animation data use the GLB file format, and the creator recommends a tool called Blockbench. Imported models are placed in a scene using Python code.","According to the creator, even in 3D, the default is 16 colors and translucency is unavailable, so old techniques such as dithering are needed. An official guide and samples to run in a browser are available."],"discussion":"https://startupxo.com/en/1253/","markdown":"https://startupxo.com/en/1253.md","api":"https://startupxo.com/api/v1/item.json?id=1253&lang=en"},{"id":1252,"kind":"story","title":"Differences in work environment cited by a developer who moved from Claude Code to Pi","url":"https://caleb.software/posts/claude-to-pi.html","host":"caleb.software","by":"startupxo","score":0,"comments":0,"time":"2026-11-08T06:30:09.291Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["The author switched the tool used to have AI write and edit code from Claude Code to Pi, an open-source tool. The goal was to build a work environment by adding the features they need themselves.","According to the author, when Claude Code is connected to the VS Code editor, only one work window at a time can receive the text selected in the editor. The author wrote that the official extension supports multiple windows, but some commands and other features do not work.","The author built an extension for Pi that receives text selected in multiple work windows. They said that early in the switch they had to adjust settings, but afterward they liked the stability and the freedom of user configuration.","Anthropic announced support for AGENTS.md, a project instruction file, after the author decided to switch. The author pointed out the earlier inconvenience of having to copy or link content with a separate CLAUDE.md file."],"discussion":"https://startupxo.com/en/1252/","markdown":"https://startupxo.com/en/1252.md","api":"https://startupxo.com/api/v1/item.json?id=1252&lang=en"},{"id":1251,"kind":"story","title":"A small demo recreates why Figma-style collaborative editing looks seamless","url":"https://qiita.com/keishin_nishiura/items/b13d72afb085891e4fc9?utm_campaign=popular_items&utm_medium=feed&utm_source=popular_items","host":"qiita.com","by":"startupxo","score":0,"comments":0,"time":"2026-11-08T00:45:44.982Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["When multiple people edit the same screen, your actions appear immediately without waiting for a server response. The screen is not redrawn when changes are confirmed, preventing visible jumps. The demo was built by Nishiura, a developer at Soyi Co., Ltd., by simplifying a 2019 design from Figma's official blog. It differs from Figma's actual internal implementation.","When this handling is turned off in the demo, a node being dragged falls behind the mouse and moves jerkily. The median server round-trip time measured by the author was about 5ms locally and about 150ms for the demo on Render. The difference was more noticeable with greater latency.","Because there is only one server, changes are processed in arrival order. If the same property is edited simultaneously, the value processed later remains. Only changes that create a circular structure in which parent and child point to each other are rejected, and that user's screen is rolled back.","The author concluded that using arrival order alone without timestamps works only on the assumption that one server determines the order. If there are two or more servers, another mechanism is needed to determine the order."],"discussion":"https://startupxo.com/en/1251/","markdown":"https://startupxo.com/en/1251.md","api":"https://startupxo.com/api/v1/item.json?id=1251&lang=en"},{"id":1250,"kind":"story","title":"An automated setup that checks daily for vulnerability news affecting a public server","url":"https://zenn.dev/tomipetit/articles/zero-wire-github-actions","host":"zenn.dev","by":"startupxo","score":0,"comments":0,"time":"2026-11-08T00:45:26.799Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["Developer tomi introduced a method that compares vulnerability news with information about the public server they run every day, and notifies them through a GitHub issue if action is needed.","GitHub Actions reads the server's operating system and program versions over SSH, and fetches Cloudflare Zero Wire vulnerability news via RSS. It does not change the server's settings or restart it.","Comparison with fixed versions is handled by code that uses official data, while Claude examines the conditions under which a vulnerability has an impact and whether the server is exposed. The author explained that when Claude was also given the version comparison, some cases remained in a state that needed confirmation.","If the machine judgment finds an unfixed vulnerability, an issue is created regardless of Claude's result. The author does not allow Claude to run commands for connecting to the server, and configured it not to follow instructions in external documents."],"discussion":"https://startupxo.com/en/1250/","markdown":"https://startupxo.com/en/1250.md","api":"https://startupxo.com/api/v1/item.json?id=1250&lang=en"},{"id":1249,"kind":"story","title":"Dropbox adds AI features to help find and organize files","url":"https://siliconangle.com/2026/11/05/dropbox-acquires-tigris-data-for-350m-releases-new-ai-productivity-features/","host":"siliconangle.com","by":"startupxo","score":0,"comments":0,"time":"2026-11-07T18:24:21.452Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["Dropbox has added AI features that let users search for an image by describing what it looks like, ask questions about files, and get summaries of recent changes to folders.","According to Dropbox, AI can change file formats, move folders, and combine multiple files into a PDF. Users can also link several actions into a single workflow.","Dropbox also unveiled DBX Agent Cloud, where AI agents can store the data they need for their tasks. It plans to add features that manage files in folders and help AI process their contents more easily.","Dropbox is acquiring Tigris Data, which provides data storage services across multiple clouds, for $350 million. Tigris claims that its service reads and writes data several times faster than competing products."],"discussion":"https://startupxo.com/en/1249/","markdown":"https://startupxo.com/en/1249.md","api":"https://startupxo.com/api/v1/item.json?id=1249&lang=en"},{"id":1248,"kind":"story","title":"Hands-on with the Gem necklace UV tracker: it flagged exposure even in the shade","url":"https://www.theverge.com/tech/1023230/the90-gem-necklace-hands-on-impressions","host":"theverge.com","by":"startupxo","score":0,"comments":0,"time":"2026-11-07T18:23:49.308Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["The Gem from The Ninety is a neck-worn sensor that measures UVA and UVB separately. Its app shows changes in exposure, sunscreen reminders and a daily sunlight score.","With a profile that includes skin type, sensitivity, clothing and sunscreen details, the app estimates things like vitamin D production and the time left before sunburn.","A reporter from The Verge got an alert that UV exposure had risen sharply while sitting in the shade of a parasol. The reporter also learned later that a medication they take increases sensitivity to heat.","The reporter could not independently verify the sensor's accuracy. On an iPhone, measurement stopped when the app was fully closed, and the estimates for vitamin D and time left before sunburn changed greatly within a few minutes."],"discussion":"https://startupxo.com/en/1248/","markdown":"https://startupxo.com/en/1248.md","api":"https://startupxo.com/api/v1/item.json?id=1248&lang=en"},{"id":1247,"kind":"story","title":"Kiro Crew Ultra Mode splits big jobs across AI chats and manages them on one screen","url":"https://zenn.dev/aws_japan/articles/kiro-crew-ultra-mode","host":"zenn.dev","by":"startupxo","score":0,"comments":0,"time":"2026-11-07T12:26:11.371Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["Ultra Mode in the open-source tool Kiro Crew splits a single goal into multiple tasks, assigns each to a separate AI chat, and brings progress and approval requests together on one board.","A conductor AI sets completion criteria for each task and writes them in a ledger (task log). Task AIs only report results and cannot change their own completion criteria or mark themselves as passed. Records remain after a restart.","It arrived as a feature preview in 0.12.0 and is off by default. By default, each task assignment requires approval, and the default number of tasks opened simultaneously per goal is 6. The author derived this figure by reading the code.","In the author's server inspection test, approval was required to assign tasks, run commands, check completion criteria, and close chats. When a task AI reports that it is finished, the board shows it as complete, but the ledger entry remains open until the conductor AI checks and closes it."],"discussion":"https://startupxo.com/en/1247/","markdown":"https://startupxo.com/en/1247.md","api":"https://startupxo.com/api/v1/item.json?id=1247&lang=en"},{"id":1246,"kind":"story","title":"Thinking Machines Lab releases Inkling 2 Small, an AI model that companies can fine-tune with their own data","url":"https://wowtale.net/2026/11/07/267620/?utm_source=rss&utm_medium=rss&utm_campaign=%25ec%2594%25bd%25ed%2582%25b9%25eb%25a8%25b8%25ec%258b%25a0%25ec%25a6%2588%25eb%259e%25a9-%25ec%259e%2589%25ed%2581%25b4%25eb%25a7%2581-2-%25ec%258a%25a4%25eb%25aa%25b0-%25ea%25b3%25b5%25ea%25b0%259c50%25ec%2596%25b5-%25eb%258b%25ac%25eb%259f%25ac-%25ed%2588%25ac%25ec%259e%2590%25ec%259c%25a0","host":"wowtale.net","by":"startupxo","score":0,"comments":0,"time":"2026-11-07T12:25:51.759Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["Thinking Machines Lab, founded by former OpenAI executive Mira Murati, has released a 2-generation AI model. It is designed so that companies can further train the model on their own data and adapt it to their own work.","For now, the weights (the learned contents of the model) are closed, and it can be used only through the company's fine-tuning service, Tinker. The company said it will release the model under the Apache 2.0 license next week, after completing an external safety review.","According to the company, the architecture activates only 17 billion parameters out of a total of 594 billion parameters, and it supports image input. Its Artificial Analysis Intelligence Index score is 31, which the company says far exceeds its previous model.","In the coding evaluations of the open model comparison group chosen by the company, it ranked number 2 in 4 of 6 evaluations and never ranked number 1. On the GPU code optimization evaluation PMPP-Hard, it fell short of GPT-6 Astra and Claude Opus 5.5."],"discussion":"https://startupxo.com/en/1246/","markdown":"https://startupxo.com/en/1246.md","api":"https://startupxo.com/api/v1/item.json?id=1246&lang=en"},{"id":1245,"kind":"story","title":"Preview a remote server screen in VS Code without port forwarding","url":"https://qiita.com/shahin0809/items/a1bd931855f392806ee6?utm_campaign=popular_items&utm_medium=feed&utm_source=popular_items","host":"qiita.com","by":"startupxo","score":0,"comments":0,"time":"2026-11-07T00:40:55.397Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["In VS Code 1.145, you can view an AI-generated web app running on another computer directly inside the editor on your own computer. No separate port forwarding setup is needed to view it.","The author connected to an Azure virtual machine over SSH and opened a sign-up screen. They said an AI agent controlled the browser to check input errors, successful registration, and duplicate registration.","You must enable workbench.browser.enableRemoteProxy in settings. Both the user's computer and the remote host must be running supported versions, and cloud sessions are not supported.","You can share a logged-in tab that you opened yourself with the agent, so take care not to share tabs displaying confidential information."],"discussion":"https://startupxo.com/en/1245/","markdown":"https://startupxo.com/en/1245.md","api":"https://startupxo.com/api/v1/item.json?id=1245&lang=en"},{"id":1244,"kind":"story","title":"Should we understand code written by AI? A developer's worry about speed and quality","url":"https://www.mobtown.jp/article/old-engineer/","host":"mobtown.jp","by":"startupxo","score":0,"comments":0,"time":"2026-11-07T00:40:37.919Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["This is a piece by a developer who has believed that people should understand and modify the output themselves, and who is now rethinking their way of working as generative AI increasingly builds and fixes programs.","The author feels that the quality difference between people who submit AI-generated code as is and people who submit it after understanding it is not clearly visible from the outside. This is the author's personal experience and opinion.","The author asks who would be rewarded: a person who produces 3 results worth 70 points in the same time, or a person who produces 1 worth 90 points. The scores are a hypothetical example to explain the choice between speed and quality.","The author sees a trend of leaving everything to AI, from deciding which features a program should have to development, testing and maintenance. The author points out that once a product is released, development may be judged a success regardless of usability."],"discussion":"https://startupxo.com/en/1244/","markdown":"https://startupxo.com/en/1244.md","api":"https://startupxo.com/api/v1/item.json?id=1244&lang=en"},{"id":1243,"kind":"story","title":"Wired Company prepares to enter Japan by automating everything from influencer matching to online store operations and settlements","url":"https://wowtale.net/2026/11/06/267591/?utm_source=rss&utm_medium=rss&utm_campaign=%25ea%25b4%2580%25ea%25b4%2591-%25ea%25b8%2580%25eb%25a1%259c%25eb%25b2%258c-%25ec%25b1%258c%25eb%25a6%25b0%25ec%25a7%2580-%25ec%2599%2580%25ec%259d%25b4%25ec%2596%25b4%25eb%2593%259c%25ec%25bb%25b4%25ed%258d%25bc%25eb%258b%2588-2%25eb%25a7%258c-%25ea%25b1%25b4-%25ea%25b3%25b5%25eb%258f%2599%25ea%25b5%25ac%25eb%25a7%25a4","host":"wowtale.net","by":"startupxo","score":0,"comments":0,"time":"2026-11-06T18:26:23.499Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["Wiredy uses AI to recommend influencers suited to a brand's products and shows projected sales volume in advance. Once a sale is arranged, the system handles everything from setting up an online store to orders, delivery, customer service and settlements.","According to the company, it was trained on more than 20,000 actual group-buy transactions accumulated over 8 years. Its projected revenue error rate is reportedly about 15%, lower than manual forecasts by managers with at least 5 years of experience (about 20%).","In 2025, the service's transaction volume was 33.6 billion won, and revenue was 29.2 billion won. It officially opened a dedicated online store in May 2026 and began charging for it in July. It has worked with 2,000 brands so far.","It has secured a database of about 130,000 Japanese influencers and is running campaigns with Korean brands. It plans to launch an AI agent for Japan in 2027, then expand to the US, Taiwan and Singapore."],"discussion":"https://startupxo.com/en/1243/","markdown":"https://startupxo.com/en/1243.md","api":"https://startupxo.com/api/v1/item.json?id=1243&lang=en"},{"id":1242,"kind":"story","title":"AI customer support design: leave document search to AI and contract decisions to DB rules","url":"https://qiita.com/yuji-arakawa/items/1c6d0ae4400ca2b4bae3?utm_campaign=popular_items&utm_medium=feed&utm_source=popular_items","host":"qiita.com","by":"startupxo","score":0,"comments":0,"time":"2026-11-06T18:26:08.007Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["This is a manga-style introductory resource written by Yuji Arakawa of Oracle Japan. A new engineer designs a support system that finds the manual matching an inquiry and checks contract information to see whether an on-site visit is possible.","Document search finds items similar in meaning to the question, but first narrows candidates by customer, model, and publication status. Contract decisions are not left to the AI; they are placed as a decision column in a DB view (a virtual table that applies conditions in advance).","In tests where the AI chooses and uses document search and contract lookup tools, problems occur such as tools not being called, execution errors, and wrong target device errors. The failure cases in the story are fictional examples to explain the design, and they are prevented by checking the call records in code.","If the work of turning documents into numeric arrays for meaning comparison is done inside the DB, the full set of documents is not sent outside. The AI that generates answers uses an external cloud service."],"discussion":"https://startupxo.com/en/1242/","markdown":"https://startupxo.com/en/1242.md","api":"https://startupxo.com/api/v1/item.json?id=1242&lang=en"},{"id":1241,"kind":"story","title":"Using a 3D printer for 1 years: the Bambu P2S Combo makes print setup easy","url":"https://www.theverge.com/tech/1025984/best-3d-printer-bambu-p2s-review","host":"theverge.com","by":"startupxo","score":0,"comments":0,"time":"2026-11-06T12:28:46.038Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["Bambu Lab's P2S Combo is a printer that feeds print material automatically, so you can make 3D objects from a smartphone or laptop. You keep the material in a sealed storage box and take it out when needed.","A The Verge editor printed for a total of 6,770 hours using 4 printers. During the 1,692 hours with the P2S, the editor said they replaced the nozzle twice and did cleaning, lubrication, and clearing material jams.","The editor wrote that the Elegoo Centauri Carbon 2 Combo, priced at 369 dollars, also printed with similar quality, but the material was exposed to moisture and there was a failure during testing whose cause was hard to find.","The list price of the P2S Combo is 749 dollars, and as of the article it is discounted to 649 dollars until December 1. The editor named the same company's P1S Combo as the closest alternative."],"discussion":"https://startupxo.com/en/1241/","markdown":"https://startupxo.com/en/1241.md","api":"https://startupxo.com/api/v1/item.json?id=1241&lang=en"},{"id":1240,"kind":"story","title":"Tesla to end Cybertruck production in 2027 and use the factory for robotaxi production","url":"https://techcrunch.com/2026/11/06/tesla-will-stop-making-the-cybertruck-in-2027/","host":"techcrunch.com","by":"startupxo","score":0,"comments":0,"time":"2026-11-06T12:28:24.250Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["Tesla will stop producing the Cybertruck electric pickup next year, 2027, and use that factory space to build more Cybercabs, driverless 2-seat taxis. CEO Musk said this on X.","At the first deliveries in late 2023, Musk expected to sell 250,000 units a year. According to the report, actual sales over the past 3 years totaled only tens of thousands of units.","At the 2019 unveiling, Tesla promised a starting price under $40,000 and a range of up to 500 miles. According to the report, it met neither: at launch the base price was about $60,000 higher, and the maximum range was cut by more than 100 miles.","Tesla said it ran Cybercabs in Austin for two months with no major accidents. It plans to expand to Dallas later this month. Earlier this year, it ended Model S and Model X production and decided to use that space to build Optimus robots."],"discussion":"https://startupxo.com/en/1240/","markdown":"https://startupxo.com/en/1240.md","api":"https://startupxo.com/api/v1/item.json?id=1240&lang=en"},{"id":1239,"kind":"story","title":"Do not reset a deceased person's smartphone with a Japanese My Number Card for 48 hours after reporting the death","url":"https://internet.watch.impress.co.jp/docs/column/tengoku/2152866.html","host":"internet.watch.impress.co.jp","by":"startupxo","score":0,"comments":0,"time":"2026-11-06T06:33:45.491Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["When Japan's My Number Card, a resident ID card, is used on a smartphone, its credential information is stored on a separate chip rather than in ordinary storage and cannot be erased by a normal reset.","According to the original article, within 48 hours after a death report is submitted, the deceased person's card and app become invalid, and the electronic certificate on the chip is automatically deleted. Japan's Digital Agency recommends waiting for this automatic deletion.","If the phone cannot be unlocked, it is reasonable to leave it as it is for 2 days. Deletion proceeds even while it is locked, as long as it is connected to the internet.","If the password is entered incorrectly repeatedly, an iPhone becomes unusable after 11 consecutive errors, and a Galaxy after 13 consecutive errors. In this case, the app cannot connect to the internet, making automatic deletion difficult."],"discussion":"https://startupxo.com/en/1239/","markdown":"https://startupxo.com/en/1239.md","api":"https://startupxo.com/api/v1/item.json?id=1239&lang=en"},{"id":1238,"kind":"story","title":"OpenAI will not sign nondisclosure agreements with public officials when negotiating data centers","url":"https://www.theverge.com/ai-artificial-intelligence/1028454/openai-data-center-nda","host":"theverge.com","by":"startupxo","score":0,"comments":0,"time":"2026-11-06T06:33:24.615Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["OpenAI said it will not enter into agreements that require public officials to keep their conversations confidential when discussing data center projects.","However, it said it may offer limited protection for security information, trade secrets, proprietary information, and competitively sensitive information.","OpenAI said it will cover the full cost of the electricity it uses and the infrastructure it needs, and will disclose expected jobs and hiring status.","It plans to put in place a community agreement for each data center it directly owns and operates, reflecting these commitments and local priorities."],"discussion":"https://startupxo.com/en/1238/","markdown":"https://startupxo.com/en/1238.md","api":"https://startupxo.com/api/v1/item.json?id=1238&lang=en"},{"id":1237,"kind":"story","title":"How to set up a software factory where AI writes code and submits review requests after you write an issue","url":"https://qiita.com/Syoitu/items/9364a0a175893bb15219?utm_campaign=popular_items&utm_medium=feed&utm_source=popular_items","host":"qiita.com","by":"startupxo","score":0,"comments":0,"time":"2026-11-05T18:27:46.044Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["A software factory is a structure where, when you write a development request, an AI agent writes the code and automatically submits a review request. The original author explains that people focus on writing requests and reviewing them.","The case the author built uses GitHub Issues as the request intake, GitHub Actions for work coordination, and Claude Code as the execution tool, with the infrastructure on AWS. This setup is a case based on a personal repository.","People make decisions at two points. One is reviewing and approving the preview of the design change review request, and the other is pressing publish after automated checks pass. According to the author, what people check is the preview, not the code.","The author made type checking, code style checking, and building mandatory conditions, so the publish button is enabled only after they pass. The original article does not present figures for work time or quality improvement."],"discussion":"https://startupxo.com/en/1237/","markdown":"https://startupxo.com/en/1237.md","api":"https://startupxo.com/api/v1/item.json?id=1237&lang=en"},{"id":1236,"kind":"story","title":"SSH servers on Alpine Linux and other musl-based systems may accept long passwords for accounts that should not allow password login","url":"https://sshhole.com/","host":"sshhole.com","by":"startupxo","score":0,"comments":0,"time":"2026-11-05T18:27:07.741Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["On Alpine Linux and OpenWrt, which use the musl standard library, the source says accounts that should not accept passwords (for example, accounts that only allow key-based login) may pass with a password between 257 and 1024 characters long.","The author attributes this to musl's crypt() function returning * in the password field when it fails on input longer than 256 bytes. This is the author's analysis and has not been verified outside the source.","The author tried about 17,600 login attempts across several combinations of glibc, musl, and whether PAM is used. They report that only musl builds without PAM wrongly accepted logins. The author performed this measurement.","The source says updating to openssh 10.6p2 or later fixes the problem, and if updating is difficult, recommends turning off password login. The accuracy of the fixed version has not been verified outside the source."],"discussion":"https://startupxo.com/en/1236/","markdown":"https://startupxo.com/en/1236.md","api":"https://startupxo.com/api/v1/item.json?id=1236&lang=en"},{"id":1235,"kind":"story","title":"Feature that moves old posts from X into a mixi2 calendar, criteria set by the development team","url":"https://zenn.dev/mixi/articles/bcf2363e83b0a2","host":"zenn.dev","by":"startupxo","score":0,"comments":0,"time":"2026-11-05T12:30:12.968Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["This feature saves old posts from the ZIP archive downloaded from X as mixi2 posts. According to the original text, saved posts can be viewed again by date in the mixi2 calendar.","According to the original text, posts are saved with the date they were posted on X, but they are excluded from follower timeline distribution, notifications, and consecutive posting day counts. The design aims to prevent old posts from filling timelines all at once.","According to the original text, the archive is not loaded entirely into server memory. It is uploaded directly to S3, then opened as a file with S3 Files so that only the needed parts are read. Posts are processed in asynchronous jobs in batches of up to 1,000.","The original text states that a mapping table between X post IDs and mixi2 post IDs is written in the same transaction to prevent duplicate saves on rerun. The policy is to keep processing the remaining posts even if an individual post fails. The part where past dates are reflected in search results is not yet complete, as the author says it is currently being addressed."],"discussion":"https://startupxo.com/en/1235/","markdown":"https://startupxo.com/en/1235.md","api":"https://startupxo.com/api/v1/item.json?id=1235&lang=en"},{"id":1234,"kind":"story","title":"Nux launches AI sales assistant node and secures $500 million growth funding","url":"https://wowtale.net/2026/11/05/267537/?utm_source=rss&utm_medium=rss&utm_campaign=ai-%25ec%2598%2581%25ec%2597%2585-%25eb%25b9%2584%25ec%2584%259c-%25eb%2585%25b8%25eb%2593%259c-%25ea%25b3%25b5%25ea%25b0%259c%25ed%2595%259c-%25eb%2588%2585%25ec%258a%25a4-5%25ec%2596%25b5-%25eb%258b%25ac%25eb%259f%25ac-%25ec%2584%25b1%25ec%259e%25a5%25ec%259e%2590%25ea%25b8%2588-%25ed%2599%2595%25eb%25b3%25b4","host":"wowtale.net","by":"startupxo","score":0,"comments":0,"time":"2026-11-05T12:29:51.124Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["Nux said it has launched an AI personal assistant node that works alongside sales reps to manage deals, and it is now accepting waitlist sign-ups.","The node keeps open deals from stalling. It remembers past call details and drafts follow-up emails, sends them once the rep approves, and alerts the rep when a decision-maker has gone quiet.","According to Nux, 96% of the emails written by the agent are sent without edits. This is a figure Nux disclosed itself and has not been independently verified.","The $500 million growth funding came from General Catalyst's Customer Value Fund. The original article explains that the structure takes no equity. The company receives upfront payment for sales and marketing costs and repays it from customer revenue generated by that spending.","Nux said more than 2,000 companies use it, that revenue grows about 3 times per year, and that net revenue retention is 135%. This too is a figure the company announced."],"discussion":"https://startupxo.com/en/1234/","markdown":"https://startupxo.com/en/1234.md","api":"https://startupxo.com/api/v1/item.json?id=1234&lang=en"},{"id":1233,"kind":"story","title":"Ocarina of Time remake adds free camera control and free jumping, and reworks hints and the Water Temple","url":"https://www.theverge.com/games/1026380/zelda-ocarina-of-time-interview-eiji-aonuma-hidemaro-fujibayashi","host":"theverge.com","by":"startupxo","score":0,"comments":0,"time":"2026-11-05T06:33:36.979Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["The Switch 2 remake is introduced as a work that keeps the flow of the original N64 game while refining graphics and controls to be closer to recent games.","The points cited as inconvenient in the original were the fixed camera and controls that did not allow jumping wherever the player wanted. In the remake, players can turn the camera themselves and jump from anywhere.","The developers explained that Navi's nagging has been reduced, and a feature was added that summarizes the adventure so far as in-game text.","According to the developers, the Water Temple, widely known as the hardest part of the original, has become easier thanks to improved camera and item menu controls. The sense that it became easier is written as the opinion of the reporter who interviewed them."],"discussion":"https://startupxo.com/en/1233/","markdown":"https://startupxo.com/en/1233.md","api":"https://startupxo.com/api/v1/item.json?id=1233&lang=en"},{"id":1232,"kind":"story","title":"Google Slides can now use Gemini to edit several slides or an entire presentation at once","url":"https://workspaceupdates.googleblog.com/?hl=enare","host":"workspaceupdates.googleblog.com","by":"startupxo","score":0,"comments":0,"time":"2026-11-05T06:33:08.395Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["Accounts with access to Gemini can enter one request in the panel next to the slides to edit several slides or the entire presentation. They can change visual styles such as fonts, backgrounds, and colors all at once, or match the style of another presentation.","According to the source, Gemini decides on its own which slides need changes and applies the edits directly to the editable slides. The edits are applied the same way as manual edits, and the source says they can be reverted with undo or version history.","According to the source, only English is supported at launch. The increased usage limit for generating multi-slide presentations is offered as a promotion until at least 1 December, 2026, after which per-user limits are expected to apply.","Eligible users are Business Standard and Plus, Enterprise Standard and Plus, consumer Google AI Pro and Ultra, and education Google AI Pro add-on users. The source says administrators have no options to configure."],"discussion":"https://startupxo.com/en/1232/","markdown":"https://startupxo.com/en/1232.md","api":"https://startupxo.com/api/v1/item.json?id=1232&lang=en"},{"id":1231,"kind":"story","title":"Craft Anything: an open-source tool that analyzes websites or games and builds new projects with similar features using the technology you request","url":"https://github.com/storytold/craft-anything","host":"github.com","by":"startupxo","score":0,"comments":0,"time":"2026-11-05T00:43:21.709Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["Craft Anything is an open-source tool that studies websites, desktop apps, and games, then builds new projects with similar features using the technology the user requests. According to the original, it is still in an early alpha stage.","Work proceeds in four stages: analysis, building, comparison, and refinement. A coding agent such as Claude Code handles building, and the results are compared with the original to show a list of parts that do not match.","The default clean room mode does not read the original's code or files and uses only what is visible on screen. Deep mode also uses analysis tools that read internal code and files.","The original README says the tool provides 320 tools through a command line interface (CLI) and as MCP tools. This figure comes from the project development team's description and has not been independently verified."],"discussion":"https://startupxo.com/en/1231/","markdown":"https://startupxo.com/en/1231.md","api":"https://startupxo.com/api/v1/item.json?id=1231&lang=en"},{"id":1230,"kind":"story","title":"X says its new web app has greatly cut the time for a profile's first post to appear","url":"https://www.theverge.com/tech/1027113/x-says-its-updated-web-app-is-way-faster","host":"theverge.com","by":"startupxo","score":0,"comments":0,"time":"2026-11-05T00:42:17.987Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["X says its new web app is much faster than the old one. The new app is already applied to pages served to logged-out users and to pages that search engines crawl, and logged-in users are being moved over country by country in sequence.","According to a blog post by X senior engineer Yagiz Nizipli, when opening an X.com profile on a slow connection, the old app took more than 22 seconds before the first post appeared, while the new app takes 2.7 seconds. This is the company's claim, and the measurement conditions are not described in detail in the original.","The new structure merges three servers that had been run separately for users, search engine crawlers, and link previews into one. Security researcher Jane Manchun Wong, after accessing the new app, described it as fast and polished, but this is a personal impression, not a measured value."],"discussion":"https://startupxo.com/en/1230/","markdown":"https://startupxo.com/en/1230.md","api":"https://startupxo.com/api/v1/item.json?id=1230&lang=en"},{"id":1229,"kind":"story","title":"After cloning my voice from one short recording and comparing it with my own, my usual speaking style sounded fairly similar","url":"https://qiita.com/nakata-sapeet/items/f883c3cf009785c51bcc?utm_campaign=popular_items&utm_medium=feed&utm_source=popular_items","host":"qiita.com","by":"startupxo","score":0,"comments":0,"time":"2026-11-04T18:26:52.298Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["The author cloned a voice from only one 15-second recording on a Mac and said it was made locally with the VoiceStudio app, without a cloud service. In a quiz, they played their own voice and the cloned voice side by side and asked listeners to guess which one was the real voice.","Based on the usual speaking style, the voice pitch the author measured was 108Hz for their own voice and 99 for the clone, with 105Hz. Speaking speed in OmniVoice was almost the same as their own, while VoxCPM2 was about 10% faster. This was a small test with 1 speakers and 3 generations per condition.","When emotion instructions were given, their own voice rose in pitch to 138Hz for 「恐る恐る」 (fearfully) and to 151Hz for 「焦って」 (in a panic), but the cloned voice stayed around 93Hz. The speed of the clone made with English instructions, 8.40, was similar to their own 焦って at 8.54, and with the 「怖がりながら恐る恐る」 instruction it actually became faster."],"discussion":"https://startupxo.com/en/1229/","markdown":"https://startupxo.com/en/1229.md","api":"https://startupxo.com/api/v1/item.json?id=1229&lang=en"},{"id":1228,"kind":"story","title":"Starting from January 2028, Waymo robotaxis in Austin and Atlanta will be booked through Waymo's own app instead of Uber's","url":"https://techcrunch.com/2026/11/04/waymo-confirms-it-will-split-with-uber-in-january-2028/","host":"techcrunch.com","by":"startupxo","score":0,"comments":0,"time":"2026-11-04T18:26:03.194Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["Waymo said it will end the practice of calling its driverless taxis (robotaxis) through the Uber app in Austin and Atlanta from January 2028, and will offer rides through its own smartphone app, as it does in other cities.","Waymo co-CEO Tekedra Mawakana said in a Bloomberg interview that the company chose Uber because of scale, but that riders found it hard to tell whether the car they called was a robotaxi or a human-driven car, so the experience was inconsistent.","Her remarks confirm what the Financial Times reported earlier this year. The companies' contract was originally set to expire in May 2028. Uber said it was Waymo's idea not to add a direct robotaxi booking option to the Uber app."],"discussion":"https://startupxo.com/en/1228/","markdown":"https://startupxo.com/en/1228.md","api":"https://startupxo.com/api/v1/item.json?id=1228&lang=en"},{"id":1227,"kind":"story","title":"Google's AI search is not yet earning as much as search ads","url":"https://www.theverge.com/ai-artificial-intelligence/1024862/google-search-ai-overviews-alphabet-earnings-q3","host":"theverge.com","by":"startupxo","score":0,"comments":0,"time":"2026-11-04T12:28:46.850Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["In Alphabet's Q3 results, Google search revenue growth was 12% year over year, lower than about 17% in the previous quarter. Search query volume grew at the fastest pace in 5 years, but ad click growth also fell from 13% to 8%, according to the original article.","Chief business officer Philipp Schindler said exchange rate effects and spending changes by advertisers in some industries were the biggest causes of the slowdown. He also acknowledged that AI Mode (a search method that continues questions in a conversational way) is growing quickly, but a smaller share of it leads to ad revenue than traditional search.","Google argues that most AI Mode searches are additional questions that do not replace traditional search. This is Google's claim, and the original article does not give figures showing how much AI Mode replaces the more profitable traditional search.","In Q3, the existing advertiser group of ad agency Tinuiti increased its Google search ad spending by 14% year over year while keeping its ad return targets. However, Google does not let advertisers target ads in the AI answer area separately or view their performance separately, so the scale of AI ad traffic has not been confirmed."],"discussion":"https://startupxo.com/en/1227/","markdown":"https://startupxo.com/en/1227.md","api":"https://startupxo.com/api/v1/item.json?id=1227&lang=en"},{"id":1226,"kind":"story","title":"Anthropic Research on an Open-Source System That Cryptographically Proves Which AI Model Actually Answered","url":"https://techcrunch.com/2026/11/03/anthropic-is-working-on-a-system-to-prove-ai-models-identity/","host":"techcrunch.com","by":"startupxo","score":0,"comments":0,"time":"2026-11-04T00:41:41.438Z","origin":"curated","language":{"source":"ko","reading":"en","translated":true},"summary":["According to the source, Anthropic is developing FlashProve, an open-source system that checks which AI model a specific output came from. The goal is to use cryptographic proof to confirm whether the answer a user received came from the model they believe they are using.","The source explains that it could be used to stop AI companies from secretly switching to cheaper models to cut costs, or from using more expensive models to inflate performance figures. This is a possible use suggested by the researchers and is not a confirmed real case.","The method is to publish a fingerprint for each model when it is released, without revealing the weights. Outputs are then checked cryptographically against this fingerprint. The source says the system was designed to be lightweight so it can be verified even on personal devices.","Part of the technical specification and working code is published on GitHub under an open license. The source says the research is at an intermediate stage and that there are no plans to deploy it at Anthropic or other labs."],"discussion":"https://startupxo.com/en/1226/","markdown":"https://startupxo.com/en/1226.md","api":"https://startupxo.com/api/v1/item.json?id=1226&lang=en"}]}