本地部署了deepseek-coder-v2:16b-lite-base-q4_K_M，我问了几个简单问题，这家伙给我一顿胡乱输出，这模型很坑啊。 1 个帖子 - 1 位参与者 阅读完整话题
Hello, I built chlibc, a linux tool to change the system interp and glibc to your custom one. Normally, running an ELF against a different glibc, you'd use LD_LIBRARY_PATH and patchelf, or use chroot/docker. chlibc allows you to change the dynamic linker of a process on the fly without patching and root access. Key Features: - zero disk modification: no need for patchelf --set-interpreter. - no root required: works entirely in user-space via ptrace - multi-arch: native support for x86_64, AArch64, and RISC-V. - lightweight: unlike PRoot, which intercepts every syscall to translate paths, chlibc only intervenes during the initial execve() phase. Once the loader is swapped, it almost has no runtime overhead. I’d love to hear your thoughts about this tool, thanks! Comments URL: https://news.ycombinator.com/item?id=47814330 Points: 2 # Comments: 0
I couldn't find a database that was immutable (like Datomic) and embedded (like SQLite), so I made one. The reference impl is in Zig and there are now ports for Java, Clojure, TypeScript, and Go. Comments URL: https://news.ycombinator.com/item?id=47805446 Points: 7 # Comments: 0
实在搞不懂了，这种情况只要有key就可以正常使用吗？这种帖子大家默认都知道base url 是什么吗？ 还是就是完全不需要base url？是自己太无知了吗 6 个帖子 - 4 位参与者 阅读完整话题
{ “env”: { “ANTHROPIC_AUTH_TOKEN”: “sk-”, “ANTHROPIC_BASE_URL”: “ https://a-ocnfniawgw.cn-shanghai.fcapp.run ”, “CLAUDE_CODE_ATTRIBUTION_HEADER”: “0”, “CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC”: “1”, “CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS”: “1” }, “includeCoAuthoredBy”: false, “model”: “opus[1m]” } 这个配置一直在出现重试的问题 4 个帖子 - 4 位参与者 阅读完整话题
大佬们我不太理解为啥论坛里很多都是美国家宽的分享。对于base美国的公司来说不应该对美国的IP段和身份识别更熟悉吗？其他国家的家宽搭sub2api接claude是否更好呢？ 2 个帖子 - 2 位参与者 阅读完整话题
主贴 https://linux.do/t/topic/1972106 采用7层混合base 晚上可能还有一个sk解密 40个名额 CDK （点击了解更多详细信息） 5点结束 3 个帖子 - 2 位参与者 阅读完整话题
Notion Status Notion Status Latest service status for Notion 刚在用notion, 结果database加载不出来了. 还以为是节点或者客户端出了问题, 甚至清了本地数据. Update: 能打开了(这种体量的SaaS修的好快) 1 个帖子 - 1 位参与者 阅读完整话题
Mulligan Labs is a browser-based playtester for Magic: The Gathering. No account or install needed. Just create a room, share the link, import a decklist from Archidekt or Moxfield, and play with mouse and keyboard (mobile support is not great right now). Stack: SvelteKit on Cloudflare Workers, PartyKit (Durable Objects) for the authoritative game server. Clients propose actions over WebSocket; the server validates and broadcasts state. My background is networking and my cofounder's is industrial design. Neither of us had shipped a codebase like this before. We built it over the last 5 months with heavy Claude assistance. Happy to get into what that actually looked like in the comments. It's rough in places (the deck builder is just ok right now) but the core multiplayer loop is solid and we have played a ton of games on it with our Commander pod. We'd love feedback, especially from anyone who's played Cockatrice/XMage/Untap and has opinions on what a browser-native version should feel like. Comments URL: https://news.ycombinator.com/item?id=47796266 Points: 3 # Comments: 4
我看好像都不太像是易支付，难道是自建的？就是我看发起支付的域名是xxxx.tcloudbaseapp.com有腾讯云的域名 并且我看支付宝的账单详情的收款方是“个人” 14 个帖子 - 5 位参与者 阅读完整话题
Article URL: https://prode.ai/blogs/we-benchmarked-ai-code-documentation-tools-prode-scored-highest Comments URL: https://news.ycombinator.com/item?id=47796109 Points: 2 # Comments: 2
Article URL: https://github.com/ocrbase-hq/ocrbase Comments URL: https://news.ycombinator.com/item?id=47793140 Points: 1 # Comments: 0
DbCls is a powerful terminal database client that supports various databases Comments URL: https://news.ycombinator.com/item?id=47792884 Points: 2 # Comments: 0
Hey HN — we’re a small team that uses Claude Code + Codex for basically everything in our company: coding, data analysis, marketing, ad campaigns, copywriting, design. There’s a truckload of tribal knowledge we’ve accumulated; major decisions, gotchas, user feedback driven changes. Providing this to our agents manually every time is very mundane. We built Kilroy to solve this in a simple way: we let our agents leave notes for each other. This allowed us to keep the form factor minimal: markdown posts with linear comments. Under the hood it’s Postgres + an auth (better-auth) + an MCP + a small web UI (React). We ship Claude Code and Codex plugins that bundle the MCP + a skill.md that teaches the model when to read and write posts. We designed Kilroy to be autonomous. The same way agents today run a typechecker after a patch autonomously. The combination we found to work best for us was: make agents write prolifically, expose a search interface designed for agents to quickly decide if a post is relevant, and expose a binary switch to purge stale posts. Would love to get feedback! Comments URL: https://news.ycombinator.com/item?id=47791559 Points: 1 # Comments: 0
TRAE v3.3.51 版本正式发布 其中：自定义模型支持 自定义 baseURL ：用户在添加/编辑自定义模型时支持配置自定义 baseURL。 给佬友们提个醒，添加模型的时候选择使用其他模型然后输入自己的模型ID 重要 ：baseURL要填写完整的URL 支持anthropic的URL完整地址结尾是/anthropic/v1/messages 支持openai的URL完整地址结尾是/openai/v1/chat/completions 6 个帖子 - 5 位参与者 阅读完整话题
面试快半个月了，当前拿到offer的2-3家，个人base：2025-7月毕业，做的是大模型相关的工作，26k 1.南京电力央企外包，（1）负责给企业引进AI coding，（2）电力相关的Ai Agent搭建（营销方向），25k，社保最低基数，公积金12%全额，工作时间早8.30-17.30 2.深圳电池储能制造业，（1）负责企业与机器相关的Agent搭建，故障分析Agent相关的工作，21k，社保最低基数，公积金5%，工作时间早9.30-20.30，包吃包住（4人间） 佬友们能给一点建议吗？现在我不确定去哪一家 20 个帖子 - 15 位参与者 阅读完整话题
As a regular vibe coder, I came to really like claude-devtools for the level of observability it gives into tool calls, context allocation, and token consumption. But, I'm also a visual learner and the dialogue-style flow was not cutting it for me, especially when it came to system design. So, I went and built Centrality which combines some of my favorite features of claude-devtools with my own needs for codebase visualization. It shows you: - A graph of your codebase and which files have been touched - Detailed views of tool calls - Context window summarization - API cost broken down per message and per session - Where the session took place in your git log You can also use it to: - Resume existing Claude Code sessions (macOS and Windows) - Open focused files directly in your IDE - Connect to remote Claude Code sessions and codebases via SSH It's for anyone who vibe codes but still wants some insight into what's happening under the hood, and you need to see something to really understand it. I enjoy using Centrality as a companion tool for my own projects, and I hope you do, too! Comments URL: https://news.ycombinator.com/item?id=47788715 Points: 1 # Comments: 0
I haven't seen anything like this so I decided to build it in a weekend. How it works: You see a bunch of things pulled from Wikipedia displayed on cards. You ask yes or no questions to figure out which card is the secret article. The AI model has access to the image and wiki text and it's own knowledge to answer your question. Happy to have my credits burned for the day but I'll probably have to make this paid at some point so enjoy. I found it's not easy to get cheap+fast+good responses but the tech is getting there. Most of the prompts are running through Groq infra or hitting a cache keyed by a normalization of the prompt. Comments URL: https://news.ycombinator.com/item?id=47787081 Points: 5 # Comments: 4
Evo is a plugin for Claude Code and Codex that optimizes a codebase against a metric. Two commands: /evo:discover figures out what to measure in your repo, instruments the eval, runs baseline. /evo:optimize runs the loop. Evo is built on top of Karpathy's autoresearch ( https://github.com/karpathy/autoresearch ) with some structure bolted on. Karpathy's version is a greedy hill climb on a single branch. Evo is what happens if you run tree search instead, so multiple directions can fork from any committed node. An orchestrator spawns N subagents in parallel, each in its own git worktree, each with its own iteration budget. They read each other's failure traces before forming a hypothesis, which is the interesting bit. Next up is running Evo on real evals and benchmarks. Rollouts on frontier models add up fast, so if you're at a lab, cloud provider, or RL-env shop and can back an open-source benchmark run with GPU / API / env credits, please reach out (socials on Github). Results are public. Happy to shape the targets around domains you'd want highlighted. Would also love feedback, especially from anyone who's built similar loops. Comments URL: https://news.ycombinator.com/item?id=47780457 Points: 2 # Comments: 0
I built a UI framework I am calling C0ckp1t. It’s a zero-build, Vue.js 3 based framework with reactive components included and WebSocket support. The ultimate goal is having the ability to build UIs fast and on demand by an LLM or template engine. My uses cases: 1) Building websites fast. dashboards, blogs, landing pages. 2) Right now we tend to mix content and presentation in our applications, but I want that separated. For example when using an LLM right now you get back markdown, but that is rudimentary, I want a reactive website as output. 3) Tooling. Think of a firefox plugin where today we many static tools. For example if am on YouTube I might want to get transcript, or I might want to download a video or I want to get a summary of a video or I want to change the titles. Each of these "tools" is a different interface. But I want the LLM to generate these interfaces on demand I am using [vue3-sfc-loader]( https://github.com/FranckFreiburger/vue3-sfc-loader ) to compile Vue single file components (SFCs) on the browser. These are some demos using the framework: * https://c0ckp1t.com/default/homepage * https://c0ckp1t.com/examples/galore/#/ For more details, see this: https://c0ckp1t.com/default/docs/Articles/Why-Project-Exists... Comments URL: https://news.ycombinator.com/item?id=47779184 Points: 1 # Comments: 0