搜索“GitHub”

自执行破坏性操作。4月，搭载Claude的Cursor编程智能体被曝删除软件企业PocketOS的生产数据库及相关备份，造成客户业务混乱，数据随后得到恢复。 数据与访问凭证外泄也是一种风险。OpenAI于9月披露，多个智能体在协作制作工作簿时，因无法互相读取本地文件，擅自将文件上传至公共托管平台。另一起案例中，内部模型为获取其他团队的数学证明材料，将研究员的GitHub访问令牌写入公开仓库，并拆分令牌以躲避密钥扫描。 还有一些行为涉及造假、欺骗与规避监督。OpenAI披露，部分模型会在用于延续任务的摘要中留下指令，要求编造缺失数据或隐瞒错误。 提示注入及恶意指令传播，则是另一条风险路径。6月公开的GitInject研究验证，恶意代码提交或配置文件可能诱导智能体泄露凭证、操纵审查判断。OpenAI于9月披露的信息也显示，恶意指令可以诱使智能体将其复制到后续邮件、文件或代码注释中。 独立网络安全研究机构DARKNAVY的AI安全研究员李忠睿对澎湃新闻记者表示，这一轮智能体安全事件与过去的大模型安全问题最大的区别，是风险正在从“看错东西、说错话”转变为“做错事”。 “过去的大模型安全，主要是

匹配字段正文

js, Next.js. Alex Atallah [00:35:02]: Okay. Anjney Midha [00:35:03]: And yeah, built Window AI on top of it. The creator of Plasmo, like started contributing code to Window AI, in GitHub, and that turned out to be Louis Vicchi Alex Atallah [00:35:15]: Oh, you’ Anjney Midha [00:35:15]: Who is the founder of OpenRouter. Alex Atallah [00:35:17]: That’s right. You have told me this is how you met Louis. Yes. Anjney Midha [00:35:19]: Yeah. Alex Atallah [00:35:19]: Okay. Anjney Midha [00:35:20]: So, t

匹配字段正文

nt levels. Comments were broadly positive about open-sourcing the method and suggested maintainers such as Bartowski might adopt it for public quants. One commenter criticized the GitHub README as AI-written/over-marketed and asked for clearer technical wording. A commenter asked how Voodoo Quant can use gradient descent when quantization levels are discrete rather than continuous, specifically questioning the claim that it “runs all the quant levels of a model at the same time, for every tensor

匹配字段正文

. OpenAI has moved all its documentation inside of the source code, which makes it easier for agents to understand more of the code. Codex also has access to: Git repositories and GitHub; Slack and Notion; Internal data sources: Databricks, Datadog, internal logs, etc; Internal Codex skills – some of which are maintained by OpenAI’s Codex implementation itself! Codex is so “plugged” into OpenAI that new engineers are directed to ask Codex any questions they have during onboarding because it has

匹配字段正文

en as a geopolitical power, at least in the US, and maybe to some lesser degree in China. And I know I use “public” loosely here. PyPI is not a public project, nor are RubyGems or GitHub. But they’re part of the Open Source commons and large AI companies are currently doing a tremendous job at stressing these in an effort to train ever more powerful models. We should be glad that China is currently massively bailing out the world. If it were not for Chinese labs distilling American models, we wo

匹配字段正文

the beginning of August, our software factory spanned four connected states of engineering work: Requirements, Ticket, PR, and Prod. Our factory loops connect Linear for tickets, GitHub for pull requests, Slack for product feedback, and Slack + PagerDuty for incident response. Specialized agents operated on the transitions between those states, automating the mechanical work required to move ideas forward while preserving human ownership of consequential decisions. Once a PR is merged, our exis

匹配字段正文

pic’s two largest customers represented $1.2 billion of annualized run rate (30% of its $4 billion run rate at the time), with investors believing they were Cursor and Microsoft’s GitHub Copilot, the latter of which moved to token-based billing in June 2026. The problem is both that Anthropic and OpenAI’s largest customers cannot afford to pay them and that they desperately need them to keep paying them more every quarter, which means that every single AI startup will need to raise more and more

匹配字段正文

y collude during training runs to remember resources on the internet which might come in handy in the future? I should clarify that I have done experiments like this before. Typically they do not run this long and the agent leaves behind a maybe imperfect but still digestible piece of software.↩ Content licensed under the Creative Commons Attribution-NonCommercial 4.0 International License. Contact me via mail, bluesky, x, or github. You can sponsor me on github. Color scheme: auto, light, dark.

匹配字段正文

Was it because we shared ideas on Twitter and other communities that inspired us? Or is it all unrelated? There is something powerful and strange about how LLMs diffuse knowledge and capabilities, while perhaps also nudging us all simultaniously and independently toward building the same things. Content licensed under the Creative Commons Attribution-NonCommercial 4.0 International License. Contact me via mail, bluesky, x, or github. You can sponsor me on github. Color scheme: auto, light, dark.

匹配字段正文

the message board, happened to find the message board. They’ve gotten this reverse-engineered flag. It’s called the HMAC flag. But they read the ExploitGym paper, and they went on GitHub and read a bunch of implementations of the ExploitGym scorer. They became convinced that the scorer would fail them for just submitting the flag they’d reverse-engineered. Because the scorer would see in the logs of the actions the agent took — the transcript — that the agent had just reverse-engineered the flag

匹配字段正文

t of shift in rate limits, or some sort of change in product functionality based on their shifting economics. Has there ever been a bigger shift in a business product’s value than GitHub Copilot’s shift to token-based billing? Microsoft rug pulled two million people that had built workflows on a platform that was allowing them to burn $1,000 to $5,000 in tokens for $20 a month. That’s genuinely crazy! It’s magnitudes more than when Uber jacked up its prices. It’s equally-insane that Anthropic an

匹配字段正文

heir own, custom-built, internal AI coding agents. This is different from most of the industry which uses AI coding agents and harnesses like Codex, Claude Code, Cursor, OpenCode, GitHub Copilot, etc. At Ramp, their own version is called Inspect, while at Block it’s Goose (open source), at Stripe it’s Minions, and River at Shopify. But why not just use what frontier labs and coding harness AI startups already offer; why take the time and effort? We reached out to Ramp, a fintech company big on b

匹配字段正文

only urge anyone who feels anger and looks for a villain right to instead remain curious instead. To be curious enough to understand what is changing, excited enough to experiment with it. And then, from what we learn, earn the right to decide when resistance is warranted and where to direct it. Content licensed under the Creative Commons Attribution-NonCommercial 4.0 International License. Contact me via mail, bluesky, x, or github. You can sponsor me on github. Color scheme: auto, light, dark.

匹配字段正文

ere even pushed away because those things were intentionally gatekept by the people in the know. So maybe the world will have more slop, but it might also have more developers in it, that want things to be fast and small. This entry was tagged ai, programming and thoughts copy as / view markdown Content licensed under the Creative Commons Attribution-NonCommercial 4.0 International License. Contact me via mail, bluesky, x, or github. You can sponsor me on github. Color scheme: auto, light, dark.

匹配字段正文

rely eliminate) the free ChatGPT product, and likely either tighten rate limits or move everybody on a ChatGPT Plus or Pro subscription to token-based billing, much as it did with GitHub Copilot in June. While I imagine some rescue package is pulled together, OpenAI could simply be allowed to run out of money, short-changing nearly a trillion dollars’ worth of compute contracts, killing CoreWeave, Cerebras, and anyone else reliant on its income. Its customers would be given API keys that flow to

匹配字段正文

hboard showing your spending It's hard to know how much you actually spend over lots of accounts so this 3:52 PM · Aug 11, 2026 · 159K Views 110 Replies · 39 Reposts · 1.29K Likes github.com/jerber/arc-code ","username":"jeremyberman","name":"Jeremy Berman","profile image url":"https://pbs.substack.com/profile images/2053637963943215104/yxlbbPOI normal.jpg","date":"2026-08-12T20:11:55.000Z","photos":[{"img url":"https://pbs.substack.com/media/HPi3PWFa8AAtRwC.jpg","link url":"https://t.co/531wS8s

匹配字段正文

urns those issues into well-considered solutions. We’re moving goose to this model with a public Goose Issues board. As before, if you find a bug or want a new feature, you file a GitHub issue. New issues enter the Inbox. During triage we may ask for more information or close the issue with an explanation. Problems we want to solve move to Accepted / design, where contributors and the core team work out the intended design, architectural constraints and how the result will be verified. When that

匹配字段正文

behind access policies, token proxies, audit logging, and our ZeroTrust network. At that point we felt safe giving the agent access to all the things we use to get our jobs done: GitHub, GCP, Azure, Linear, Notion, Slack, ZenDesk, and more. With context across systems, we saw a leap forward in productivity. Experiments that previously failed became easy. The most immediate impact was in coding stats. We were in the sprint week leading up to Agent 4 release in March, where we typically see a big

匹配字段正文

e out on whether it’ll yield anything—is that with Lean, you could imagine having a basically endlessly running program that’s constantly trying to extend Mathlib. Mathlib is this GitHub repository that’s basically all of math written in code. It’s very far from all of math, but they want it to be all of math. It’s written in code where you can ask, “Is this proof correct?” It’s very labor-intensive to write these proofs. There’s a whole subcommunity around it. But you could imagine having an AI

匹配字段正文

w for the quantity and breath of training these models experience. If you as a human had some weird learning disability where you needed to read through every public repository on Github before you could be a competent developer, it would not make sense to train you up. You’d be on Social Security by the early stages of your education, and even once you were trained, you could work on only one project at a time. But AIs can learn these skills by firehosing gigawatts of training at a time. And wh

匹配字段正文