AllyshipContent Radar

Deep dives 738

Paper và bài viết kỹ thuật

Crawl live-1795068900000 · 13:22 19 thg 11, 2026 · Xuất bản 13:30 19 thg 11, 2026

Tác tử

v0.22.3

<h2>What's Changed</h2> <ul> <li>fix(core): align conditional approvals with validated tool arguments by <a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/users/seratch/hovercard" data-octo-click="hovercard-link-click" data-octo-dimensions="link_type:self" href="https://github.com/seratch">@seratch</a> in <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="5484896104" data-permission-text="Title is private" data-url="https://github.com/openai/openai-agents-python/issues/5066" data-hovercard-type="pull_request" data-hovercard-url="/openai/openai-agents-python/pull/5066/hovercard" href="https://github.com/openai/openai-agents-python/pull/5066">#5066</a></li> <li>fix(core): deliver tool-not-found output on server-managed resume by <a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/users/hyeonsang010716/hovercard" data-octo-click="hovercard-link-click" data-octo-dimensions="link_type:self" href="https://github.com/hyeonsang010716">@hyeonsang010716</a> in <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="5412414529" data-permission-text="Title is private" data-url="https://github.com/openai/openai-agents-python/issues/4947" data-hovercard-type="pull_request" data-hovercard-url="/openai/openai-agents-python/pull/4947/hovercard" href="https://github.com/openai/openai-agents-python/pull/4947">#4947</a></li> <li>fix(sandbox): keep command paths POSIX on a Windows host by <a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/users/hyeonsang010716/hovercard" data-octo-click="hovercard-link-click" data-octo-dimensions="link_type:self" href="https://github.com/hyeonsang010716">@hyeonsang010716</a> in <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="5425677074" data-permission-text="Title is private" data-url="https://github.com/openai/openai-agents-python/issues/4958" data-hovercard-type="pull_request" data-hovercard-url="/openai/openai-agents-python/pull/4958/hovercard" href="https://github.com/openai/openai-agents-python/pull/4958">#4958</a></li> <li>fix(sessions): handle concurrent async SQLite startup by <a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/users/00200200/hovercard" data-octo-click="hovercard-link-click" data-octo-dimensions="link_type:self" href="https://github.com/00200200">@00200200</a> in <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="5439630528" data-permission-text="Title is private" data-url="https://github.com/openai/openai-agents-python/issues/4987" data-hovercard-type="pull_request" data-hovercard-url="/openai/openai-agents-python/pull/4987/hovercard" href="https://github.com/openai/openai-agents-python/pull/4987">#4987</a></li> <li>fix(sessions): do not forward session limit as Conversations page size by <a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/users/tiagovilasboas/hovercard" data-octo-click="hovercard-link-click" data-octo-dimensions="link_type:self" href="https://github.com/tiagovilasboas">@tiagovilasboas</a> in <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="5429411248" data-permission-text="Title is private" data-url="https://github.com/openai/openai-agents-python/issues/4961" data-hovercard-type="pull_request" data-hovercard-url="/openai/openai-agents-python/pull/4961/hovercard" href="https://github.com/openai/openai-agents-python/pull/4961">#4961</a></li> <li>fix(tracing): do not cache a missing OPENAI_API_KEY by <a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/users/hyeonsang010716/hovercard" data-octo-click="hovercard-link-click" data-octo-dimensions="link_type:self" href="https://github.com/hyeonsang010716">@hyeonsang010716</a> in <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="5449456992" data-permission-text="Title is private" data-url="https://github.com/openai/openai-agents-python/issues/5017" data-hovercard-type="pull_request" data-hovercard-url="/openai/openai-agents-python/pull/5017/hovercard" href="https://github.com/openai/openai-agents-python/pull/5017">#5017</a></li> <li>fix(extensions): include shell and apply_patch calls in AdvancedSQLiteSession's tool usage by <a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/users/rioyu123/hovercard" data-octo-click="hovercard-link-click" data-octo-dimensions="link_type:self" href="https://github.com/rioyu123">@rioyu123</a> in <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="5433079345" data-permission-text="Title is private" data-url="https://github.com/openai/openai-agents-python/issues/4982" data-hovercard-type="pull_request" data-hovercard-url="/openai/openai-agents-python/pull/4982/hovercard" href="https://github.com/openai/openai-agents-python/pull/4982">#4982</a></li> <li>fix(extensions): reap subprocesses on early stream close by <a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/users/Hughhhhcoder/hovercard" data-octo-click="hovercard-link-click" data-octo-dimensions="link_type:self" href="https://github.com/Hughhhhcoder">@Hughhhhcoder</a> in <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="5310233997" data-permission-text="Title is private" data-url="https://github.com/openai/openai-agents-python/issues/4804" data-hovercard-type="pull_request" data-hovercard-url="/openai/openai-agents-python/pull/4804/hovercard" href="https://github.com/openai/openai-agents-python/pull/4804">#4804</a></li> <li>fix(extensions): preserve exact IDs in AdvancedSQLiteSession.delete_branch by <a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/users/jiangLLM/hovercard" data-octo-click="hovercard-link-click" data-octo-dimensions="link_type:self" href="h

Nguyên văn nguồn
Chi tiết nguồn
  • v0.22.3 openai/openai-agents-python releases · 18 thg 9, 2026
Kỹ thuật

How Claude Watermarks AI-Generated Text

Câu hỏi And now there’s going to be this watermarking, and there’s this, I guess, fear of watermarking making text worse, or what’s actually the benefit of this watermarking?

I recently posted a Substack note about Claude’s new watermarking process and implementation. Since it’s such a popular topic and sparked such a lively discussion, I thought it might be interesting to go into a bit more detail when explaining how it works. Instead of the usual text article, I recorded a little lecture on the topic (to change it up a bit from my usual articles). So, below is the video along with a transcript. Originally, I planned to make 10 slides and record a short 10-min video. However, while putting it together, I added some crucial details here and there, resulting in >50 slides and a 48 min recording. I hope that this now explains it well, though! Happy watching! I also have a YouTube version if you prefer using the YouTube player And here is a link to the slides Subscribe now Video Transcript Note: The transcript below is slightly edited and cleaned up for readability but preserves the overall order and flow of the video lecture above. 0:00 How Claude’s Text Watermarking Works Slide 2 of 52, time stamp 0:00 Hi everyone. So, a few days ago, Anthropic announced that they will watermark the text outputs of their Claude models. I then did a social media post briefly explaining how that works. And yeah, this was quite the popular post. So not the watermarking itself was popular, but I guess the explanation or the mechanism behind it. Then, it might be worthwhile expanding this a bit to explain it in more detail, because this post only had one figure, and there were a lot of questions and discussions. So, I thought, well, let’s make a few more figures. I actually originally planned to do like 10 slides and walk you through it. It ended up being 50 slides, but I hope this really explains how this watermarking technique works well, how watermarking itself can fail or be removed, and so forth. So I think it might be an interesting topic because a lot of people use LLMs these days and also consume a lot of text on the Internet that might be generated by LLMs. And now there’s going to be this watermarking, and there’s this, I guess, fear of watermarking making text worse, or what’s actually the benefit of this watermarking? And so what does it mean? And I think if we understand a bit better what watermarking is, that goes a long way, and then we can make up our own minds about whether that’s a good thing or not, and so forth, like the pros and cons. So, my goal here is really to explain how the underlying mechanism works and how they are going to implement this type of watermarking, text watermarking. Slide 2 of 52, time stamp 1:41 It’s also a great example to illustrate why understanding things from scratch is actually quite useful. This watermarking technique is also a nice way to explain how conventional models or LLMs in general work under the hood. So yeah, you may know I like doing things from scratch. Like, I have my books: Build a Large Language Model From Scratch, Build a Reasoning Model From Scratch. I have some articles labeled from scratch. So, for me, “ from scratch often includes coding. So this one will not be coding-related, but coding from scratch is actually a very, very useful technique because it really helps you understand how something is implemented. And then from that we can derive our understanding, figures, concepts, because if we don’t really implement things, if there’s no code, it’s really sometimes ambiguous. And of course, you know, as I realized, not everyone is coding from scratch anymore. Like back in the day, coding something from scratch was all we had. I mean, there were only humans coding. Nowadays, coding can be done by LLMs. However, that doesn’t mean reading code is no longer useful, because it carries a lot of information. So in this case here with this watermarking, spending some time coding an LLM from scratch really makes you realize how this sampling inside is implemented. We still have some relevant code snippets. And then that really, in turn, helps us understand, oh, the watermarking is applied at this position, and this has so-and-so consequences and so forth. So I think even though people may not be coding from scratch, at least not all the time anymore, it is still useful being able to, let’s say, build something from scratch for educational purposes to understand something deeply and then also for research purposes to manipulate this in a transparent way that is not hidden away in tons of layers of abstraction. But that aside, I think it’s just a coincidental nice relationship here because, for this slide deck, I actually used a lot of figures from my from-scratch coding materials. Slide 3 of 52, time stamp 3:58 So a few days ago (this is August 14), there was this article, How Claude’s Text Watermark Works, and there was this article here; it’s just like a screen recording, so it can have everything in the slides, but there’s plenty of detail. They updated it actually a couple of times, so originally when I read this, it was a way shorter. Still, it is very, I guess, conceptual; there’s like this overview, and there’s, I mean, there’s not a single figure in there. And so it’s kind of still hard to understand what they’re trying to do. So they explain a lot about why they’re going to do it, but they don’t explain how. They’re linking to one paper somewhere there, which is very technical also. So I do think it makes sense maybe to take a step back and start at the beginning to kind of understand what they’re trying to implement here with this watermarking technique. And so the motivation, by the way, of watermarking is for them to identify if someone posts some text that they can say, oh, this text was generated by our Claude Opus 4.8 model, for example, so that they have a way to tell, OK, this text is AI-generated because it carries this watermark. And this watermark is invisible to users, so only they can decode it and find out whether the text has their watermark. Why can only they do it? We will get to that later in thi

  • However, while putting it together, I added some crucial details here and there, resulting in >50 slides and a 48 min recording.
Nguyên văn nguồn
Chi tiết nguồn
Kỹ thuật

Introducing the Life Sciences Verification Program

The Life Sciences Verification Program (LSVP) gives life science professionals access to Claude Mythos, Opus, and Sonnet models with a refined set of safeguards more permissive for biology-related work.

Nguyên văn nguồn
Chi tiết nguồn
Kỹ thuật

Mistral and Mozilla are bringing open, private and multilingual AI to your web browser

Open, private and multilingual AI is coming to your web browser. Mistral and Mozilla team up to put powerful, trustworthy AI where you already browse.

Nguyên văn nguồn
Chi tiết nguồn
Kỹ thuật

Introducing Claude Mythos and Fable 5.1

They’re the world’s most advanced models for coding and knowledge work—and their research capabilities offer an early glimpse of how AI models will contribute to scientific progress.

Nguyên văn nguồn
Chi tiết nguồn
Huấn luyện

Together AI expands fine-tuning service with more models, live metrics, and finer controls

Together Fine-Tuning adds the latest open-weight models, live experiment tracking, Expert LoRA, early stopping, tokenized dataset previews, pre-flight validation, and lower training prices on selected models.

Nguyên văn nguồn
Chi tiết nguồn
Kỹ thuật

Cloudera and Mistral Partner to Bring Specialized, Sovereign Intelligence to Enterprise Data

Cloudera and Mistral join forces to bring specialized, sovereign AI intelligence to enterprise data, helping regulated industries innovate on their own terms.

Nguyên văn nguồn
Chi tiết nguồn
Tác tử

Cursor Projects

Today we're launching Projects in Cursor. Projects lets you take on larger bodies of work, such as a feature, a migration, or a full app. It maintains context over months of work, delegates tasks to thousands of subagents, and performs recurring work without being prompted. Access Projects from the left-hand nav. The coordinator agent in a project doesn't write code itself; it plans the work, delegates it to agents that implement it, and brings the finished work back to you to check. Coordinators create and manage agents on your behalf, running as many in parallel as the work needs. Powered by Cloud Agents A Project runs on its own computer in the cloud, so closing your laptop doesn't stop it. When something needs testing on your machine, the coordinator spins up a local agent to run it there. Shared context You shouldn't have to onboard an agent every time you start a task. Each Project maintains a set of files that sync across every cloud and local machine its agents use. Agents add research and artifacts, along with what they learn about the codebase and how you prefer work to be done. If one agent figures out how to test a service, for example, every future agent can use those instructions. The shared context grows with the Project, making the coordinator more effective over time. Subscriptions Tell the coordinator agent to watch a Slack channel, run on a schedule, or follow all your PRs. This way it can take action based on signals it detects, without waiting for your prompt. Connect Slack and point it at a bug-report channel, and it starts delegating each time a bug comes in. Projects are available in beta and rolling out to all users starting today.

Nguyên văn nguồn
Chi tiết nguồn
Kỹ thuật

To Infinity and Beyond: ThunderKittens Now on NVIDIA Vera Rubin NVL72!

We ported ThunderKittens to NVIDIA's Vera Rubin NVL72 and rebuilt our NVFP4 GEMM around the new hardware, taking it from 42% of roofline to over 22 PFLOPS — competitive with cuBLAS and CuTe DSL. Here is what changed in the ISA and how we used it.

Nguyên văn nguồn
Chi tiết nguồn
Tác tử

v0.22.2

<h2>What's Changed</h2> <ul> <li>feat: support current image generation tool options by <a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/users/seratch/hovercard" data-octo-click="hovercard-link-click" data-octo-dimensions="link_type:self" href="https://github.com/seratch">@seratch</a> in <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="5392006170" data-permission-text="Title is private" data-url="https://github.com/openai/openai-agents-python/issues/4932" data-hovercard-type="pull_request" data-hovercard-url="/openai/openai-agents-python/pull/4932/hovercard" href="https://github.com/openai/openai-agents-python/pull/4932">#4932</a></li> <li>fix(sandbox): prevent UnixLocal file API symlink races by <a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/users/seratch/hovercard" data-octo-click="hovercard-link-click" data-octo-dimensions="link_type:self" href="https://github.com/seratch">@seratch</a> in <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="5391970420" data-permission-text="Title is private" data-url="https://github.com/openai/openai-agents-python/issues/4931" data-hovercard-type="pull_request" data-hovercard-url="/openai/openai-agents-python/pull/4931/hovercard" href="https://github.com/openai/openai-agents-python/pull/4931">#4931</a></li> <li>fix(sessions): reset compaction response chain after pop by <a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/users/rksharma-owg/hovercard" data-octo-click="hovercard-link-click" data-octo-dimensions="link_type:self" href="https://github.com/rksharma-owg">@rksharma-owg</a> in <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="5394300046" data-permission-text="Title is private" data-url="https://github.com/openai/openai-agents-python/issues/4934" data-hovercard-type="pull_request" data-hovercard-url="/openai/openai-agents-python/pull/4934/hovercard" href="https://github.com/openai/openai-agents-python/pull/4934">#4934</a></li> </ul> <h3>Documentation &amp; Other Changes</h3> <ul> <li>docs: document v0.22.1 behavior updates by <a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/users/seratch/hovercard" data-octo-click="hovercard-link-click" data-octo-dimensions="link_type:self" href="https://github.com/seratch">@seratch</a> in <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="5219327869" data-permission-text="Title is private" data-url="https://github.com/openai/openai-agents-python/issues/4577" data-hovercard-type="pull_request" data-hovercard-url="/openai/openai-agents-python/pull/4577/hovercard" href="https://github.com/openai/openai-agents-python/pull/4577">#4577</a></li> <li>release: 0.22.2 by <a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/users/seratch/hovercard" data-octo-click="hovercard-link-click" data-octo-dimensions="link_type:self" href="https://github.com/seratch">@seratch</a> in <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="5397784189" data-permission-text="Title is private" data-url="https://github.com/openai/openai-agents-python/issues/4935" data-hovercard-type="pull_request" data-hovercard-url="/openai/openai-agents-python/pull/4935/hovercard" href="https://github.com/openai/openai-agents-python/pull/4935">#4935</a></li> </ul> <p><strong>Full Changelog</strong>: <a class="commit-link" href="https://github.com/openai/openai-agents-python/compare/v0.22.1...v0.22.2"><tt>v0.22.1...v0.22.2</tt></a></p>

Nguyên văn nguồn
Chi tiết nguồn
  • v0.22.2 openai/openai-agents-python releases · 9 thg 9, 2026
Huấn luyện

v0.30.0

<h2>What's Changed</h2> <h3>Notable Changes</h3> <ul> <li><code>wandb.InternalApi</code>, the <code>wandb.PublicApi</code> and <code>wandb.apis.PublicApi</code> aliases, and the <code>wandb.apis.internal</code> and <code>wandb.sdk.internal.internal_api</code> modules have been removed. Use <code>wandb.Api()</code> instead.</li> <li><code>wandb.api</code> and <code>wandb.ensure_configured()</code> are deprecated. Use <code>wandb.login(prompt=False)</code> to check whether the user is logged in, and <code>wandb.Api()</code> for everything else.</li> <li>The legacy <code>wandb sync</code> options have been removed. Run <code>wandb sync</code> with no arguments instead of <code>wandb sync --sync-all</code>, and use <code>--job-type</code> instead of <code>--job_type</code>. See <code>wandb sync --help</code>.</li> </ul> <h3>Added</h3> <ul> <li><code>wandb.login(prompt=False)</code> uses the configured credentials when there are any and returns <code>False</code> otherwise, without asking for an API key or switching to offline mode. Use it to check whether the user is logged in without blocking (<a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/users/dmitryduev/hovercard" data-octo-click="hovercard-link-click" data-octo-dimensions="link_type:self" href="https://github.com/dmitryduev">@dmitryduev</a> in <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="5354300146" data-permission-text="Title is private" data-url="https://github.com/wandb/wandb/issues/12741" data-hovercard-type="pull_request" data-hovercard-url="/wandb/wandb/pull/12741/hovercard" href="https://github.com/wandb/wandb/pull/12741">#12741</a>)</li> <li>The automations API supports sending a prompt to ARIA as an automation action with <code>SendPromptToAria</code> (<a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/users/gdecarvalhovaz-lgtm/hovercard" data-octo-click="hovercard-link-click" data-octo-dimensions="link_type:self" href="https://github.com/gdecarvalhovaz-lgtm">@gdecarvalhovaz-lgtm</a> in <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="5261609338" data-permission-text="Title is private" data-url="https://github.com/wandb/wandb/issues/12594" data-hovercard-type="pull_request" data-hovercard-url="/wandb/wandb/pull/12594/hovercard" href="https://github.com/wandb/wandb/pull/12594">#12594</a>)</li> <li>W&amp;B LEET TUI: <ul> <li>Charts with custom x-axes set with <code>run.define_metric(...)</code>, including glob definitions such as <code>run.define_metric("train/*", step_metric="train/step")</code>. The axis name is shown as <code>[x: train/step]</code> in the chart header. Applies to runs opened from local <code>.wandb</code> files (<a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/users/dmitryduev/hovercard" data-octo-click="hovercard-link-click" data-octo-dimensions="link_type:self" href="https://github.com/dmitryduev">@dmitryduev</a> in <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="5240910608" data-permission-text="Title is private" data-url="https://github.com/wandb/wandb/issues/12568" data-hovercard-type="pull_request" data-hovercard-url="/wandb/wandb/pull/12568/hovercard" href="https://github.com/wandb/wandb/pull/12568">#12568</a>, <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="5345813404" data-permission-text="Title is private" data-url="https://github.com/wandb/wandb/issues/12728" data-hovercard-type="pull_request" data-hovercard-url="/wandb/wandb/pull/12728/hovercard" href="https://github.com/wandb/wandb/pull/12728">#12728</a>)</li> <li>Filter console logs: with the logs pane focused, press <code>/</code> and type a pattern (regex, or glob after <code>Tab</code>) to show only the matching lines. New matches are followed as they arrive, and <code>ctrl+/</code> clears the filter (<a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/users/dmitryduev/hovercard" data-octo-click="hovercard-link-click" data-octo-dimensions="link_type:self" href="https://github.com/dmitryduev">@dmitryduev</a> in <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="5353176604" data-permission-text="Title is private" data-url="https://github.com/wandb/wandb/issues/12736" data-hovercard-type="pull_request" data-hovercard-url="/wandb/wandb/pull/12736/hovercard" href="https://github.com/wandb/wandb/pull/12736">#12736</a>)</li> <li>The metrics, system metrics and runs filters are remembered for each wandb directory and applied again the next time you open it, in both the workspace and the single-run view. They are stored in <code>.wandb-leet.json</code> inside the directory. Clearing a filter forgets it (<a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/users/dmitryduev/hovercard" data-octo-click="hovercard-link-click" data-octo-dimensions="link_type:self" href="https://github.com/dmitryduev">@dmitryduev</a> in <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="5353176206" data-permission-text="Title is private" data-url="https://github.com/wandb/wandb/issues/12735" data-hovercard-type="pull_request" data-hovercard-url="/wandb/wandb/pull/12735/hovercard" href="https://github.com/wandb/wandb/pull/12735">#12735</a>)</li> <li>The border or separator line you drag to resize a pane or sidebar is highlighted while you drag it (<a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/users/dmitryduev/hovercard" data-octo-click="hovercard-link-click" data-octo-dimensions="link_type:self" href="https://github.com/dmitryduev">@dmitryduev</a> in <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="5273609369" data-permission-text="Title is private" data-url="https://github.com/wandb/wandb/issues/12617" data-hovercard-type="pull_req

Nguyên văn nguồn
Chi tiết nguồn
  • v0.30.0 Weights & Biases — SDK releases · 9 thg 9, 2026
Kỹ thuật

The Open Source AI Stack

A deep dive into the open model AI stack — model, inference, gateways and routers, harness, and tools — and how keeping each layer independent lets you swap in a new open model in minutes instead of rebuilding your workflow.

Nguyên văn nguồn
Chi tiết nguồn
701–720 / 738
Trang 36 / 37