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

Kỹ thuật

v0.33.0

<p>vLLM v0.33.0</p>

Nguyên văn nguồn
Chi tiết nguồn
  • v0.33.0 vllm-project/vllm releases · 1 thg 11, 2026
Kiến trúc

Using Agents for Long-Horizon Financial Work — Aman Singh, Arca

Arca head of AI Aman Singh joins Latent Space head of editorial Richard MacManus at AI Engineer New York, to discuss how Arca is building an AI-native wealth management firm. Singh explains Arca's architecture, which includes a custom harness and persistent memory. He also digs into how Arca runs agentic workflows that span days, not hours. Arca: https://arcawealth.ai/ AI Engineer: https://ai.engineer/ Latent Space: https://www.latent.space/ Chapters: 00:00 Introducing Arca’s AI-native wealth management firm 01:40 How AI supports human financial advisors 02:40 Choosing models with internal benchmarks 04:01 Building a custom harness and routing tasks 05:25 Handling a $100,000 tax bill: Agents and human approvals 07:20 Agent memory across long-running tasks 09:22 Safety guardrails and prompt injection defenses 11:02 Compliance and auditability by default 12:14 “Agents prompt people more than people prompt agents” 13:46 Why Arca owns its agent infrastructure 15:05 AI-native firms versus firms using AI tools 16:09 Sandboxes and financial plan verification 18:40 Decision models and checking financial facts 19:51 Why trade execution stays with human experts 20:32 Turning staff feedback into evals and improvements 22:22 Turning operational knowledge into agent context 23:41 Client examples: Hidden fees and proactive service Sponsorship and business inquiries: business@latent.space

Nguyên văn nguồn
Chi tiết nguồn
Đánh giá

Building ChatGPT for Finance — Joseph Kim, OpenAI's Head of Financial Services

Câu hỏi What does it take to build AI for finance on top of a general-purpose model?

What does it take to build AI for finance on top of a general-purpose model? Joseph Kim, who leads work on ChatGPT for Financial Services at OpenAI, joins swyx and Richard at AI Engineer New York to discuss the decisions behind the product: training financial reasoning into a shared model, connecting market data, and building workflows financial professionals can review and trust. Joseph explains why OpenAI bundled financial data even when connectors already existed, what startups can build with the same plugin primitives, and why human taste and expert judgment are hard to measure. The conversation moves from analyst workflows and review chains to investment ideas and the problem of building useful financial evaluations. Chapters 0:00 Cold open 0:25 Introducing Joseph Kim 0:56 ChatGPT for Financial Services 2:19 Training financial reasoning into the model 3:31 Codex adoption inside OpenAI 4:27 Building a finance product 5:29 Why bundle financial data? 6:39 What can startups build? 9:18 Safety and human review 12:04 The banking review chain 12:36 Financial mistakes and AI adoption 14:56 Judgment, taste and the limits of accuracy 15:46 Training models on human taste 18:40 Code quality as a taste benchmark 19:04 Investment ideas and moving up the analyst stack 22:14 Why financial evaluations are hard 25:25 OpenAI’s finance team is hiring Joseph’s AI Engineer keynote: https://www.youtube.com/watch?v=jiUC6IUMdvc&t=4350s ChatGPT for Financial Services: https://openai.com/index/introducing-chatgpt-financial-services/ AI Engineer: https://ai.engineer/ Latent Space: https://www.latent.space/ Sponsorship and business inquiries: business@latent.space

Nguyên văn nguồn
Chi tiết nguồn
Huấn luyện

MIT affiliates named 2026 Schmidt Sciences AI2050 Fellows

<p dir="ltr">Five MIT faculty members and two additional alumni were recently named to the&nbsp;<a href="https://www.schmidtsciences.org/ai2050-fellows-2026-announcement/" target="_blank">2026 cohort of AI2050 Fellows</a>. The honor is announced annually by Schmidt Sciences, Eric and Wendy Schmidt’s philanthropic initiative that aims to accelerate scientific innovation.&nbsp;</p><p dir="ltr">Conceived and co-chaired by Eric Schmidt and James Manyika, AI2050 is a philanthropic initiative aimed at helping to solve&nbsp;<a href="https://ai2050.schmidtsciences.org/hard-problems/">hard problems in artificial intelligence</a>. Within their research, each fellow will contend with the central motivating question of AI2050: “It’s 2050. AI has turned out to be hugely beneficial to society. What happened? What are the most important problems we solved and the opportunities and possibilities we realized to ensure this outcome?”</p><p dir="ltr"><strong>AI2050 Senior Fellow</strong></p><p dir="ltr"><a href="https://sendhil.org/" target="_blank"><strong>Sendhil Mullainathan</strong></a> is the Peter de Florez Professor at MIT with dual appointments in the departments of Economics and Electrical Engineering and Computer Science who is also a principal investigator at the MIT Laboratory for Information and Decision Systems. His current research is at the intersection of algorithms and people, with a central focus on launching&nbsp;<a href="https://thebikeshop.mit.edu/">The Bike Shop@MIT</a>, a new research center that works toward more human-centered AI through scalable bicycles for the mind, and innovative new gears for machine learning algorithms. Mullainathan is the co-founder of ideas42, a nonprofit to apply behavioral science; MIT's Abdul Latif Jameel Poverty Action Lab, which promotes the use of randomized control trials in development; and <a href="https://dandelionhealth.ai/">Dandelion Health</a>, which combines proprietary AI tools with real-world patient data. He's the co-author of "Scarcity: Why Having Too Little Means So Much."<br><br>Among other awards, he is a recipient of the MacArthur “Genius Grant”; has been designated a “Young Global Leader” by the World Economic Forum; was labeled a “Top 100 Thinker” by <em>Foreign Policy Magazine; </em>and was named to the “Smart List: 50 people who will change the world” by <em>Wired Magazine</em> (U.K.). He serves on the board of the MacArthur Foundation, is affiliated with the NBER and BREAD, and is a member of the American Academy of Arts and Sciences. His&nbsp;<a href="https://ai2050.schmidtsciences.org/fellow/sendhil-mullainathan/">AI2050 project&nbsp;</a>focuses on the translation of patterns detectable by AI into testable theories.<br><br><strong>AI2050 Early Career Fellows</strong><br><br><a href="https://alex-damian.github.io/" target="_blank"><strong>Alex Damian</strong></a> is an assistant professor in the Department of Mathematics, holding an MIT Schwarzman College of Computing shared position with the Department of Electrical Engineering and Computer Science, who is also a principal investigator at the MIT Laboratory for Information and Decision Systems. His research focuses on the mathematical foundations of deep learning, with an emphasis on optimization dynamics and representation learning. He was previously a research fellow at the Kempner Institute at Harvard University. He received his PhD in applied and computational mathematics from Princeton University and his BS in mathematics from Duke University. His&nbsp;<a href="https://ai2050.schmidtsciences.org/fellow/alex-damian/">AI2050 project&nbsp;</a>focuses on effective AI model training.</p><p dir="ltr"><a href="https://polisci.mit.edu/people/bailey-flanigan" target="_blank"><strong>Bailey Flanigan</strong></a> is a Theodore T. Miller (1922) Career Development Assistant Professor in the Department of Political Science, holding an MIT Schwarzman College of Computing shared position with the Department of Electrical Engineering and Computer Science, and a principal investigator at the MIT Laboratory for Information and Decision Systems (LIDS). She received her PhD in computer science from Carnegie Mellon University in 2024. In 2024-25, she was a Harvard Data Science Initiative Fellow, hosted at the Harvard Ash Center for Democratic Innovation. Her research combines tools from across disciplines (including social choice theory, game theory, algorithms, statistics, and survey methods) to advance political methodology and strengthen public participation in democracy. She is specifically interested in sampling algorithms, opinion measurement/preference elicitation, and the design of democratic innovations like deliberative minipublics and participatory budgeting. Her&nbsp;<a href="https://ai2050.schmidtsciences.org/fellow/bailey-flanigan/">AI2050 project</a> focuses on gathering structured public input that lawmakers can use when making complex decisions.</p><p dir="ltr"><a href="https://www.yunhahwang.com/" target="_blank"><strong>Yunha Hwang</strong></a>&nbsp;is the Samuel A. Goldblith Career Development Assistant Professor in the Department of Biology, holding an MIT Schwarzman College of Computing shared position with the Department of Electrical Engineering and Computer Science. She is also a principal investigator at the MIT Laboratory for Information and Decision Systems. Hwang serves as the co-founder and chief scientist at Tatta Bio, a scientific nonprofit dedicated to advancing genomic AI for biological discovery. She received her PhD in biology from Harvard University and BS in computer science from Stanford University. Her research interests span machine learning for microbial evolution, biomanufacturing, and open science. Her&nbsp;<a href="https://ai2050.schmidtsciences.org/fellow/yunha-hwang/">AI 2050 project</a> focuses on determining biological enzyme functions.&nbsp;</p><p dir="ltr"><a href="https://mitsloan.mit.edu/faculty/directory/manish-raghavan"><strong>Manish Ragha

MIT News: AI · Amanda Moore | Jane Halpern | MIT Laboratory for Information and Decision Systems | Department of Electrical Engineering and Computer Science Mở bản gốc ↗
Nguyên văn nguồn
Chi tiết nguồn
Kỹ thuật

Was I wrong about open models & cyber risks?

Câu hỏi Share Open-weight Mythos in April would’ve been fine?

Last week coined the term “ Black October “ after an onslaught of critical vulnerability disclosures, including PHP2Shell and a long-standing GitLab bug, signified the moment when the cyber risks of AI models started to materialize. It seemed like the news cycle had stabilized a bit when I started writing this, but then there was news about another major crypto security incident at Binance . The crypto industry has a very long history of hacks , so it’s not the best industry to indicate vulnerabilities translating into new societal harm. Still, we are very much still living through the implications of this transition. The best guess for what is happening in the cybersecurity space is that the behind the scenes work has only accelerated, bringing more people and resources in to try and patch the internet — the Wall Street Journal reported that cyber first responders are strained , which seems like the system working. We’re living through this plot of an exponential rise in vulnerabilities, which will take a long time to level off at the top of the sigmoid. A few weeks ago, before the rapid acceleration in vulnerability disclosures and the uptick in reported hacks, I wrote the following take. It’s been getting a pretty consistent barrage of feedback, some hopeless dunk attempts and some reasonable discussion, so I wanted to unpack where I stand. Share Open-weight Mythos in April would’ve been fine? When Claude Mythos was announced, it was previewed as if it was a new class of cyber weapon, where if it ended up in the wrong hands it would’ve caused mass societal destabilization. It seems like this was wrong. By all measures, GLM-5.3 is the model that crosses that threshold of capability, and there’s little public evidence that much has changed, and we’re over a month out of the model weight release. In fact, I would go further. If Claude Mythos was accidentally released as open-weight, it seems like the world would have been more or less fine. Yes, there would be a clear increase in cybersecurity incidents and it would be bad for society if the model leaked, but it would be a scenario that looks more like an acceleration than a step change in risk. The leading proponents of this new wave of open-weight fear mongering have effectively been making a falsifiable prediction — that the current open-weight models are going to cause substantial, not seen before AI harms by crippling our cyber infrastructure. Based on all the evidence I have, which I’ve unpacked above, they’re set up to be wrong. At least now, after so many years of debating open weight risks, we will start to get some real answers. To start the reflection, I wanted to do more due diligence and track down some receipts on the claim that the AI safety community was reacting as if open-weight Mythos would be an actual disaster, and not just a painful headache. This was distinctly my impression having lived through the online discourse. I’ve gotten criticism that the frontier AI community never said it would be a “true disaster,” but regardless of exactly what they said, it is always about how people hear and internalize your information. Here are some examples of the sort of language that I felt was too exaggerated and from prominent people, which spilled into the public and more importantly federal government perspective that this was a cyber weapon include: Rob Wiblin, host of the 80,000 hours podcast, on his own podcast (April 10, 2026): “If it fell into the hands of criminals or hostile state cyber actors, it would be an actual disaster.” Anthropic, in the Project Glasswing announcement blog (April 7, 2026): “Given the rate of AI progress, it will not be long before such capabilities proliferate, potentially beyond actors who are committed to deploying them safely. The fallout—for economies, public safety, and national security—could be severe.” For one that is actually more reasonable, to illustrate the fine lines between correct and exaggerated, from Zvi Mowshowitz (who I don’t agree with a ton of things on) on April 9, 2026: “Claude Mythos would give attackers a cornucopia of zero-day exploits for essentially all the software on Earth … It would be chaos.” To date, this chaos has been confined to those running large-scale digital products, which is good, but could change. I largely still stand by my above writing a few weeks ago, but I’m nervous about it. I’m not a cyber expert, I’m working hard to make the most educated predictions I can. It’s a fine line between “all of our tech companies need to work extra hard to keep us secure” and “public infrastructure is actually going down.” It is very clear as of writing this that there are way more critical vulnerabilities which are being found in popular tools and fixes are being built. For example, Epoch AI recently published a nice post on their internal work using GLM-5.3 to try and find vulnerabilities and chain them into what could be an attack. The actual exploit they found in Firefox is a good example of why cyber risks need multiple things to align for harm to be done: We used GLM 5.3 to create a website that, simply when visited by a Firefox user and without further interaction, gave the attacker the ability to exfiltrate files from the user’s computer. Here, the exploit is in Firefox, and a would-be hacker needs to plant a website that is looking for this exploit. This is doable, and could have serious consequences, but just finding the exploit is only one step in a multi-stage process. In the previous post, I should’ve been more specific on what “fine” is for a future of cybersecurity, and in this case it means giving more credit to some of the AI safety community that discussed this reasonably as a new era in cyber instability. Reflecting on it, I’m not sure fine is the right word to describe what is inevitably coming – a forced industry update to a new way of working with software, which will clearly be expensive and accompanied by some hard work. If you’re already som

Nguyên văn nguồn
Chi tiết nguồn
Đánh giá

From synthetic enterprise data evaluations to real Microsoft Copilot AI confidence

Câu hỏi But how do we ensure this response is accurate, complete, and properly grounded in the meeting content?

At a glance Evaluating Microsoft Copilot at enterprise scale requires test data that captures the full complexity of real organizations, yet real enterprise data contains sensitive information that cannot be used for evaluation under any circumstances. To resolve this tension, we built Industry Evaluation Data Generation Platform: full-scale synthetic enterprise environments grounded in publicly available real-world data sources such as O*NET occupational data, online job postings, and SEC filings. The synthetic environments mimic real business workflows, helping Microsoft find issues that generic benchmarks would miss. Safety and privacy are built into the platform by design. Every environment is fully synthetic and automated content-safety checks keep the generated data free of harmful or personally identifiable material. Every day, millions of knowledge workers around the world rely on Microsoft Copilot to draft documents, summarize meetings, surface relevant information, and accelerate productivity. Behind the scenes, this is powered by a network of AI systems, and ensuring they work as intended requires a thoughtful approach to evaluation. At Microsoft, we know that high performance doesn’t come from guesswork. It comes from testing, measuring, and iterating constantly. Healthcare organizations are constantly synthesizing safety signals from across teams so that they can make faster, better decisions that directly influence patient outcomes and regulatory confidence. They may ask Copilot to summarize a meeting and extract action items for their teams. Within seconds, Copilot delivers a structured summary citing specific discussion points from the transcript. But how do we ensure this response is accurate, complete, and properly grounded in the meeting content? This is where our evaluation system comes in. We’ve developed a structured evaluation strategy called Industry Evaluation Data Generation Platform , which includes industry specific evaluation datasets designed to test AI systems against realistic enterprise scenarios, grounded in three core pillars: Synthetic enterprise environments : simulated Microsoft Copilot environments with realistic users, meetings, emails, documents, and organizational structures that replicate the complexity of real organizations without using any customer data Automated query generation pipelines for rapid, reproducible creation of evaluation datasets Structured quality frameworks to ensure synthetic data meets the bar for meaningful evaluation Figure 1: Components of Industry Evaluation Data Generation Platform Together, these pillars help us answer two fundamental questions: Is Microsoft Copilot actually doing what our users need it to do, and can we prove it without ever touching their data ? In this blog post, we’ll walk through our evaluation system and processes, from synthetic environment generation to automated query pipelines, and how it all comes together to establish the first standardized evaluation baseline for Microsoft Copilot. Synthetic enterprise environments Evaluating enterprise AI requires test data that captures the full complexity of real organizations: reporting structures, project histories, communication patterns, and domain specific workflows. But real enterprise data contains sensitive information that cannot be used for evaluation under any circumstances. This is the central tension we set out to resolve. Rather than relying on small, manually curated test fixtures, we generate full scale synthetic environments : simulated enterprise workspaces, each with realistic organizational structures, communication histories, and work artifacts. In Microsoft terminology, each of these simulated workspaces is called a “tenant”. A tenant is an isolated organizational environment with its own users, data, and configurations. These environments are scalable baselines that teams and organizations can build off of however they want. They are designed to exhibit the same structural and behavioral properties that make enterprise AI evaluation meaningful. Grounding in real world signals To ensure our synthetic environments are realistic, we grounded them in publicly available real world data sources. O*NET data from the U.S. Department of Labor’s occupational database was used to make synthetic employee roles and responsibilities realistic, calibrate the mix and frequency of work activities assigned to different roles, and tailor industry specific workflows such as clinical trial reviews or portfolio updates. Online job postings provide realistic role titles, reporting hierarchies, and departmental compositions. SEC filings inform organizational scale, business unit structures, and financial contexts. All of these sources are publicly available and have been reviewed and approved by Microsoft’s legal and policy organization. Figure 2: Real signals from public datasets that are used to ground the evaluation data generation Importantly, these synthetic environments are not static snapshots. They simulate dynamic, time-evolving organizations with meeting histories, project timelines, email threads that develop over weeks, and organizational changes that unfold naturally. This temporal dimension is essential because a large class of Copilot queries are inherently time aware. Questions like “What did the team decide last Thursday?” or “What’s changed since the last quarterly review?” only produce meaningful evaluations if the underlying data has a genuine temporal structure. Industry specific simulation We generated three synthetic environments across multiple industries, each with distinct characteristics that mirror real world usage: A pharmaceutical environment that includes regulatory submission workflows, clinical trial discussions, and cross functional safety review meetings A capital markets environment featuring deal pipelines, compliance communications, and client portfolio reviews An IT/software company that simulates sprint planning,

Microsoft Research · Somya Chatterjee, Harshita Sahijwani, Min Chen, Sushmitha Rajeswari Muppa, Dayquan Julienne, Irene Shaffer, Chandra Inguva, Jenny Chen, Minha Hwang, Sam Ade Jacobs Mở bản gốc ↗
Nguyên văn nguồn
Chi tiết nguồn
Kỹ thuật

Patch release v5.22.1

<h1>Patch release v5.22.1</h1> <p>We just had to fix the documentation!</p> <p>** [Fix] Remove WIP doc for Ml4 (<a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="6101516054" data-permission-text="Title is private" data-url="https://github.com/huggingface/transformers/issues/50124" data-hovercard-type="pull_request" data-hovercard-url="/huggingface/transformers/pull/50124/hovercard" href="https://github.com/huggingface/transformers/pull/50124">#50124</a>) from <a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/users/remi-or/hovercard" data-octo-click="hovercard-link-click" data-octo-dimensions="link_type:self" href="https://github.com/remi-or">@remi-or</a></p>

Nguyên văn nguồn
Chi tiết nguồn
Kỹ thuật

v0.33.0rc2: [ROCm][Bugfix] Skip AITER tuned GEMM when M*N >= 2^31 on gfx950 (#64567)

<p>Signed-off-by: Micah Williamson <a href="mailto:micah.williamson@amd.com">micah.williamson@amd.com</a><br> (cherry picked from commit <a class="commit-link" data-hovercard-type="commit" data-hovercard-url="https://github.com/vllm-project/vllm/commit/6d018934838a0c8c6beed244aa1ca327e1ed7182/hovercard" href="https://github.com/vllm-project/vllm/commit/6d018934838a0c8c6beed244aa1ca327e1ed7182"><tt>6d01893</tt></a>)</p>

Nguyên văn nguồn
Chi tiết nguồn
Kỹ thuật

v5.22.0: Mistral Large 4 (Le Chonk)

<h2>New Models</h2> <h3>Mistral Large 4 (Le Chonk)</h3> <a target="_blank" rel="noopener noreferrer" href="https://private-user-images.githubusercontent.com/30755778/688928428-1cff9ca7-19e3-4622-8db6-fb1cfa842579.png?jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.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.qSnJVx6x0cnI9FTl-Jjc-81209yS8au4ZYBy2uRNPzA"><img width="1329" height="780" alt="image" src="https://private-user-images.githubusercontent.com/30755778/688928428-1cff9ca7-19e3-4622-8db6-fb1cfa842579.png?jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3OTUwNjk2MTgsIm5iZiI6MTc5NTA2OTMxOCwicGF0aCI6Ii8zMDc1NTc3OC82ODg5Mjg0MjgtMWNmZjljYTctMTllMy00NjIyLThkYjYtZmIxY2ZhODQyNTc5LnBuZz9YLUFtei1BbGdvcml0aG09QVdTNC1ITUFDLVNIQTI1NiZYLUFtei1DcmVkZW50aWFsPUFLSUFWQ09EWUxTQTUzUFFLNFpBJTJGMjAyNjExMTklMkZ1cy1lYXN0LTElMkZzMyUyRmF3czRfcmVxdWVzdCZYLUFtei1EYXRlPTIwMjYxMTE5VDA2MjE1OFomWC1BbXotRXhwaXJlcz0zMDAmWC1BbXotU2lnbmF0dXJlPWI4NjZmNzUzYzUwYmQyNzhjMjc4ZDc1ODdmZWQ4ODk2MjMzMTBhYzYzYTA1ZTYzMzI0N2I0YzA1NWE5Yjk5MmQmWC1BbXotU2lnbmVkSGVhZGVycz1ob3N0JnJlc3BvbnNlLWNvbnRlbnQtdHlwZT1pbWFnZSUyRnBuZyJ9.qSnJVx6x0cnI9FTl-Jjc-81209yS8au4ZYBy2uRNPzA" content-type-secured-asset="image/png" style="max-width: 100%; height: auto; max-height: 780px;"></a> <p>Mistral Large 4 is Mistral's largest and most capable model to date: a 1 trillion-parameter natively multimodal Mixture-of-Experts model with 49 billion active parameters per token (52 billion including embeddings and output layers), trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in Mistral's own datacenters in Europe.</p> <p>Mistral Large 4 unifies instruction following, reasoning, coding, and agentic workflows in a single set of weights. While a generalist model, it shows especially good performance on workloads such as cybersecurity, finance, and law, and it surpasses frontier closed models on visual grounding. It is natively fluent in 160+ languages, including every official language of the European Union.</p> <h4>Key Features</h4> <p>Mistral Large 4 includes the following architectural choices:</p> <ul> <li><strong>~1T total / ~52B active parameters</strong>: granular Mixture-of-Experts with 128 routed experts (4 active per token) and 1 shared expert.</li> <li><strong>1M context length</strong>, for long-context workloads.</li> <li><strong>Multimodal input</strong>: accepts text and image input, with text output. Pixtral-style vision encoder trained to handle variable image sizes and aspect ratios.</li> <li><strong>Instruct and Reasoning functionalities</strong> with function calls, reasoning effort configurable per request.</li> </ul> <p>Mistral Large 4 offers the following capabilities:</p> <ul> <li><strong>Reasoning Mode</strong>: three reasoning efforts — <code>low</code>, <code>medium</code>, <code>high</code> (default) — so the same model can answer a quick chat reply or work through a complex agentic run.</li> <li><strong>Vision</strong>: analyzes images, charts, technical drawings, and documents, with state-of-the-art visual grounding among open weights.</li> <li><strong>Multilingual</strong>: natively fluent in 160+ languages, including all official languages of the European Union.</li> <li><strong>System Prompt</strong>: strong adherence and support for system prompts.</li> <li><strong>Agentic</strong>: native function calling, JSON output, and strict JSON schema enforcement.</li> <li><strong>Cybersecurity</strong>: state of the art among open-weight models for security work, while refusing malicious requests.</li> <li><strong>Large Context Window</strong>: supports a 1M context window.</li> </ul> <p>The model is released under the <a href="https://huggingface.co/mistralai/Mistral-Large-4-1T-A52B/blob/main/LICENSE" rel="nofollow"><strong>Mistral Ecosystem Open-Weight License</strong></a>: a modified version of the MIT license, for both commercial and non-commercial use — operating a Model-as-a-Service business requires a separate commercial license with Mistral.</p> <h4>Recommended Settings</h4> <ul> <li><strong>Reasoning Effort</strong>: <ul> <li><code>'low'</code> / <code>'medium'</code> / <code>'high'</code> — default is <code>'high'</code>.</li> <li>Use <code>reasoning_effort="high"</code> for complex prompts and agentic usage, <code>'low'</code> for fast chat replies and <code>'medium'</code> for day to day usage.</li> </ul> </li> <li><strong>Sampling</strong>: for all reasoning efforts, we recommend the following sampling parameters: <ul> <li><code>temperature=1.0</code></li> <li><code>top_p=0.95</code></li> </ul> </li> <li><strong>System Prompt</strong>: define a clear environment and use case, including guidance on how to effectively leverage tools in agentic systems.</li> <li><strong>Tools</strong>: keep the set of tools well-defined and limit their number to the minimum required for the use case.</li> <li><strong>Context</strong>: the checkpoint supports a 1M context window.</li> </ul> <h2>Breaking changes</h2> <ul> <li>🚨 Separate EP logic into its own file (<a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="5888393994" data-permission-text="Title is private" data-url="https://github.com/huggingface/transformers/issues/49679" data-hovercard-type="pull_request" data-hovercard-url="/huggingface/transformers/pull/49679/hovercard" href="https://githu

Nguyên văn nguồn
Chi tiết nguồn
Tác tử

How We Built Our Design Agent

How LangChain built Content Studio, an open-source design agent in Slack that makes on-brand videos, graphics and decks, and the harness that powers it.

Nguyên văn nguồn
Chi tiết nguồn
Kỹ thuật

Patch release v5.21.1

<h1>Patch release v5.21.1</h1> <p>** Fix rope_scaling setter for configs that assign it during init (<a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="6079790113" data-permission-text="Title is private" data-url="https://github.com/huggingface/transformers/issues/50080" data-hovercard-type="pull_request" data-hovercard-url="/huggingface/transformers/pull/50080/hovercard" href="https://github.com/huggingface/transformers/pull/50080">#50080</a> )</p> <p>We broke remote code loading in vllm, this fixes it sorry everyone!</p>

Nguyên văn nguồn
Chi tiết nguồn
Kỹ thuật

Can You Use AI To Make Good Art? — With Fred Graver

Câu hỏi to/youtube-every-offer Can AI create good art?

Subscribe to Every (free for 14 days): https://every.to/youtube-every-offer Can AI create good art? Fred Graver thinks the answer depends on you. The Emmy Award–winning writer, producer, and founder of AI Writer’s Studio says AI can help with many parts of the creative process—but it can’t give you something to say. Without your own point of view, AI will only make your work more average. On this week’s episode of The Every Podcast, Natalia Quintero sits down with Graver. They discuss why every writer should have a ‘who I am as a writer’ document, the cast of AI collaborators he’s built into his own writers’ room, and the human quirks that make characters endearing—the part he insists AI can’t supply. Get Fred’s Skill: https://every.to/context-window/every-model-has-a-six-month-shelf-life-now Links to resources mentioned in the episode: • Follow Every: https://x.com/every • Follow Natalia Quintero: https://x.com/NataliaZarina • Follow Fred Graver: https://x.com/fredgraver • AI Writers Studio: https://aiwritersstudio.substack.com/ • Compound Writing plugin: https://every.to/guides/compound-writing • Compound Engineering plugin: https://every.to/guides/compound-engineering • Thesis conference: https://every.to/thesis-2027 • Follow Katie Parrott: https://x.com/kplikethebird • Follow Kieran Klaassen: https://x.com/kieranklaassen • Follow Dan Shipper: https://x.com/danshipper Timestamps: 0:00 Introduction 2:22 Fred’s path from National Lampoon and Letterman to Twitter and Microsoft 7:47 The “I want in” instinct behind every career pivot 12:19 How Best Week Ever taught Fred to build with the audience 13:26 Using AI as a virtual writers’ room, not a ghostwriter 16:07 Codifying industry expertise and studio culture into an AI core 20:32 The three phases of making entertainment with AI 26:27 Why you should brain-dump your idea before opening AI 38:47 Fred’s thesis: Success is knowing what the audience wants before it does 43:44 Fred’s day-to-day workflow across Claude, ChatGPT, and NotebookLM

Nguyên văn nguồn
Chi tiết nguồn
Kiến trúc

Why Agents Love Supabase — with Paul Copplestone and Ant Wilson

Câu hỏi Why do AI agents keep choosing Supabase—and what changes when the agent becomes the user?

Why do AI agents keep choosing Supabase—and what changes when the agent becomes the user? Supabase co-founders Paul Copplestone (CEO) and Ant Wilson (CTO) join swyx for their first podcast together after a $0 to $10B speedrun as the preeminent open source, fully remote Postgres platform company. We have watched them grow from a humble Firebase alternative launched during Covid, to now being a full preferred backend for AI-built applications: long-running compute beside the database, declarative schemas, Postgres storage and reliability, Turso, developer memes, community-led distribution, and product-led sales. The episode opens with swyx’s commentary on Supabase Select and the company’s journey from zero to a $10B valuation in six years as an open-source, fully remote company. swyx also discusses his seed investment in Supabase and its sponsorship of Latent Space. Chapters 0:00 Welcome and why agents love Supabase 5:23 Paul Copplestone and Ant Wilson 6:39 Compute beside the database 8:05 OrioleDB and Multigres 12:50 Scaling up, scaling out, and Turso 17:53 Developer marketing and memes 23:15 The launch and the migration 31:03 Vectors, growth, and AI recommendations 41:43 Product-led sales and enterprise growth 49:38 Live migrations and analytics 58:43 Agent-ready local development 1:06:43 Community and company workflows Links and show notes Supabase: https://supabase.link/AYNPC9B Supabase Select: https://select.supabase.com/ Build from code and Supabase Compute: https://supabase.com/blog/select-2026-build-anything Operate with your own agent: https://supabase.com/blog/select-2026-operate-with-confidence Multigres and OrioleDB: https://supabase.com/blog/select-2026-scale-without-limits Turso acquisition: https://supabase.com/blog/supabase-is-acquiring-turso Declarative schemas: https://supabase.com/blog/declarative-schemas-2-0-pg-delta Alpha launch postmortem: https://supabase.com/blog/alpha-launch-postmortem Supabase’s Series F: https://supabase.com/blog/supabase-series-f Latent Space: https://latent.space/

Nguyên văn nguồn
Chi tiết nguồn
Kỹ thuật

3 Questions: Why infrastructure is entering a defining era

Câu hỏi The question then becomes: How do we use the remaining resources more efficiently?

<p><em>For centuries, infrastructure has evolved alongside technology. New materials and engineering techniques enabled taller buildings, longer bridges, and larger cities. Computing and digital technologies later transformed how infrastructure was designed, monitored, and operated.</em></p><p><em>Now, the pace of change is accelerating, and challenges are arriving simultaneously. Climate change is increasing pressure on water, energy, and other resources. Artificial intelligence is beginning to transform engineering and design, and its energy needs, along with aging grid infrastructure, is creating significant challenges, as well as new opportunities. In addition, geopolitical complexity is exposing vulnerabilities in global supply chains, and demand for infrastructure is rising, particularly in the developing world.</em></p><p><em>The result is a fundamental challenge: How can the world build the infrastructure it needs without simply repeating the extractive, carbon- and resource-intensive model of the 20th century?</em></p><p><em>In this interview, Ali Jadbabaie, JR East Professor and head of the MIT Department of Civil and Environmental Engineering, says the answer will require rethinking not only how infrastructure is designed and built, but how it is financed and operated, and bringing together the many stakeholders who shape those decisions. On Oct. 28, Jadbabaie will join experts from around the world in Madrid to explore these topics at the&nbsp;</em><a href="https://www.worldinfrastructureforum.org/" target="_blank"><em>World Infrastructure Forum</em></a><em>, which was co-founded by MIT’s Department of Civil and Environmental Engineering, UPM, the World Bank Group, and the University of Texas at Austin.</em></p><p><strong>Q: </strong>Infrastructure has always evolved in response to technological advances. What makes this era different?</p><p><strong>A:</strong> In the 18th century, civil engineering emerged as the civilian counterpart to military engineering, building the dams, bridges, and canals of its time. Industrialization in the 19th century brought railways and sewage systems. In the early<sup>&nbsp;</sup>20th century, the focus scaled up to buildings, highways, tunnels, and metro systems, and by mid-century the infrastructure needed to support megacities and skyscrapers.</p><p>Every time technology has changed, with information technologies, computing, sensing, and, more recently, digital transformation, it has changed the shape of infrastructure. It has even changed the definition of what infrastructure is. Today, we consider the power grid, semiconductor manufacturing plants, and data centers part of infrastructure, along with wireless infrastructure like cellphone towers and the infrastructure for high-speed internet, just as much a part of our critical infrastructure as roads, water systems, and power grids.&nbsp;</p><p>What is different now is the combination and speed of the changes we're facing. Climate change, resource constraints, and geopolitical instability are putting pressure on the way we designed infrastructure in the 20th century. At the same time, many people are moving out of poverty because of recent economic growth in the developing world. In fact, according to models, by 2028, for the first time in the history of humanity, the median human inhabitant of the planet will live off $10 a day. As more people move out of abject poverty, there will also be an increase in the demand for infrastructure.</p><p>If we try to meet that demand using the same resource-intensive model we used in the 20th century, we simply won't have the resources to do it. The question then becomes: How do we use the remaining resources more efficiently? We need to find a way to provide more infrastructure while using less water, energy, and materials, and produce less carbon.</p><p>This is where digital transformation and AI become potentially transformative. Faster and cheaper sensing, greater access to data, and better sensing, actuation, and control can help us make infrastructure decisions more efficiently. AI is slowly permeating the physical world, and we expect it to revolutionize the engineering design process, allowing us to explore many more possibilities, much faster and at a lower cost.&nbsp;</p><p>But the story of that hasn't been written yet, and the outcomes could still be very positive, depending on what we do.</p><p><strong>Q: </strong>Can you give an example of infrastructure systems becoming more interconnected in a way that changes how we build or manage them?</p><p><strong>A:</strong> The revolution in information technology and digital transformation over the past 30 years has made infrastructure systems more interconnected. You see this in cities. If the power goes out, water and gas systems can stop functioning, internet fails, and transportation system comes to a halt. If communications networks fail, many other services are affected. We saw this during Covid, when disruptions in supply chains affected energy networks, critical materials, and manufacturing. We see it when geopolitical conflicts disrupt oil and energy markets and those disruptions propagate through other parts of the global economy.</p><p>Infrastructure systems that were once managed separately are increasingly dependent on one another. As we connect more things together for more efficiency, we also increase systemic risk and expose the systems to rare but catastrophic failures. They have become rare, but when they happen, they're much bigger and more consequential.</p><p>The really tricky thing is that developing infrastructure involves much more than designing and building it. The whole value chain includes financing and operating it, as well as understanding who will use it and what those users need and prefer.</p><p>So, designing infrastructure is a highly transdisciplinary, sociotechnical endeavor, involving social, technical, and economic considerations. Understanding how these sys

MIT News: AI · Stephanie Martinovich | Department of Civil and Environmental Engineering Mở bản gốc ↗
Nguyên văn nguồn
Chi tiết nguồn
Huấn luyện

Modal CEO on Agent Maxxing, the Future of Inference & Unpacking the Chaos of the GPU Market

Jacob Effron and Pat Chase sit down with Erik Bernhardsson, co-founder and CEO of Modal, for his second appearance on Unsupervised Learning, now with the company past a $1B run rate. They dig into how the AI compute market actually works behind the scenes, why Modal’s early bets on sandboxes and low-level infrastructure paid off, and Erik’s takes on NeoClouds, NeoLabs, and where GPU prices are headed next. 0:00 Intro 1:26 The Compute Market Today 5:16 Modal’s Path to LLM Inference 10:02 Building the Hard Infrastructure 12:32 Unsolved Problems in Inference 16:07 Chips & Model Sizes 21:21 What Wins in Inference 24:56 Prioritizing a Multi-Product Roadmap 30:20 Agents as Modal’s Users 33:19 The Next Big Bet After Sandboxes 38:58 Robotics & Multimodal Inference 41:04 Scaling From $2M to $1B 42:51 NeoClouds & NeoLabs 48:15 RL as a Service 51:40 Quickfire Jacob is an AI investor at Redpoint Ventures. He's led Redpoint's investments in companies like Abridge, Physical Intelligence & Legora. Follow Jacob on Twitter: https://x.com/jacobeffron Pat is an AI investor at Redpoint Ventures. He’s led Redpoint’s investments in companies like Modal, Serval & Hex. Follow Pat on Twitter: https://x.com/patrickachase On Unsupervised Learning we probe the sharpest minds in AI in search for the truth about what's real today, what will be real in the future and what it all means for businesses and the world. If you're a builder, researcher or investor navigating the AI world, this podcast will help you deconstruct and understand the most important breakthroughs and see a clearer picture of reality. Subscribe to this show to stay up to date on our latest episodes.

Unsupervised Learning · Unsupervised Learning: With Jacob Effron Mở bản gốc ↗
Nguyên văn nguồn
Chi tiết nguồn
Kỹ thuật

European Defense Sovereignty Starts in the Factory: Stark’s Uwe Horstmann and Johannes Schaback

Two years after its founding, Stark has roughly 700 employees and strike drones in the field in Ukraine. CEO Uwe Horstmann and CTO Johannes Schaback argue that the real lesson of that war is not any single weapon: the innovation ecosystem behind it is the wonder weapon. They explain why Stark designs every system for mass manufacturing from day one, runs 10 to 15 programs like a venture portfolio, and views the staff sergeant who operates a system as just as important as the general who buys it. Johannes recounts the publicly reported failed trial that became a forging moment, as engineers worked around the clock at a freezing test range until they landed three consecutive hits. They discuss NATO’s principles of responsible AI use and the hard-coded constraints in Stark’s autonomy stack. And Uwe lays out his next big idea: defense contracts structured like cloud contracts with surge capacity built in, so production itself becomes deterrence. Hosted by Pat Grady and Luciana Lixandru, Sequoia Capital Chapters 0:00 – Introduction 1:03 – Billions in contracts in 18 months: Stark's four-part playbook 2:28 – The staff sergeant matters as much as the general, and wood beats plastic 7:42 – Why Europe is ready for a neo prime, and why procurement reform will last 11:45 – Startup DNA, venture-style programs, and "as complex as a washing machine" 16:27 – Born out of fire: the failed trial and the round-the-clock comeback 21:13 – Where Stark finds talent, and why defense is no longer off-limits 23:48 – Anatomy of a strike, from the wooden crate to the kill zone 27:22 – Vertical integration, multiple domains, and shifting left on manufacturing 32:53 – After the Hugging Face hack: guardrails on autonomous weapons 35:56 – Matching the primes on R&D and scaling to 700 people in two years 39:24 – Deterrence by contract, and Europe's next chapter with the US

Nguyên văn nguồn
Chi tiết nguồn
Kỹ thuật

v0.33.0rc1: [CI] Fix stale NVFP4 decode test selector on DGX Spark (#64107)

<p>Signed-off-by: grYe99 <a href="mailto:guorongye99@gmail.com">guorongye99@gmail.com</a><br> Co-authored-by: grYe99 <a href="mailto:guorongye99@gmail.com">guorongye99@gmail.com</a><br> (cherry picked from commit <a class="commit-link" data-hovercard-type="commit" data-hovercard-url="https://github.com/vllm-project/vllm/commit/bfa2a433453675b0c7abcb5df163fd5e59ffe2a3/hovercard" href="https://github.com/vllm-project/vllm/commit/bfa2a433453675b0c7abcb5df163fd5e59ffe2a3"><tt>bfa2a43</tt></a>)</p>

Nguyên văn nguồn
Chi tiết nguồn
501–520 / 562
Trang 26 / 29