Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
ASML is committing 1.3 billion euros to Mistral's 1.7 billion euro fundraise, making it set to become Mistral's top shareholder. It is expected to get a board seat at Mistral. The funding round will make Mistral the most valuable AU company in Europe. The cash could help Mistral make Europe less reliant on US and Chinese AI models.
OpenAI has raised its projected cash burn through 2029 to $115 billion, $80 billion higher than the company previously expected. It will burn more than $8 billion this year, $1.5 billion more than it projected earlier this year. The company plans to develop its own data center server chips and facilities to power its technology and control its soaring costs. Its first chip, developed in partnership with Broadcom, will only be used internally and won't be made available to customers.
Hallucinations persist because current evaluation methods reward wild guessing that might occasionally be correct over admitting a lack of knowledge. The fix is simple: calibrate rewards during training based on certainty, penalize confidently wrong errors more, and give partial credit when models hedge or express doubt.
Building a popular AI safety movement risks undermining existing organic public support by making it appear coordinated rather than grassroots. What has worked so far is authentic opposition to specific policies and elite-focused expert advocacy. The author argues that broad movements inevitably get captured by adjacent political causes like labor or environmental issues, diluting safety-specific goals while exposing credible organizations to guilt-by-association with more extreme protesters.
The era of rapid compute scaling that drove AI progress since 2020 is over. GPUs from large cloud providers are almost entirely saturated, and frontier labs are now racing to bring new data centers online and reserve as much chip production as they can. Scaling, and therefore investment, is hampered by physical reality and the years required to build more infrastructure, meaning incremental scaling is the only option ahead, despite headlines of massive upfront funding rounds.
Sponsored
Use this practical guide [12] to design your AI governance program. Learn how to build on AI principles aligned with global frameworks like the EU AI Act and NIST AI RMF, establish clear policies and oversight, and embed governance in daily workflows. Get inspiration from OneTrust's real-world approach to risk, collaboration, and
NVIDIA and AMD have been racing to solve data movement at the cluster scale to create a software moat. Voltron Data's Theseus is the first distributed runtime to put movement at the center of scheduling, memory, and I/O, and then prove the payoff with benchmarking results. It is open, composable, and already running on CUDA and ROCm. Theseus runs across both Nvidia and AMD ecosystems for datacenter-scale analytics and AI.
Dozens of open source projects are currently testing different recipes for how to work with AI productively. These efforts show that AI works best when it is given structure. The frameworks are converging on a future where AI becomes a set of teammates to be managed. The more structure you give, the more you get back.
Alibaba has introduced Qwen3-Max-Preview, a 1 trillion-parameter model now accessible via Qwen Chat and Alibaba Cloud. Early benchmarks highlight improvements in instruction following, dialogue, and agentic task performance.
The Agentic Systems Series is a complete guide for building AI coding assistants that actually work in production. It covers everything from fundamental concepts to implementing enterprise-ready collaborative systems. The series provides deep analysis of real production systems, including Amp, Claude Code, and open-source implementations like anon-kode. It is assumed readers are familiar with system design concepts and have a basic understanding of AI/LLM integration and experience with either TypeScript/Node.js or similar backend technologies.
SEO jobs are changing to include the distribution of content. They now focus more broadly on the discoverability of a website. Focus on methods that drive brand awareness growth. Pivot to tracking LLM visibility, structure content for LLM digestibility, change content approaches, and think about the future of conversion to prepare for the next few years.
Companies like Anysphere face soaring LLM inference costs, forcing price hikes and prompting exit strategies like reverse acqui-hires. The coding AI market's rapid growth, driven by reasoning models, is now pressured by expensive compute costs, leading vendors to adopt usage-based pricing.
Sponsored
Following several public incidents regarding ChatGPT's personality, OpenAI is absorbing the Model Behavior team into the larger Post Training group, suggesting that personality is now considered a core capability to measure and hone during model development.
GPT-5 Thinking leads to far more comprehensive results, but it's a lot slower.
Anthropic's researchers have embraced moderate accelerationism, advocating for measured and responsible AI advancement.
Starting January, Medicare will implement an AI pilot program in six states to assist in making coverage decisions through the WISeR Model.
Moonshot AI has released Kimi-K2-Instruct-0905, a mixture-of-experts (MoE) model featuring 32B activated parameters out of 1T total.
Sponsored
In today's AI boom, shipping fast gets attention—but building trust gets results. Startups now face higher expectations earlier than ever, with investors demanding clean financials and enterprise buyers requiring strong security before deals close.
OpenAI has introduced a new hiring platform powered by AI to match companies with AI-fluent talent and launched OpenAI Certifications to validate AI skills.
Atlassian is seeking to create an AI-powered browser optimized for knowledge workers, building off its existing suite of productivity SaaS applications. The Browser Company will continue developing its upcoming Dia browser under independent operations.
The Chinese startup is developing a multi-step AI agent software similar to OpenAI's Agent tool that can handle complex tasks without human supervision.
AI Artists give AI systems full creative control, while AI engineers operate under constraints to optimize for quality. The distinction between the two is a spectrum, and most people fall somewhere in the middle, closer to one side or the other. While large language models are very powerful fools, in the absence of the ability to do fine-grained learning in the way a human does, they need stricter guardrails than a truly general-purpose system would ideally want. Companies should focus on building trust with customers by doing well-scoped, expected things.
Users tend to abandon agents after their first real problems. The real challenge for product managers is making architectural decisions that shape how users experience and begin to trust the agent. This post walks through the different layers of AI agent architecture to help readers understand why some agents feel magical while others feel frustrating. Product managers should architect for the magical experience.
Sponsored
From smarter audience segmentation to automated media buying, real-time optimization, and precise performance measurement, artificial intelligence is reshaping every stage of the media campaign lifecycle. Read about what's coming next in OneTrust's State of Data 2025 [10] and the actionable Companion Guide [10].
Robix is a unified vision-language model for robots that combines reasoning, task planning, and natural language interaction in one framework. It enables robots to handle complex, long-horizon tasks and communicate naturally with humans through verbal responses and proactive dialogue.
Google's EmbeddingGemma is a compact, multilingual embedding model with 300M parameters optimized for on-device performance.
OpenAI and Broadcom have partnered to produce an artificial intelligence chip. The chip will be used internally at OpenAI and won't be made available to external customers. OpenAI's move follows efforts by Google, Amazon, and Meta, which have all built custom chips to handle AI workloads. Broadcom expects AI revenue growth for fiscal 2026 to improve significantly.
Sierra announced raising $350M at a $10B valuation to expand its AI agent platform, backed by Greenoaks. The platform supports hundreds of major companies, reaching 90% of Americans in retail and 50% of US families in healthcare. New funding will enhance AI development capabilities and international expansion, strengthening Sierra's global presence.
Sponsored
You'll probably like our flagship newsletter. It's all about tech, science, and programming.
Google Photos has integrated its latest video generation model, Veo 3, into the new Create tab, enhancing its image-to-video capabilities.
Jules can now reply to PR comments, create new repositories, and receive files.
Sponsored
Metronome interviewed leaders across pricing, product, and ops to understand how AI is reshaping SaaS monetization [4]. For many teams, the focus for pricing AI features isn't to boost revenue. What they're doing is pricing to avoid financial disaster:
Mistral AI is finalizing a €2 billion investment that will boost its valuation to $14 billion. Founded by ex-DeepMind and Meta researchers, it develops open-source language models and the European AI chatbot Le Chat. This marks Mistral's first major fundraising since June 2024.
xAI has seen a number of high-profile departures in recent weeks. Mike Liberatore, xAI's chief financial officer, left the company at the end of July. Robert Keele, xAI's general counsel, announced he was leaving on August 7. Raghu Rao, a senior lawyer who oversaw commercial legal affairs for xAI, left around the same time as Liberatore and Keele. Igor Babuschkin, a veteran of Google's DeepMind AI research lab and OpenAI, announced on August 13 that he was leaving to start his own venture-capital firm focused on AI safety.
Apple plans to enhance Siri with an AI search feature, potentially relying on Google's Gemini AI model. The feature, dubbed “World Knowledge Answers,” will generate summaries using web results, integrating various media formats. Apple has agreed with Google to test Gemini, while also considering its own and Anthropic's models for Siri's development.
Rather than nationalizing AI companies or creating new international bodies, we should create publicly-traded AGI corporations with stricter government oversight. Foreign countries and citizens would purchase voting stock in companies like OpenAI, giving them financial stakes and board influence instead of racing to build competing systems.
While general methods that leverage computation seem more effective, data still plays a very important role. You need 40% more data when doubling GPUs, otherwise you're just lighting cash on fire. AI companies have already consumed the internet. The leaders who win the next era of AI will be scaling data itself. This could be done by architecting better use of limited data or alchemizing entirely new ultimate data universes.
AI companies haven't actually done a good job of scaling up RL - they've scaled up to compute, but with low-quality data. While some argue that once these companies actually scale RL up, there will be a big jump in AI capabilities, it is more likely that progress will be the result of numerous hard-won advances. Higher quality RL environments will be a substantial part of what drives progress over the next year or two, and maybe beyond that, but significant improvements will come from several advances, not just scaling RL.
Sponsored
Teams can't see into their AI models - and many can't even answer basic questions like "which prompts are costing us the most money?" AI observability [11] is the way to stop guessing and starting getting answers - with automated dashboards for prompt frequency, response times, drift indicators, and cost impact. Try it out in the Dynatrace
Frontier models can write optimized GPU kernels for Apple devices to speed up inference. This post discusses how researchers generated Metal kernels that were 1.87 times faster across 215 PyTorch mod duties, with some workloads running hundreds of times faster than baseline. The results show that AI can take on portions of optimizations, leaving human engineers to focus on the most complex optimizations.
A curated list of papers on agentic reinforcement learning with large language models.
In this 400 page preprint, a senior Google exec lays out the 21 fundamental design patterns for AI agents. They provide concrete code examples of prompt chaining, memory implementation, multi-agent coordination, and more from production applications. While it's still very early into agentic tooling, the core engineering foundations for the post-chatbot generation of AI are taking shape.
OpenAI is allowing current and former employees to sell more than $10 billion worth of stock in a secondary share sale. The sale will be at a $500 billion valuation, which is in line with expectations. OpenAI was valued at $300 billion in a fundraising round earlier in the year. Staff who have held shares for more than two years have until the end of September to decide whether to participate.