Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
xAI is expanding into video game development. Elon Musk has confirmed that the company is hiring game developers and that it is starting a new game studio to 'make games great again'. Musk founded xAI in 2023 with the goal of understanding the true nature of the universe. The company has since developed a chatbot and partnered with X to power the conversational AI on the platform.
Waymo spent 16 years gathering data before going mainstream. Its robotaxis have driven 96 million miles with 91% fewer serious injury accidents than humans. It's a rare example of safety in AI and extreme caution in Silicon Valley: Tesla's Austin robotaxis crashed three times in 7,000 miles, and Cruise dragged a pedestrian 20 feet before GM killed the business. Waymo still limits rides to city zones despite years of successful tests on highways.
Sponsored
Mirage Studio helps you generate on-brand, expressive videos with AI actors. Videos look, sound, and feel real, and you can even add
Microsoft is using Nebius' datacenter facilities for its own internal teams, as it reserves its own internal datacenter infrastructure for renting out to new clients.
While Nvidia offers the best price-performance, Microsoft's CTO Kevin Scott says he's willing to entertain anything in order to meet demand.
ChatGPT is shifting toward more reliable, verifiable sources of information.
OpenAI called the lawsuit part of Musk's ongoing harassment and said talented xAI employees are leaving “in droves" with every right to work elsewhere.
Google's LearnLM team developed a system that adds adaptive layers to textbooks—like multimedia representations and personalization—without compromising content quality.
Sponsored
WorkOS [5] presents Enterprise Ready Conf: the flagship full-day event for product and engineering leaders shaping the future of Enterprise AI.
Thinking Machines has launched Tinker, a flexible API for fine-tuning open-weight language models. It abstracts infrastructure complexity, supports large MoE models like Qwen-235B, and comes with a low-level API and open-source Tinker Cookbook for implementing advanced post-training techniques.
The launch of Sora 2, a TikTok-style feed for AI-generated videos, prompted mixed reactions from OpenAI researchers, fearful of repeating social media's addictive patterns. Sam Altman defended the launch by claiming that consumer revenue will fund AGI research, but many internally and externally are skeptical that OpenAI can grow entertainment products without compromising its nonprofit charter.
Anthropic claims that Claude Sonnet 4.5 is the best coding model in the world. While there are still clear use cases where versions of GPT-5 seem likely to be better, Claude Sonnet does seem like the best coding model in the world right now. GPT-5 still seems to make better use of extended thinking time than Claude Sonnet 4.5 does. Unless you must have an open model or need to keep inference costs very low, there is no reason to consider anything other than Claude Sonnet 4.5 or GPT-5.
Claude Code is a terminal-based application that trades accessibility for native Unix command integration. The commands that power Unix happen to be perfectly suited for use by large language models. Having file system access allows Claude Code to write notes to itself, accumulate knowledge, keep running tallies, and think beyond a single conversation. Claude Code works as a blueprint for building reliable agentic systems because it captures model capabilities instead of limiting them through over-engineered interfaces.
Sponsored
Running AI on your PC is transformative for productivity - but like all emerging tech, it comes with some security risk. Learn about the tactics attackers use to gain entry [12] to your endpoints, and how
Slack is reshaping how AI agents access and use enterprise data with a new platform that allows developers to tap directly into the rich conversational data flowing through workplace channels. Its new real-time search API and Model Context Protocol server give third-party developers secure, permission-aware access to Slack conversations, messages, and files. Slack is in increasingly intense competition with Microsoft Teams. Both companies are embedding AI throughout their collaboration platforms while taking markedly different approaches. Slack is focused on becoming an integration hub for other software experiences, while Microsoft is building a comprehensive suite of productivity tools.
Tunix is a JAX-native library that makes the path from pre-trained model to a fully aligned, production-ready large language model simpler. It gives developers a comprehensive and developer-friendly toolkit for aligning models at scale. Tunix is built for performance on TPUs, especially when combined with MaxText. The initial release provides modular and easy-to-use APIs for the most common post-training workflows.
Google DeepMind has built a custom generative image model fine-tuned for prototyping capable of producing design outputs that reflect specific stylistic features while enabling exploration of novel concepts.
YouTube plans to leverage AI to transform video creation, enabling creators to use AI tools like DeepMind's Veo 3 for video enhancements. CEO Neal Mohan highlights the potential for AI-generated content to democratize creation, although it raises questions about authenticity. Despite concerns, YouTube remains committed to AI advancements, aiming to stay at the forefront of video innovation.
The deal gives Meta access to Nvidia's latest GB300 systems and offers much-needed diversification to CoreWeave, which relies heavily on Microsoft and OpenAI. CoreWeave shares jumped 15% on the news.
Sponsored
GenAI is now mainstream, and companies are placing an increased focus on security, data quality, and cost efficiency. Read the Google Cloud
The AI chip maker claims 20x faster inference speeds than Nvidia GPUs and serves trillions of tokens monthly to major companies and government agencies.
After receiving a cease-and-desist from Disney, Character.AI has removed various Disney-owned characters from its chatbot platform.
Samsung and SK Hynix say that OpenAI's anticipated demand could grow to 900,000 DRAM wafers monthly.
Sponsored
You've tried Cursor, Codex, and Claude Code. Now try the coding agent that beats them.
OpenAI introduced Sora 2, its next-generation video generation model with improved physical realism, finer detail, and greater control. The update supports synchronized dialogue, sound effects, and user interaction through a new app interface.
Periodic Labs emerged from stealth to build AI scientists that conduct physical experiments autonomously in robotic labs. Founders Ekin Dogus Cubuk (who led Google Brain's materials team that discovered 2 million new crystals with AI) and Liam Fedus (former OpenAI VP who helped create) aim to automate scientific discovery, starting with superconductors, and claim that LLMs need fresh data for physical experiments because they've exhausted the internet as a training source.
Meta is acquiring chip startup Rivos to strengthen its AI hardware capabilities and reduce dependence on external suppliers. The deal gives Meta more control over custom silicon, signaling a push toward deeper vertical integration in its AI infrastructure.
AI is fundamentally advanced by scaling, but AI researchers continue to work on algorithms, architecture, and data as if scaling laws were yet to be discovered. More compute and more energy is the most reliable path to advancing AI. Many labs are aiming for recursive self-improvement, but the notion that AI will algorithmically self-improve ad infinitum ignores scaling laws and that research is compute-bound. This makes problems like autonomous science a better problem for AI researchers to work on.
Quietly and all at once, frontier AIs crossed a threshold into economically valuable work. OpenAI tested experts against AI on multi-hour tasks, and humans won, but barely, with AI's weakness being formatting rather than accuracy. Claude Sonnet 4.5 reproduced an economics paper's findings in minutes, work that would take academics hours of tedious conversion, potentially solving academia's replication crisis at scale. The difference between transformation and waste lies entirely in human judgment about what's worth doing, not merely what can be done.
Sponsored
Leading enterprises are embedding AI at the core of their operations to unlock flexibility, free resources, and accelerate growth. This new model moves organizations from pilots to enterprise-wide impact, creating real competitive advantage.
Agents can now directly exercise the code they are writing, correct errors, dig through existing implementation details, and even run experiments to find effective code solutions to problems. Coding agents are brute force tools for finding solutions to coding problems. Reducing problems to clear goals and sets of tools that can iterate towards those goals can help coding agents brute-force their way to an effective solution. The art of using agents well is to carefully design the tools and loops for them to use.
Logicstar shrank SWE-Bench Verified from 240 GiB to just 5 GiB. It can now be downloaded in under a minute, making large-scale evaluation and trace generation on cloud machines fast and painless. The team achieved this result by restructuring layers, trimming unnecessary files, and compressing the results. This article gets into details about how the smaller SWE-Bench Verified was created.
Compute for pretraining grows 4× annually, but web data increases just 1.03× per year, forcing a shift toward more efficient learning algorithms. Regularization parameters 30× higher than standard practice prevent overfitting when models see the same data repeatedly. Ensembling independently trained models achieves lower loss than simply making models larger. The combination cuts data requirements by 5× to match baseline performance and directly translates to improvements on pretraining benchmarks.
Large AI developers must publicly publish safety frameworks, report critical incidents to the state, and face civil penalties for noncompliance. Anthropic endorsed the law, but Meta and OpenAI lobbied against it. The law establishes a public computing cluster, protects whistleblowers, and mandates annual updates.
Prime Intellect has released INTELLECT-2, a 32B parameter model trained via distributed reinforcement learning instead of traditional centralized GPU clusters. This breakthrough demonstrates that advanced AI models can be developed using a network of volunteer computing resources spread worldwide, potentially democratizing AI development by eliminating the need for massive co-located data centers that currently give tech giants their competitive advantage in the race for more powerful AI systems.
Sponsored
You'll probably like our flagship newsletter. It's all about tech, science, and programming.
Shallot is a lightweight system designed to help maintain clean, scalable interactions with Claude Code.
Google Search's AI Mode now includes a new visual exploration feature that helps users imagine, find, and shop for items through a more interactive, image-driven interface.
Robots often fail in real-world scenarios due to overfitting to specific locomotion strategies.
Semantic speech tokenizers often break under minor audio noise, even when speech remains intelligible.
for driving adoption and measuring ROI.
Sponsored
Employees switch between apps +1,200 TIMES PER DAY, killing productivity. This "toggle tax" is avoidable: AI + integrations can give users the information they need, without rummaging through 37 open tabs.