AI news for builders and product teamsUpdated Nov 17, 2026, 19:01 UTC
AI news, straight from the source.
Releases, research, and ideas for developers and product teams. Updated every hour.
Across the feeds
Newest first
Anthropic launched OSS Scanner, a free opt-in service that uses its strongest models, including Claude Mythos, to generate vulnerability reports and candidate patches for open-source projects, delivered to maintainers without human review.

inDrive signed a memorandum of understanding with Pony.ai and Nvidia to test Pony.ai's Gen-7 robotaxis in Astana, Kazakhstan, starting in the first half of 2027, initially with a safety operator aboard.

A new MIT history paper by Benjamin Lindquist in the journal Past and Present traces AI hallucinations to mid-20th century engineers who deliberately injected randomness into algorithms to simulate creativity. Lindquist argues that suppressing hallucination will inevitably cost some of the surprising output those pioneers valued.

MIT researchers left 13 AI agents running on Claude Opus 4.7 unsupervised on a simulated desert island for 30 hours, and the agents invented their own games and challenges; when later given specific tasks, the agents that had explored generally outperformed identical agents that had not.

Anima is using GPT-6 and GPT-Live to cut patient call wait times by 75%, according to OpenAI, giving care teams more capacity to respond to patients.

OpenAI and NASK, Poland's national research institute, announced a partnership covering AI safety, security and technical research aimed at strengthening safeguards and a more resilient digital ecosystem.

PyTorch Conference North America 2026 drew more than 3,500 attendees to San Jose on October 20–21, where the PyTorch Foundation and partners announced new hardware backends including TorchTPU for Google TPUs and TorchNeuron for AWS Trainium, plus nine new member organizations.

Microsoft and Hugging Face released ThinkingBox-Training, an open-source stack that reuses the same executable database-state checks used for grading agents as RL rewards. Fine-tuning Qwen3.8-27B with GRPO for 50 steps raised pass@1 from 51.70% to 60.78% and tasks solved on all 20 attempts from 38 to 107.

OpenAI is reportedly raising up to $30 billion at a $1.25 trillion pre-money valuation, led by Coatue and UAE investors, ahead of a planned IPO. The terms are still under negotiation and could change.

Canva has launched Canva Craft, its first in-house AI image model, which uses a two-stage architecture to plan a design before generating the image, supporting up to 2K resolution and more than 30 aspect ratios.

A sponsored MIT Technology Review piece from EXL argues that enterprise AI competitiveness increasingly hinges on the serving harness—architecture, memory management, scheduling and optimization around the model—rather than model choice alone. It urges teams to design from workload constraints and measurable production economics.

SpaceX and Tesla have formed special committees to discuss a possible merger, according to The Verge, a move tied to Elon Musk's stated ambitions around space travel and consciousness.

Simon Willison flagged De Staat's new music video "Freedom," where a visual effect at the 00:44 mark appears to involve generative AI, credited to Matties Grooten of Fake it or Leave it.

Databricks has published a piece on AI cost management, framing enterprise AI adoption over the past two years largely in terms of the spending and efficiency questions that come with running models in production.

1Password has launched 1Password for Agents, a tool that lets AI agents sign into a user's accounts while the password manager fills in credentials the agent never sees, with the user confirming each login attempt.

OpenAI has made its Decisions API generally available, using a reconfigured ChatGPT Luna to cut response times from one to three seconds down to roughly 100–180 milliseconds and compete with TypeSafe's Jev in the emerging decision-model market.

Canva released its first in-house image generation model, Canva Craft, available to all users through the Dream Lab tab, with support planned for the Canva API and third-party chatbots via Canva's MCP.

Jing Wang, principal at Vanguard's Center for AI and Insights, is the featured speaker in an MIT Technology Review EmTech Future session on building scalable AI foundations in a regulated financial environment.

Microsoft's MAI-DxO diagnostic system correctly identified 83% of complex simulated cases versus 36% for individual generalist physicians, per a study in Science, though much of the gain traces to the underlying AI models rather than the orchestration layer.

Databricks published a piece describing a "Fraud Analyst AI Coworker" for detecting telecom fraud in real time on a single platform, citing the Communications Fraud Control Association's 2025 survey as background.