Every story we have kept, newest first.
Looking for the daily editions? → Past editions
Page 92 of 1373
Dreame is reportedly shutting down its automotive division, Project Starry Sky, after government funding dried up. The project once employed more than 1,000 people, but only a small legal and HR team reportedly remains to wind it down.
The Verge · 23d ago
This LangChain guide shows how to build multimodal RAG applications for slide decks using GPT-4V. It compares implementation approaches, evaluates them with benchmarks, and explains deployment through LangChain templates for visual question answering.
LangChain Blog · 23d ago
OpenAI says its new Jalapeno chips performed better than Nvidia’s current lineup, while Nvidia prepares to report earnings. Apple also announced upgraded Mac mini and Mac Studio desktops with major processor improvements.
Bloomberg Technology · 23d ago
Autonomous trucking startup Gatik has raised $200 million to transition from technology development to commercial-scale operations. The company plans to deploy hundreds of driverless trucks by year-end, with North America remaining its primary market.
Bloomberg Technology · 23d ago
Apple is upgrading the Mac mini and Mac Studio as demand increases for desktop computers capable of running AI applications locally. Memory and silicon shortages are contributing to higher prices for the new systems.
Bloomberg Technology · 23d ago
Nvidia is expected to report another quarter of exceptional growth, but investor sentiment may depend on more than headline financial results. Franklin Templeton portfolio manager Sara Araghi says Nvidia needs to clarify its expanding investments across the AI ecosystem and address concerns about circular financing.
Bloomberg Technology · 23d ago
Lyft used LangGraph and LangSmith to build a self-serve AI agent platform for customer support. The platform reduced agent development timelines from months to weeks.
LangChain Blog · 23d ago
The article introduces context engineering as the practice of building dynamic systems that give LLMs the right information, tools, and format. Its goal is to help models complete tasks more reliably.
LangChain Blog · 23d ago
Emerald AI has raised $150 million in an oversubscribed Series A at a $1.05 billion valuation. Its software helps data centers reduce electricity use on demand, bringing the company's total funding above $220 million.
The Next Web (TNW) · 23d ago
LangSmith provides tracing and analysis capabilities for debugging complex deep-agent workflows. Developers can inspect traces, optimize prompts with Polly, and improve agent performance.
LangChain Blog · 23d ago
Samsung unveiled LPDDR5X-PIM, an LPDDR5X memory product with logic integrated directly into the memory. Samsung claims it delivers 3.01x faster AI inference and 8x the bandwidth of standard LPDDR5X.
Tom's Hardware · 23d ago
LangChain Structured Tools enables agents to work with tools that accept multiple inputs. Developers can define complex tool schemas to support more capable and flexible AI agent workflows.
LangChain Blog · 23d ago
LangChain presents an implementation guide for using a Multi-Vector Retriever with tables, text, and images in RAG systems. The accompanying cookbooks cover retrieval over semi-structured and multimodal data.
LangChain Blog · 23d ago
Stability AI raised $76 million in a Series B round backed by Universal Music Group, Warner Music Group, Sony Music Group, Electronic Arts, AMD Ventures, and Pacific Alliance Ventures. The investment gives three major record companies equity stakes in the generative AI company.
The Next Web (TNW) · 23d ago
Alphabet旗下的自動駕駛公司 Waymo 計劃在德國慕尼黑推出無人駕駛出租車服務。這項計畫延續了 Waymo 數月前在德國設立實體後的歐洲市場布局。
cnBeta (Full RSS) · 23d ago
NVIDIA 在 2026 年科隆遊戲展期間推出採用第二代 Transformer 模型的 DLSS 4.5 光線重建技術。該功能已透過 NVIDIA App 開放早期體驗，支援 GeForce RTX 20 系列及後續所有 RTX 顯示卡。
cnBeta (Full RSS) · 23d ago
Deep Agents introduces RubricMiddleware, which adds a self-evaluation loop to agent runs. Developers can define a rubric and configure a grader to improve reliability on tasks where correctness matters.
LangChain Blog · 23d ago
OpenAI presented benchmarks claiming its 700W Jalapeño ASIC outperforms Nvidia’s 1,400W GB300 flagship GPU in efficiency and latency. Co-developed with Broadcom, the chip reportedly delivers up to 1.9x higher throughput per kilowatt and 3.6x lower latency.
Tom's Hardware · 23d ago
LangSmith Engine now detects agent issues more than twice as effectively and proposes stronger fixes. The update also adds Slack and Linear workflows and supports self-hosted deployments.
LangChain Blog · 23d ago
LangChain explains how to use agent skills with the Deep Agents CLI to build token-efficient AI agents. Skills can be discovered, loaded, and executed dynamically.
LangChain Blog · 23d ago
.png)
LangChain presents an agentic engineering architecture that mirrors real software teams through coordinated AI agent swarms. Built on LangGraph, the approach reportedly cuts debugging time by 93% and compresses cross-team delivery.
LangChain Blog · 23d ago
Republican Senate candidate Ken Paxton has proposed a “Texas first” plan to address the perceived negative impacts of datacenters. The announcement comes amid growing political backlash and featured an X graphic flagged as “Made with AI.”
The Guardian Technology · 23d ago
Anthropic is adding shared memory between Claude chat and Cowork. Users can now preserve project details, preferences, and other context instead of repeatedly briefing the AI.
TechCrunch AI · 23d ago
McKinsey reports that enterprise AI investment is increasing and may finally be moving toward measurable return on investment. However, reported effects on enterprise earnings remain largely flat, highlighting a persistent gap between spending and realized business value.
The Register - AI/ML · 23d ago
LangChain explains why dependable agents require a task-specific harness rather than relying on model quality alone. The post examines the core agent loop, shows how layered loops improve capability, and discusses instrumentation with LangChain primitives.
LangChain Blog · 23d ago
An exploratory study by OWASP LLM security leaders found that prompt injection ranks first in expert assessments but only 12th in a record of 6,639 labeled real-world incidents. The gap likely reflects poor visibility, since prompt injection exploits model behavior and legitimate tool permissions rather than conventional software flaws.
VentureBeat · 23d ago
OpenAI says its new Jalapeno chips outperformed Nvidia’s current lineup in testing. The chips reportedly led in AI workload per unit of power and response speed.
Bloomberg Technology · 23d ago
Former Google researchers have founded a new nonprofit focused on keeping humans at the center of AI development. The organization aims to promote safer AI and reduce the risk of systems escaping human control.
Bloomberg Technology · 23d ago
Anthropic has expanded Claude's memory so it works across both chats and Cowork sessions. The company is also giving users more control over which memories Claude saves.
Engadget · 23d ago
Anthropic is merging the memory systems of Claude chat and Cowork, allowing information remembered in one product to be used in the other. The change is enabled by default, though users can opt out, raising questions about privacy and the benefits of persistent AI memory.
ZDNet AI · 23d ago