AI News on June 16, 2026

Open Source

Meituan Unveils LongCat-Next: Open-Sourcing a Native Multimodal Model for Physical World AI

Meituan's technical team has announced the release and open-sourcing of LongCat-Next, a native multimodal model designed to bridge the gap between artificial intelligence and the physical world. By treating vision and speech as "native languages," the model aims to fundamentally enhance how AI perceives, understands, and interacts with its environment. Alongside the core model, Meituan has open-sourced its discrete tokenizer, providing the global developer community with the essential infrastructure to build sophisticated AI systems capable of real-world action. This move represents a strategic milestone in Meituan's exploration of embodied AI, focusing on the seamless integration of multiple sensory inputs to create more intuitive and functional artificial intelligence that can operate beyond digital constraints.

美团技术团队•June 16, 2026

Read AnalysisSource

Industry News

Meituan's Breakthroughs at ACL 2026: Redefining Generative Paradigms through Evaluation and Reasoning Optimization

Meituan's technical team has achieved a significant milestone at ACL 2026, the premier international conference for computational linguistics and natural language processing. With six papers accepted, Meituan's research spans critical frontiers including large model evaluation, complex process reasoning, competition-level mathematical thinking optimization, reinforcement learning, and generative recommendation systems. These contributions highlight a strategic shift toward building a new generation of AI paradigms that emphasize both the robustness of model assessment and the depth of logical reasoning. By addressing high-level challenges such as mathematical problem-solving and the evolution of recommendation engines, Meituan is bridging the gap between theoretical academic research and practical industrial application, setting a new standard for generative AI development.

美团技术团队•June 16, 2026

Read AnalysisSource

Open Source

Meituan Open-Sources LongCat-Video-Avatar 1.5: Bridging the Gap Between Research and Commercial Digital Human Applications

Meituan's technical team has officially announced the open-source release of LongCat-Video-Avatar 1.5, a digital human video model that marks a significant transition from experimental State-of-the-Art (SOTA) performance to practical, commercial-grade utility. This update introduces comprehensive improvements across five critical dimensions: lip-synchronization, physical plausibility, long-video stability, multi-person interaction, and inference efficiency. By addressing the limitations of previous experimental models, LongCat-Video-Avatar 1.5 is designed to deliver stable, natural, and high-quality content even within complex commercial environments. The release signifies a strategic move to transition digital human technology from controlled "rehearsal" settings to the "real stage" of diverse, real-world applications, providing a robust and scalable solution for the industry.

美团技术团队•June 16, 2026

Read AnalysisSource

Industry News

Meituan LongCat Team Launches General 365: A New Benchmark Revealing AI Reasoning Limitations

The Meituan LongCat team has officially released General 365, a new evaluation benchmark specifically designed to measure the reasoning capabilities of large language models. In an extensive test involving 26 mainstream models, the benchmark has highlighted a significant performance gap in the current AI landscape. According to the results, Gemini 3 Pro emerged as the top performer but only managed an accuracy rate of 62.8%. Strikingly, the vast majority of the tested models failed to reach the 60% threshold, which is typically considered a passing grade. This development suggests that while AI has made strides in general tasks, complex reasoning remains a formidable challenge for even the most advanced systems currently available on the market.

美团技术团队•June 16, 2026

Read AnalysisSource

Industry News

Managing AI Coding with Agent Evaluation Logic: Lessons from a 310,000-Line AI Refactoring Project

As AI-generated code accounts for over 90% of system development, the primary challenge has shifted from production speed to the effective constraint of AI capabilities. Without unified standards, AI risks exponentially increasing system chaos. This analysis explores the practice of the Meituan technical team in refactoring 310,000 lines of code by applying Agent evaluation logic to AI coding management. By implementing a structured framework consisting of technical debt sorting, rule construction, Refactoring Standard Operating Procedures (SOPs), and Pre-PR mechanisms, the team successfully transformed high-cost refactoring into a continuous, iterative daily process. This approach ensures that AI-driven development remains orderly and sustainable, preventing the accumulation of unmanaged technical debt while maintaining high code quality across large-scale systems.

美团技术团队•June 16, 2026

Read AnalysisSource

Research Breakthrough

Meituan Technical Team Releases LARYBench: A New Standard for Evaluating Latent Action Representations in Embodied AI

The Meituan Technical Team has officially introduced LARYBench (Latent Action Representation Yielding Benchmark), a systematic evaluation framework designed to guide the learning of universal latent action representations from large-scale visual data. This benchmark represents a significant step in embodied AI, often compared to the 'ImageNet' for action representation. Experimental results released alongside the benchmark reveal that general-purpose vision models significantly outperform specialized embodied AI expert models in both action generalization and control precision. Furthermore, the research demonstrates that embodied action representations can successfully emerge from large-scale human video data, suggesting that specialized datasets may not be the only path toward developing sophisticated robotic control systems.

美团技术团队•June 16, 2026

Read AnalysisSource

Research Breakthrough

Meituan LongCat Team Unveils LongCat-AudioDiT: Revolutionizing Zero-Shot Voice Cloning via Waveform Latent Space Diffusion

The Meituan LongCat team has introduced LongCat-AudioDiT, a breakthrough model designed to push the boundaries of zero-shot Text-to-Speech (TTS) voice cloning. By fundamentally changing the traditional synthesis pipeline, the model bypasses intermediate representations such as Mel-spectrograms. Instead, it operates directly within the waveform latent space using a diffusion-based approach. This strategic shift aims to eliminate cascade errors typically introduced during data conversion processes. By allowing the AI to learn the inherent patterns of sound directly, LongCat-AudioDiT offers a more streamlined and accurate method for replicating voices without prior training on specific target speakers, marking a significant advancement in audio synthesis technology and addressing long-standing technical bottlenecks in the field of AI-generated speech.

美团技术团队•June 16, 2026

Read AnalysisSource

Open Source

Meituan Technical Team Open-Sources LongCat-Flash-Prover for Rigorous Mathematical Theorem Proving and Formalization

The Meituan Technical Team has announced the open-source release of LongCat-Flash-Prover, a specialized AI model designed to tackle the complexities of mathematical formalization and theorem proving. Unlike conventional AI models that prioritize reaching a correct final numerical value, LongCat-Flash-Prover focuses on the construction of rigorous logical chains. The model addresses a critical challenge in AI reasoning: the tendency for natural language ambiguity to undermine the validity of a proof. By shifting the focus from "guessing answers" to "rigorous proof," this initiative aims to enhance the capabilities of AI in handling complex reasoning tasks where precision and formal logic are paramount. The release marks a significant contribution to the field of automated reasoning and formal verification.

美团技术团队•June 16, 2026

Read AnalysisSource

Industry News

Meituan LongCat Team Open-Sources WBench: The First Systematic Multi-Round Benchmark for Interactive Video World Models

The Meituan LongCat team has officially introduced and open-sourced WBench, a pioneering evaluation framework designed to test the limits of interactive video world models. Positioned as the first systematic multi-round benchmark in its category, WBench functions as a diagnostic tool—likened to a "CT scanner"—to identify specific technical hurdles as AI transitions from passive video generation to active, interactive environmental simulation. By focusing on the boundaries between "passive viewing" and "active interaction," WBench provides a rigorous methodology for assessing how models maintain consistency across complex, multi-step scenarios. This open-source contribution aims to standardize the evaluation of world models, offering insights into their performance in diverse settings ranging from lunar landscapes to futuristic urban environments.

美团技术团队•June 16, 2026

Read AnalysisSource

Industry News

Meituan BI Evolution: Building a Metric-Centric Architecture for Enhanced Data Consistency and Performance

Meituan's Data Platform team has introduced a next-generation Business Intelligence (BI) architecture centered on a unified metric platform. This strategic shift addresses the inherent flaws of traditional BI systems, which often suffer from inconsistent data definitions and sluggish query performance due to their reliance on fragmented, personalized datasets. By implementing two core technical pillars—automatic semantics and enhanced calculation—Meituan has successfully streamlined its data analysis process. This new framework ensures that data "mouthpieces" (definitions) remain consistent across the organization while significantly boosting the efficiency of complex analytical queries, marking a significant milestone in the company's data engineering capabilities.

美团技术团队•June 16, 2026

Read AnalysisSource

Open Source

NVIDIA SkillSpector: A Dedicated Security Scanner for AI Agent Skills and Vulnerability Detection

NVIDIA has introduced SkillSpector, a specialized security scanner designed to identify and mitigate risks within the burgeoning ecosystem of AI agent skills. As AI agents gain autonomy through specialized 'skills'—modular capabilities that allow them to interact with tools and data—the potential for security breaches increases. SkillSpector aims to address these concerns by scanning for vulnerabilities, malicious patterns, and broader security risks. This release, hosted on GitHub, signals a significant step by NVIDIA to provide developers with the tools necessary to ensure the integrity and safety of agentic AI workflows. By focusing on the 'skills' layer, SkillSpector provides a targeted defense mechanism against exploitation in automated AI environments.

GitHub Trending•June 16, 2026

Read AnalysisSource

Product Launch

Meta Launches AI Mode Search on Facebook Utilizing Public User Posts for Results

Meta has officially introduced "AI Mode" for Facebook search, a new feature that leverages public user posts to generate AI-driven search results. Appearing alongside traditional search categories like "People" and "Marketplace," AI Mode is part of a broader suite of AI tools being rolled out, which also includes creative photo presets such as jersey-swapping capabilities. This update marks a significant shift in how Meta utilizes user-generated content to power its internal AI systems, providing users with a more integrated and generative search experience directly within the Facebook platform. The rollout begins today, signaling Meta's commitment to embedding advanced artificial intelligence into the core functionality of its social media ecosystem while utilizing existing public data to inform its models.

The Verge•June 15, 2026

Read AnalysisSource

Industry News

The Prospect of a Peopleless Economy: Analyzing the Technical Possibility of Total AI Replacement

In a provocative analysis, George Malandrakis explores the concept of a 'peopleless economy,' challenging the widely held belief that AI cannot fully replace the human workforce due to the necessity of consumption. Many assume that if AI replaces all workers, the economy would collapse from a lack of consumers; however, Malandrakis argues this may be a logical delusion. By examining the philosophical foundations of human logic, the author suggests that our economic theories are built on implicit, abstract axioms rather than concrete facts. The article posits that concepts such as 'Justice' and 'Money' are often ill-defined, leading to dubious logical conclusions. Ultimately, the text suggests that a peopleless economy is not technically impossible, as the fundamental assumption requiring human participation in the economic cycle may be flawed.

Hacker News•June 15, 2026

Read AnalysisSource

Industry News

The Emotional Connection to Computing: Why the 'I Love the Computer' Sentiment Resonates Amidst AI Hype

In a reflective piece inspired by the Aftermath Podcast, technologist Michael Enger explores the deep-seated passion for computing that stands in stark contrast to the current AI hype cycle. The article centers on a quote from editor Chris Person—'I love the computer'—which serves as a rallying cry against the 'snake oil salesmen' and 'insatiable avarice' currently perceived in the tech industry. Enger recounts his formative experiences in Norway during the early 1990s, where his journey began with an IBM 486 DX6 running Windows 3.0. This personal history highlights a time when technology was a daunting yet enthralling tool for discovery, rather than a vehicle for commercial exploitation. The analysis delves into the tension between genuine technological appreciation and the 'social crime' of modern industry trends.

Hacker News•June 15, 2026

Read AnalysisSource

Industry News

LinkedIn Job Offer Security Alert: Developer Discovers Hidden Backdoor in Malicious GitHub Coding Task

A developer recently exposed a sophisticated backdoor embedded in a GitHub repository shared by a recruiter on LinkedIn. The recruiter, purportedly representing a crypto startup, invited the developer to review a codebase to address "deprecated Node modules." By utilizing a secure VPS and an AI agent for inspection, the developer identified malicious code hidden within a test file. The script assembles a remote URL from fragmented strings to fetch and execute payloads from a command-and-control server. The attack is designed to trigger automatically through the "prepare" script in the project's package.json file. This incident serves as a critical warning for technical professionals regarding social engineering and the risks of running untrusted code from potential employers.

Hacker News•June 15, 2026

Read AnalysisSource

Industry News

Why South Korea Leads in AI Integration: From Unmanned Immigration to Daily Commutes

This analysis explores the pervasive nature of artificial intelligence in South Korea, as observed through the lens of Michelle Kim's recent arrival in Seoul. The report highlights the seamless transition from international travel to local life, facilitated by advanced automated systems. Key observations include the use of unmanned immigration checkpoints that utilize facial recognition and passport scanning technology, as well as the integration of AI within the public subway system. These developments suggest a societal infrastructure that is deeply intertwined with AI, prioritizing efficiency and automation in high-traffic public spaces. The article examines the implications of such widespread technological adoption and what it reveals about the daily experience in one of the world's most tech-forward nations.

MIT Technology Review - AI•June 15, 2026

Read AnalysisSource

Product Launch

Meta Launches ‘AI Mode’ on Facebook Using Cross-Platform Public Data to Boost Engagement

Meta has announced the rollout of a new 'AI Mode' and a suite of AI-driven features for Facebook. This strategic move is designed to leverage public information from across Meta’s various platforms to power its AI capabilities. The initiative serves two primary purposes: helping Meta catch up with competitors in the rapidly evolving AI race and increasing user engagement on its flagship social media platform. By integrating data from its broader ecosystem, Meta aims to provide a more sophisticated and interactive experience for Facebook users, signaling a significant shift in how the company utilizes its vast data resources to remain competitive in the modern technology landscape.

TechCrunch AI•June 15, 2026

Read AnalysisSource