Weekly Digest — 2026-W23

Generate more accurate investment dashboards and reports

Local natural language search across all your files

Control your robots from anywhere in the world

A smart LLM proxy that shows where you're overpaying

Local AI workspace to build and work with your computer

Chat with your business data inside Claude, ChatGPT and more

If Gusto, OpenClaw, and Claude Cowork had a baby...

Build, manage, and ship customer messaging from Slack

A fun new way to create & manage brands.

Use iPhone to open apps, Apple Shortcuts and websites on Mac

Your Mac's camera is now a heart-rate sensor

Make your business citable by ChatGPT, Claude & Perplexity

Large Language Model (LLM) search agents have shown strong promise for knowledge-intensive language tasks through multiple rounds of reasoning and information retrieval. Most existing systems access information using a retriever that takes a keyword or natural language query and returns a ranked list of documents using an index of pre-computed document representations. In this work, we explore a complementary perspective in which the search agent treats the corpus itself as the search environment and finds evidence by issuing executable shell commands. We introduce GrepSeek, an optimized direct corpus interaction (DCI) search agent that trains a compact search agent to find, filter, and compose evidence from large text corpora. To address the instability of learning behavior directly with reinforcement learning on large corpora, we propose a two-stage training pipeline. First, we construct a cold-start dataset using an answer-aware Tutor and answer-blind Planner to generate verified, causally grounded search trajectories. Second, we refine the initialized policy with Group Relative Policy Optimization (GRPO), allowing the agent to improve its task-oriented search behavior through direct interaction with the corpus. To make DCI practical at scale, we further use a semantics-preserving sharded-parallel execution engine that accelerates shell-based retrieval by up to 7.6times while preserving byte-exact equivalence with sequential execution of the shell command. Experiments across seven open-domain question answering benchmarks show that GrepSeek achieves the strongest overall token-level F_1 and Exact Match. Our analysis also highlights the limitations of purely lexical interaction on queries with substantial surface-form variation, suggesting DCI as a practical and competitive method for search agents that can complement existing retrieval paradigms in the real world.

LLM agents are increasingly expected not only to complete isolated tasks, but also to carry bounded representations of human expertise, judgment, and interaction style. Building such person-grounded agents remains difficult because actionable knowledge associated with a person or role is usually embedded in heterogeneous traces rather than written as clean instructions. Existing memory and persona systems capture fragments of this evidence, while skill frameworks provide portable packaging formats; however, there is no end-to-end workflow for distilling these traces into inspectable, correctable, and agent-usable skills. We present an automated trace-to-skill distillation system for generating person-grounded AI skills via expert knowledge distillation. Given materials from a target person or role, COLLEAGUE.SKILL produces a versioned skill package with two coordinated tracks: a capability track for practices, mental models, and decision heuristics, and a bounded behavior track for communication style, interaction rules, and correction history. The package can be inspected, invoked, updated through natural-language feedback, rolled back, installed across agent hosts, and optionally prepared for controlled distribution. We describe the artifact contract, generation workflow, correction lifecycle, deployment surface, and domain presets implemented in the open-source system. At the time of writing, the public repository has approximately 18.5k GitHub stars; the gallery lists 215 skills from 165 contributors and more than 100k cumulative stars across listed skill cards. The system illustrates how person-grounded skills can be represented as portable, correctable packages rather than opaque prompts or hidden memories.

On-policy distillation (OPD) trains a student on prefixes sampled from its own policy while matching a stronger teacher. This addresses the prefix mismatch of offline distillation, but early student rollouts can still be poor, placing teacher supervision on weak or low-quality prefixes. We propose Trust-Region behavior Blending (TRB), a warmup method that replaces the early rollout policy with the closest-to-teacher behavior policy inside a student-centered KL trust region, while keeping the per-prefix reverse-KL OPD loss unchanged. The KL budget is annealed to zero, so training returns to pure student rollouts after warmup. Across two math-reasoning distillation settings, TRB attains the strongest average among the compared methods.

Zero-shot text-to-speech (TTS) has improved substantially for single-speaker synthesis, yet expressive long-form multi-speaker dialogue remains difficult. A common workaround is to synthesize each turn with a monologue TTS model and stitch the outputs together. This adds inference cost and often breaks acoustic consistency, conversational coherence, and affective continuity across turns. Recent dialogue TTS systems have begun to address this setting, but they still struggle to keep expressive coherence, controllable speaker switching, and monologue quality at the same time. We present SwanData-Speech and SwanVoice. SwanData-Speech builds monologue and dialogue corpora from in-the-wild audio, using Swan Forced Aligner for pause-aware word-level alignment and RobustMegaTTS3 for pronunciation-hard cases. Built on these data, SwanVoice is a zero-shot TTS model for 1--4 speakers, combining a 25 Hz VAE, raw-text conditioning with pause-aware symbols and pinyin substitution, and a flow-matching DiT with speaker-turn conditioning. Training starts from monologue speech, moves through mixed and real dialogue data, and then uses DiffusionNFT post-training with phone-level and speaker-similarity rewards. On SwanBench-Speech, SwanVoice obtains higher richness and hierarchy scores than all evaluated open-source baselines in both monologue and dialogue settings, while content accuracy remains the main limitation. Audio demos are available at https://swanaigc.github.io//#swanvoice.

We present Mellum 2, an open-weight 12B-parameter Mixture-of-Experts (MoE) language model with 2.5B active parameters per token. Mellum 2 is a general-purpose language model specialized in software engineering, spanning code generation and editing, debugging, multi-step reasoning, tool use and function calling, agentic coding, and conversational programming assistance, and it is the successor to the completion-focused 4B dense Mellum model. The architecture builds on the Mixture-of-Experts (64 experts, 8 active) and combines Grouped-Query Attention with 4 KV heads, Sliding Window Attention on three of every four layers, and a single Multi-Token Prediction head that doubles as both an auxiliary pre-training objective and a built-in draft model for speculative decoding; each choice was validated by ablation with inference efficiency on commodity GPUs as a design constraint. Pre-training spans approximately 10.6 trillion tokens through a three-phase curriculum that progressively shifts the mixture from diverse web data toward curated code and mathematical content, optimized with Muon under FP8 hybrid precision and a Warmup-Hold-Decay schedule with linear decay to zero. The pre-trained base is extended to a 128K context window via a layer-selective YaRN and then post-trained in two stages (supervised fine-tuning followed by RLVR), yielding two released variants: an Instruct model that answers directly and a Thinking model that emits an explicit reasoning trace before its final answer. Across code generation, math and reasoning, tool use, knowledge, and safety benchmarks, Mellum 2 is competitive with open-weight baselines in the 4B-14B range while running at the per-token compute of a 2.5B dense model. We release the base, instruct, and thinking checkpoints, together with this report on the architecture decisions, data pipeline, and training recipe behind them, under the Apache 2.0 license.

Real-world image restoration (IR) is bottlenecked by the scarcity of high-quality paired training data. Synthetic datasets are abundant but often fail to model real-world degradations, while real-world paired datasets are expensive and difficult to capture. As a result, IR models trained on these datasets show limited generalization in real-world scenarios. In this work, we propose Generative Ground Truth (GGT) by using generative multimodal foundation models (MFMs) to produce high-quality (HQ) targets from real-world low-quality (LQ) images. We first conduct a systematic evaluation of nine state-of-the-art MFMs, including Nano-Banana-2 and GPT-Image-2, on images of various scenes and degradation types. The results demonstrate that Nano-Banana-2 with VLM-based adaptive prompting shows the highest capability to synthesize perceptually realistic and content-faithful HQ targets, which can serve as the GGT for the LQ input. We then employ Nano-Banana-2 to build a GGT synthesis pipeline, which involves multi-stage quality control to ensure data reliability, and construct GGT-100K, an LQ-HQ paired dataset comprising 103,707 training pairs and covering diverse scenes and complex real-world degradations. A test set of 500 image pairs is also established. Extensive experiments show that GGT-100K consistently improves the real-world generalization of a wide range of IR models, with particularly strong benefits for finetuning generative models for IR tasks. Our results suggest that MFMs can serve as practical tools for restoration-oriented data generation, and GGT-100K is a useful resource to expand the generalization boundaries of real-world IR models.

As agent capabilities advance, existing benchmarks, such as τ^2-Bench, are becoming increasingly saturated. Yet constructing new benchmark tasks remains complex, costly, and labor-intensive. Moreover, the standard approach, in which scenarios are first written in natural language and then mapped to tool sequences, captures only a narrow subset of the tool-use patterns agents exercise. In this paper, we address these problems by reversing the task construction process. We propose TASTE: Task Synthesis from Tool Sequence Evolution, an automatic method that generates challenging tasks with broader tool-use coverage. TASTE utilizes an Adaptive Contrastive n-gram model trained on LLM-judged validity signals. This enables sampling valid tool sequences that cover a vast range of tool combinations. TASTE then selects representative sequences from the pool via clustering, instantiates them into complete benchmark tasks, and refines them through iterative difficulty evolution. Using TASTE, we construct τ^c-Bench, a challenging extension of the three domains of τ^2-Bench. We evaluate 11 agent/user LLM pairs and find that models nearly saturating τ^2-Bench suffer severe performance drops on our tasks (e.g., Gemini-3-Flash falls from 0.82!-!0.94 to 0.28!-!0.61). Beyond increasing difficulty, our generated tasks more than double the number of unique tool combinations agents must execute. Our results suggest high scores on existing benchmarks often reflect saturation rather than robust task-solving ability. By automating the generation of difficult, high-coverage benchmarks, TASTE enables continuous, scalable evaluation of future agents.

Search agents are often trained as policies over growing transcripts: the model must decide how to search while also remembering what it has seen, which evidence is useful, which constraints remain open, and which claims have actually been checked. We argue that this formulation puts too much routine state management inside the policy: reinforcement learning is forced to optimize both semantic search decisions and recoverable bookkeeping that the environment can maintain more reliably. We introduce Harness-1, a 20B search agent (retrieval subagent) trained with reinforcement learning inside a stateful search harness. The harness maintains environment-side working memory, including a candidate pool, an importance-tagged curated set, compact evidence links, verification records, compressed and deduplicated observations, and budget-aware context rendering. The policy retains the semantic decisions: what to search, which documents to keep or discard, what to verify, and when to stop. Across eight retrieval benchmarks spanning web, finance, patents, and multi-hop QA, Harness-1 achieves 0.730 average curated recall, outperforming the next strongest open search subagent by +11.4 points and remaining competitive with much larger frontier-model searchers. Its gains are especially strong on held-out transfer benchmarks, suggesting that reinforcement learning over explicit search state can produce retrieval behaviors that generalize beyond the training domains. Our code is available at https://github.com/pat-jj/harness-1.

Speculative decoding accelerates LLM inference by drafting multiple tokens and verifying them in parallel with the target model. However, its practical speedup is constrained by the trade-off between draft quality and drafting cost: autoregressive drafters model causal dependencies among draft tokens but incur sequential overhead, while parallel drafters reduce drafting cost but weaken intra-block dependency modeling. In this paper, we propose Domino, a speculative decoding framework that decouples causal dependency modeling from expensive autoregressive draft execution. Domino first uses a parallel draft backbone to produce preliminary draft distributions for the entire block, and then applies a lightweight Domino head to refine them with prefix-dependent causal information. To stabilize teacher-forced causal encoding, we further introduce a base-anchored training curriculum that first strengthens the parallel backbone and then gradually shifts optimization toward the causally corrected final distribution. Experiments on Qwen3 models show that Domino achieves up to \(5.49\times\) end-to-end speedup under the Transformers backend and up to \(5.8\times\) throughput speedup under SGLang serving.

Watermarking embeds statistical signatures in AI-generated text for detection and attribution. We reveal a fundamental vulnerability: when users access multiple models (today's reality), watermarks trivially fail. Watermarks perturb output distributions away from the original, and in competitive markets, these perturbations are typically independent across providers. We theoretically prove that averaging output probability distributions recovers the unwatermarked distribution with up to a second-order error term. Empirically, simply averaging 3-5 models cancels out these perturbations. We introduce WASH (Watermark Attenuation via Statistical Hybridisation), which solves practical challenges in ensemble generation: vocabulary misalignment and tokenisation differences across heterogeneous models. Experiments across six watermarking schemes and three LLMs show that averaging across 3 models suppresses detection z-scores from 5-300 to below 2 (below the detection threshold of 4) and reduces TPR at 5% FPR to below 50%, while improving quality by 27.5% and running 6 times faster than the best baseline on the long sequence generation. Our results suggest that robust AI-text detection via watermarking requires either accepting this fundamental vulnerability or unprecedented coordination among model providers.

The recent "Reasoning with Video" paradigm utilizes Video Generation Models (VGMs) to generate temporally coherent visual trajectories to complete reasoning tasks. Although state-of-the-art VGMs excel at visual quality, they often struggle to understand and follow task-specific rules, leading to logical failures across diverse reasoning scenarios. Existing efforts try to utilize Vision-Language Models (VLMs) as problem pre-solvers to produce or refine textual guidance for the VGM. However, textual descriptions fail to capture intricate spatiotemporal details, and VGMs often struggle to faithfully execute fine-grained or long-tail instructions even with a valid plan. While VLMs struggle as solvers, they possess strong perception capabilities to evaluate process-constraint satisfaction and final-goal achievement. Leveraging this strength, we introduce a paradigm shift that transitions the role of VLMs to "teachers". Specifically, a VLM teacher extracts task-specific rules to formulate differentiable rewards, guiding a VGM Reasoner via test-time online optimization of a lightweight LoRA module. This strategy enables adaptive test-time optimization and extends the reasoning capabilities beyond the VGM's intrinsic boundaries. Evaluations on symbolic (VBVR-Bench) and general-purpose (RULER-Bench) video reasoning benchmarks show that the proposed method yields a 16.7-point average performance gain, outperforming the VLM-as-Solver paradigm (+0.4 points) and Best-of-N scaling (+2.2 points) by a large margin at comparable test-time cost. These findings reveal that integrating VLMs as test-time teachers offers a promising paradigm for achieving generalizable video reasoning. Project Page: https://VLM-as-Teacher.github.io/

Multi-agent LLM workflows route inference through specialized roles to lift end-task accuracy, but jointly training those roles with reinforcement learning is unstable in ways that are poorly understood. We study when end-to-end RL training of multi-agent LLM workflows improves over their base models, comparing Shared-Policy training, where all roles update one policy, with Isolated-Policy training, where each role has its own parameters. Our experimental matrix spans Eval-Opt, Voting, and Orch-Workers workflows, math and code tasks, and three model scales (0.6B, 1.7B, 4B). We find that multi-agent RL usually improves over base models, but gains depend jointly on workflow, task, and scale, not on policy sharing alone. Isolated-Policy tends to reach higher peak accuracy yet more often falls off a terminal accuracy cliff, while Shared-Policy training does not eliminate failure; it redistributes failure into qualitatively different patterns. We then explain the strongest of these patterns through role-level gradient dynamics induced by workflow topology and policy routing: under Isolated-Policy, parallel same-role agents on shared prompts amplify per-role gradients and drive terminal degradation in Voting and Orch-Workers workflows; under Shared-Policy, asymmetric per-step gradient mass causes the shared policy to be captured by the dominant role, producing different failure signatures by task and workflow. Together, the empirical map and its underlying mechanisms show that policy sharing routes training pressure through different channels rather than offering uniform stability, making it a design choice with workflow- and task-conditional tradeoffs.

Bloomberg : Alphabet is raising $80B in equity offerings, including a $10B investment deal with Berkshire Hathaway, to help raise money for its AI spending plans —  Google parent Alphabet Inc. is raising $80 billion in equity offerings, including an investment deal with Berkshire Hathaway Inc. …

Brody Ford / Bloomberg : HPE reports Q2 revenue up 40% YoY to $10.7B, vs. $9.74B est., Server revenue up 33%, forecasts revenue for FY26 and FY27 above est.; HPE jumps 30%+ after hours —  Hewlett Packard Enterprise Co. shares soared in extended trading after the company gave an outlook for annual sales that topped estimates …

Michelle Ma / Bloomberg : Gigascale Capital, a climate tech VC firm co-founded by former Meta CTO Mike Schroepfer, closed a $250M fund to back early-stage startups supporting the AI boom —  Gigascale Capital closed a $250 million fund for early-stage startups.  —  Climate tech venture capital firm Gigascale Capital …

Rohan Prabhu / Step Security Blog : Researchers find packages in the @redhat-cloud-services npm namespace shipped malware that harvests credentials for GitHub Actions, AWS, GCP, Azure, and others —  Several packages in the @redhat-cloud-services npm scope were found to carry malicious payloads that fire via a preinstall hook on every npm install.

Jason Koebler / 404 Media : Hackers say they used Meta's AI support chatbot to change emails tied to Instagram accounts, amid a wave of high-profile account takeovers; Meta fixed the issue —  The exploit shows the extreme risk of offloading technical support to AI.  —  Hackers say that they used Meta's AI support chatbot …

Brody Ford / Bloomberg : Source: Salesforce has a stake in Anthropic worth ~$5B; Salesforce first invested about $50M in an early 2023 round and has continually invested in rounds since —  Salesforce Inc. has a stake in Anthropic PBC worth about $5 billion after repeatedly investing in the ascendant AI startup.

Gavin Finch / Reuters : The US sanctions Nobitex, Iran's largest crypto exchange, accusing it of helping Iran's government and blacklisted state institutions evade Western sanctions —  The United States announced sanctions on Iran's biggest cryptocurrency exchange on Tuesday, accusing it of enabling the Iranian government …

Samantha Subin / CNBC : Palo Alto Networks reports Q3 revenue up 31% YoY to $3B, including $388M from CyberArk and Chronosphere, vs. $2.94B est., and forecasts Q4 revenue above est. —  - Palo Alto Networks topped Wall Street's third-quarter estimates, citing AI advancements

Aisha Malik / TechCrunch : Meta is testing a Series feature, letting select creators make episodic Reels that are placed in a dedicated hub on their profile, using both old and new Reels —  Meta is testing a new “Series” feature for Reels that's designed to make it easier to keep up with serialized content on Instagram and Facebook, the company told TechCrunch.

Gowri Gurumurthy / Bloomberg : Source: a CoreWeave-tied data center raised $900M via five-year junk bonds, priced at par to yield 7.5%, as the sector increasingly turns to high-yield bonds —  A data center tied to CoreWeave Inc. raised $900 million from a high-yield note offering, joining a wave of junk issuers tapping debt markets …

Sawdah Bhaimiya / CNBC : Marvell shares closed up 32.52% on Tuesday after Jensen Huang hailed the chipmaker as the “next trillion-dollar company” at Computex —  Nvidia's CEO Jensen Huang hailed Marvell Technology as the next trillion-dollar firm, sending its shares up 32% on Tuesday.

Ram Iyer / TechCrunch : Microsoft releases ASSERT, an open-source framework that lets developers generate and run AI behavior tests using natural-language descriptions —  AI researchers and labs have advanced by leaps and bounds in evaluating AI models for everything from safety and compliance to sycophancy and alignment.

The International Aids Vaccine Initiative(IAVI)、牛津大学以及 Moderna 公司正在研发针对埃博拉病毒的疫苗。IAVI 表示正在刚果民主共和国爆发的埃博拉疫情可能是至今最严重的。疫情发生在冲突地区，已经报告了逾千例疑似病例，邻国乌干达已确诊 9 例。目前已知有六种埃博拉病毒株，只有三种会引发疫情。最常见的 Zaire 毒株已有针对性的疫苗，但此次爆发的是比较罕见的 Bundibugyo 毒株，目前还没有针对它的疫苗。Moderna 公司宣布将利用 mRNA 技术研发针对 Bundibugyo 毒株的疫苗。

最近发表的两项研究显示，巴西亚马逊地区开始出现此前预测几十年后才会出现的情景，包括旱季延长和降雨模式改变。如果没有采取应对措施，情况可能会迅速恶化，对生物多样性、天然水库的补充以及森林功能构成威胁。其中一项研究表明，亚马逊地区的旱季正从四个月延长至六个月，期间降水量减少逾 150 毫米。第二项研究分析了 2023 年至 2024 年间亚马逊地区的干旱情况。研究结果显示，过火面积增加了 9%，森林退化预警增加了 19%，在干旱高峰期，多达 420 万公顷的森林受到火灾影响。结果表明，干旱、火灾和退化的循环在加剧，削弱了生态系统的恢复能力。亚马逊雨林的面积也可能会减少。

去年 10 月的一天，Dong Hui 突然决定试试能不能握笔写字。6 年前他因为车祸导致的脊髓损伤而颈部以下瘫痪。他缓慢而坚定的写下了自己的名字、谢谢和日期。他能做到这一切来自他参加的脑机接口芯片试验。2024 年 11 月 Dong Hui 成为中国首批接受脑部手术植入侵入式脑机接口芯片的患者之一。今年三月他使用的植入式脑机接口产品获得了商业使用批准。他植入的脑机接口设备被称为 NEO，由上海初创公司 Neuracle Technology 和清华大学合作研发。手术历时约 1.5 小时，收集脑电信号的传感器植入放置在他的硬脑膜上。植入物会将信号传输到计算机。计算机将信号翻译成指令，控制他每天 2.5 小时训练期间佩戴的软体机器人手套，帮助他学习抓握。手术后大约一周他开始康复训练，“训练的第九天，我的右手成功不用手套抓住了一个球，那真是个奇迹。”悉尼科技大学的脑机接口研究员 Avinash Singh 表示，NEO 迅速获得批准的原因之一是其侵入性相对较小，它的 8 个传感器放置在大脑保护膜之上，相比下马斯克（Elon Musk）所创办的 Neuralink 公司开发的 N1 脑机芯片直接穿透了大脑皮层。NEO 的出血、胶质瘢痕形成和长期信号衰减的风险较低。中国还着手将脑机接口列入医保，将其与量子技术、人形机器人等列为对中国未来科技竞争力至关重要的六大关键产业之一。信息科学家 Meicen Sun 表示，中国一大优势是患者乐于接受新技术。美国初创公司 Axoft 正与中国公司合作在中国对四名患者进行脑机接口测试，并计划扩大规模。

胰腺癌是最致命的癌症，大部分现有疗法的效果甚微。现在名为 daraxonrasib 的药物公布了 III 期临床试验结果，有 500 名胰腺癌已扩散的患者参与了试验，其中 248 名患者每日服用 daraxonrasib，其余 252 名接受化疗。结果显示，服药组的中位生存期为 13.2 个月，化疗组为 6.6 个月，也就是药物将患者的生存期延长了一倍，而且副作用更少。研究报告公布在芝加哥举行的美国临床肿瘤学会年会上，专家认为这种药物有望引领一场治疗革命。Daraxonrasib 的作用机制是靶向名为 Kras 的蛋白质，这种蛋白质驱动了几乎所有胰腺癌。药物通过粘合分子去捕获并抑制 Kras 蛋白，从而阻止肿瘤的生长。

由 Amazon、Cisco, Google、Intel、Microsoft、Mozilla 和 Netflix 等联合组建的开放媒体联盟 AOMedia 正式发布了 AV1 的后继者 AV2 编解码器。AV2 在 AV1 继续上提高了压缩效率，以更低的比特率实现高质量视频传输，为流媒体、广播和实时视频会议不断变化的需求进行了优化。AV2 增强了对 AR/VR 应用的支持，支持多节目分屏播放，改进屏幕内容处理，能在更宽的视觉质量范围内运行。

马来西亚新网络安全法规星期一（6 月 1 日）生效，要求各大社交媒体平台验证用户年龄，并禁止 16 岁以下儿童注册账户。这项新法规适用于在马来西亚拥有至少 800 万用户的社媒供应商，包括 Facebook、Instagram、TikTok、YouTube 等。该国通信监管机构表示将给予社媒平台一段宽限期实施这些措施，但未说明宽限期的截止日期。新《网络安全法》的相关规定包括新的《儿童保护法》和《风险缓解法》，并要求社媒平台“加强内容管理”。通信与多媒体委员会说，未能遵守这两项守则的公司可面临最高 1000 万令吉的罚款。

法国生化学家 Sébastien Fontaine 15 年来一直试图杀死土壤，他想要了解没有任何生命的土壤能释放多少碳。 他的团队将土壤密封在罐子内，用伽马射线进行灭菌照射。然后等待土壤释放的二氧化碳——这是微生物呼吸持续进行的标志——下降。他们等待了几周，几个月。在显微镜下，经辐射处理的土壤没有显示任何生命迹象，但它仍在继续释放二氧化碳。土壤拒绝停止呼吸。Fontaine 的实验室重复了实验得到了相同的结果。研究人员开始寻找无生命土壤中的呼吸来源。Fontaine 的团队如今报告，他们的土壤样本在六年内持续消耗氧气并释放二氧化碳。他们提出，为生命提供能量的代谢过程也可能发生在活细胞之外。他们的实验表明，即使没有通常组织土壤的生物蛋白质，这种代谢过程也能在土壤中发挥作用。如果他们的假设正确，那么部分生化反应如释放富碳糖分子能量的反应，可能并非生物所独有。此类反应甚至可能在地球生命出现前就已经存在。

2015 年在加拉帕戈斯群岛进行深海考察的科学家在查看遥控潜水器拍摄的影像时，发现了一只体型娇小、通体呈蓝色的章鱼，它位于水下约 1773 米处。科学家捕捉了这只章鱼以进行进一步分析。研究人员如今得出结论：这只体型小到可以放在手掌的可爱小生物属于一个全新物种。研究报告发表在《Zootaxis》期刊上。小章鱼被保存在储藏室中。由于它的独一无二，且极不可能采集到另一只，科学家不愿意对其解剖进行彻底的物种鉴定分析。因此研究团队选择了 mini-CT 扫描，研究表明这种生物手臂很短，臂上的吸盘很少，没有墨囊，皮肤光滑，且有一颗巨大的脊齿。他们将该物种命名为 Microeledone galapagensis。

迁徒鸟、海龟等动物似乎具有感知地磁场的能力，能利用地磁场进行导航。根据发表在《科学》期刊上的一项研究，信鸽肝脏中的富铁免疫细胞可能赋予了其磁罗盘的能力。对信鸽组织薄片的分析发现，其肝脏巨噬细胞富含铁蛋白，但它在脾脏中很少，且在喙和大脑中完全不存在。电子显微镜的进一步观察发现，巨噬细胞紧邻神经元，而这些神经元都与中枢神经系统相连。研究人员设计了一个试验检验富含铁的巨噬细胞是否能像磁罗盘一样为信鸽指引方向：他们使用名为 clodronate liposomes 的药物抑制巨噬细胞的活性。研究团队训练了 34 只信鸽。白天信鸽利用太阳的位置确定方向。当阴天或完全被云层遮蔽时，它们依靠磁感应辨别方向。研究团队给 18 只信鸽注射了 clodronate，24 小时后当云完全遮蔽阳光时将它们逐一放飞。这些信鸽都佩戴了 GPS 发射器，研究团队能实时追踪其飞行轨迹。所有 18 只信鸽都迷路了，直到天空放晴才返回。16 只对照组的信鸽都没有迷路。研究人员表示，如果铁蛋白辅助导航机制得到证实，那么它可能具有普适性，适用于从蜜蜂到蝙蝠，到鲸鱼和鲨鱼等各种动物。

NASA 宣布，由洛克希德马丁臭鼬工厂设计的 X-59 Quess 低音爆超音速飞机将在本月首次尝试突破音速。X-59 设计能突破音速但同时不会有超音速飞机通常会产生的音爆，它会产生更安静的“砰砰声”，类似室内听到关车门的声音。它没有前向窗户，而是通过摄像头和显示屏为飞行员提供飞机前方的增强现实的外部视觉系统。如果 X-59 成功它有望对超音速飞行和航空业产生革命性影响，解除目前对超音速飞行的限制。X-59 于 2025 年 10 月完成首飞，2026 年 3 月以来进行了 14 次试飞，本月的超音速飞行计划在 16.7 公里高度实现 1.4 马赫。

中国正在打击引发食品安全问题的幽灵外卖。幽灵外卖指的是在外卖平台上提供外卖服务但没有实体店的商家。根据周一生效的新规，外卖平台上的商家信息必须与实体店相符，商家还必须注明是否提供堂食服务。去年北京一男子投诉称他通过外卖平台订购的蛋糕质量不佳，上面装饰着不可食用的花朵。此事引发了对“幽灵外卖”的关注。调查发现，他订购蛋糕的连锁店在各大电商平台上列出了近 380 家门店，但实际上却没有一家实体店。其网店还使用了伪造的营业执照。进一步调查显示，从网店订购的蛋糕实际上外包给一个订单转运平台，该平台会将订单分配给出价最低的第三方商家。当局在两个订单转运平台上共查获了 360 万份蛋糕订单。当局还在七大外卖平台上发现了 6.7 万家“幽灵店铺”，这些店铺与订单转运网站“相互勾结，形成非法供应链”。今年四月，市场监管总局宣布对拼多多、美团、京东、饿了么、抖音、淘宝、天猫 7 家电商平台“幽灵外卖”系列案罚款 36 亿元。

中国扩大商业秘密保护范围，将数据和算法纳入其中，以加强防范技术外流。中国国家市场监督管理总局修订的《商业秘密保护规定》在星期一（6月1日）正式施行。这是中国法律首次明确将数据、算法等数字资产纳入商业秘密保护范围。新规也对远程办公和跨境企业合作提出更严格的安全要求。企业必须采取保护措施，包括按照员工职级限制文件访问权限、隐藏敏感信息，以及记录用户操作行为等。规定还将境外实施的侵犯商业秘密行为纳入规制范围，但未明确具体执法机制。配合新规实施，中国国家市场监管总局星期一启动为期一个月的专项执法行动，重点针对生物医药、半导体和人工智能等关键领域，严厉打击“恶意挖角”以及员工跳槽时携带商业秘密等行为。