Weekly Digest — 2026-W33

A MagSafe AI Recorder That Acts for You

A coding agent that can refine its own harness

Spend up to 85% less and run 3× longer coding agent sessions

Figma, but on your production app. Test variants and ship.

Healthspan score from your wearables, labs, and DNA.

Type a goal, join a live voice call with six AI minds

30-60% faster than Claude Code and Codex

The secure environment for agents, apps, and automations

Compare your startup equity grant for free.

The agentic development environment built by Spotify

Duolingo for college applications

Menu bar app for live annotations, whiteboard overlay + more

Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) exhibit fundamentally different behaviors in enhancing multi-task reasoning for large language models (LLMs). Our preliminary experiments revealed a phenomenon: SFT suffers from severe task conflicts under multi-stage training, whereas RL enables stable coexistence across diverse tasks. Empirically, we trace this to the parameter level, observing that RL induces sparse and approximately orthogonal updates across tasks. We provide a theoretical explanation for this mechanism by analyzing multi-task gradient interference. Our results reveal a distinction: interference in SFT is norm-limited, scaling with the absolute gradient magnitude, whereas interference in RL is variance-limited, bounded by the gradient variance induced by advantage normalization and on-policy optimization. This small variance bound yields near-orthogonal optimization directions across tasks. Leveraging this insight, we propose Parallel-RL, a paradigm that decouples multi-task training, significantly improving efficiency and flexibility.

Recent works train agents by constructing large-scale multimodal environment pools. However, we find that simply increasing the number of multimodal environments does not always benefit. We further analyze the limitations in current multimodal environment distributions through a series of experiments. Based on these findings, we study how to build more effective training environment distributions from two dimensions: **diversity** and **difficulty structure**. For diversity, we propose **Ability-aware Environment Selection (AES)** to obtain diverse environment sets. For difficulty structure, we propose **Hierarchical Difficulty Curriculum (HDC)**, which organizes curriculum learning through two difficulty levels: harness weakening and state-scale progression. Experiments show that AES and HDC effectively improve multimodal agent training.

World-Action Models (WAMs) improve end-to-end autonomous driving by transferring video dynamics priors to action prediction, but existing methods require costly future generation at inference. We present SimWAM, a simple yet effective WAM that uses video generation purely as a training signal. It co-trains a pretrained video expert and a lightweight action expert with joint flow matching. An isolated attention mask keeps action prediction independent of future frames, allowing the video branch to be discarded after training and leaving a self-contained planner that directly predicts trajectories. Since the two experts share no parameters and interact only through a unified attention interface, the video backbone could be replaced and the action expert scaled independently without modifying the learning objective or inference pipeline. We further apply reinforcement learning to optimize a compositional driving reward beyond trajectory imitation. Our SimWAM achieves 91.5 PDMS on NAVSIM, surpasses state-of-the-art WAM-based planners with substantially lower latency, and transfers zero-shot to nuScenes. These results position SimWAM as a simple yet solid baseline that could readily benefit from advances in video generation for efficient autonomous driving. The code and model weights are available at https://github.com/H-EmbodVis/SimWAM/

Generic parameter-efficient fine-tuning (PEFT) methods transferred from language models can fail silently on real-time detectors, whose heterogeneous operators and detection-specific components impose placement constraints absent from regular Transformer stacks. We propose YOLO-PEFT, a structure-aware framework that formulates adapter placement as an auditable constraint-planning problem. Given a detector graph, a PEFT request, and a resource budget, YOLO-PEFT assigns operator and semantic roles, evaluates explicit operator-validity, detector-semantic, graph-interface, and deployment predicates, records a reason code for each excluded module, and either emits a budgeted target-module plan or returns Refuse before training. Under the official VOC07+12 trainval-to-VOC07 test protocol, planner-selected RS-LoRA reaches 0.7138 and 0.7307 mAP50-95 on YOLO11s and YOLO12s, respectively, compared with 0.6428 and 0.6662 for Full-SFT. On RT-DETR-L, all seven evaluated LoRA-family configurations cross the predefined catastrophic threshold, supporting a calibrated Refuse-to-Full-SFT decision within the evaluated coverage. A controlled YOLO11 audit further shows that LoRA reduces peak training memory by 43.9 percent, although training takes 1.72 times longer. Within the evaluated detector families, placement policies, and calibration coverage, YOLO-PEFT replaces manual target-module trial and error with explicit, inspectable planning while preserving verified train-save-merge-export paths; refusal on unseen detector architectures remains an open validation problem. Project Page: github.com/Tencent/YOLO-Master

Deploying autonomous multimodal agents in continuous, real-world environments requires them to ingest unbounded audio-visual streams and maintain hour-scale memory. However, current evaluations predominantly rely on brief clips and multiple-choice formats. This design allows minimal baselines that process only the last four frames to match or surpass complex streaming models, while answer options also expose language shortcuts. We introduce StreamArena, a benchmark for hour-scale, interactive streaming video understanding. StreamArena contains 243 full-length videos averaging 88.8 minutes and 3,646 rigorously annotated, open-ended question-answer pairs that evaluate real-time perception, historical retrospection, proactive interaction, and multimodal tool utilization. Evaluation across diverse systems exposes a tension between continuous interaction and long-horizon multimodal comprehension. Methods that retain only recent frames cannot recover distant events, methods that convert past observations into text lose visual evidence, and methods that repeatedly compress visual memory struggle to preserve fine-grained details over time. We address this tension with StreamMind, a two-tier architecture that assigns latency-critical interaction and proactive monitoring to independently scheduled frontend workers, while backend workers asynchronously construct persistent multimodal memory and perform historical recall and external search. StreamMind outperforms existing streaming baselines across all four capabilities and reduces query-to-answer latency by reusing persistent state.

CLI-based software-engineering agents have matured rapidly, yet the open ecosystem has converged on a single training environment: trajectory datasets used to fine-tune open models are collected almost exclusively under OpenHands. Models fine-tuned on this data score well under OpenHands but degrade substantially when deployed under any non-training scaffold. Untrained base models do not show this divergence, indicating the gap is fine-tuning-induced and tied to the conventions of the training scaffold. We argue that a load-bearing scaffold-specific behavior is planning structure, in two senses this paper distinguishes: explicit planning, a pre-execution plan produced as a first-class artifact, and implicit planning, the structural conventions that shape execution throughout the agent loop. Under this hypothesis, closing the gap requires moving planning from a fixed scaffold artifact to a learned model capability. We introduce Decoupling CLI Agent Scaffolding (DCAS), a backend-substitution interception layer that routes API traffic between any CLI scaffold and any backend model without modifying the scaffold, enabling cross-scaffold evaluation and planning-aware trajectory collection. Using DCAS, a controlled plan-source intervention confirms planning quality is a high-leverage component, with gains exceeding the cross-scaffold drops we observe. A model fine-tuned on a small set of DCAS-collected planning-aware trajectories under a single scaffold gains consistently across non-training scaffolds, and the two senses of planning are empirically separable in training data.

We introduce BDH-CQ, a reasoning model that combines in-context learning with recurrent latent reasoning. Inputs presented at inference time continuously update the model's recurrent memory; the model then solves a query through iterative computation in a high-dimensional latent space, without verbalizing its intermediate reasoning. We evaluate the model on the public ARC-AGI-1 evaluation set and use controlled ARC-like interventions to study what it learns from demonstrations, how consistently it applies an inferred transformation, and which concepts remain difficult. A 150M-parameter configuration reaches 29.5% pass@2 at a computed inference cost of \$0.0007 per task. This operating point breaks through the previously reported ARC-AGI-1 cost-accuracy Pareto frontier, establishing a new state of the art in benchmark cost efficiency.

Macaron-V1 is an open agent-model family for experiential intelligence: learning from experience in real environments and continuing to learn after deployment. It is organized around two system goals. Adaptation is pursued through recursive improvement of versioned model-harness pairs, where experience from one configuration is evaluated under an external contract and used to construct its successor. Collaboration is pursued via the Mixture-of-LoRA (MoL) architecture that freezes a base model, composes specialist LoRA adapters, and selects one LoRA per user turn. The flagship Macaron-V1-Venti combines a 744B GLM-5.2 base with four LoRAs for chat, agent, coding, and GenUI; the Qwen3.6-based Macaron-V1-Tall (50B) uses the same design for local deployment. This report presents Macaron-V1 as a co-designed system spanning architecture, algorithms, and infrastructure. The MoL architecture supports continual learning through extensible LoRA specialists. The algorithm combines Model-Harness Co-design and recursive self-improvement loop, including the UI4A component-native GenUI harness, a stateful action substrate, versioned HCP contract, and the agentic RL framework MindForge. The supporting infrastructure includes the post-training platform MinT, the long-context RL method LongStraw, and stability techniques for sparse MoE and DSA base models. We evaluate Macaron-V1 on Personal Intelligence, GenUI, and general capability benchmarks against frontier baselines. Our results validate the current system, while compounding gains from continual learning and collective intelligence remain open questions.

As AI coding agents take on increasingly complex, long-horizon software engineering tasks, existing benchmarks are rapidly saturating and their evaluation quality has come under serious scrutiny: a recent audit found that nearly 60% of unsolved SWE-bench Verified instances contain flawed tests -- either overly narrow tests that reject correct solutions or overly broad tests that check unstated requirements -- and that frontier models can verbatim reproduce gold patches from training data. Code refactoring, which requires coordinated, behavior-preserving changes across many files, offers a substantially harder and more realistic test of agent capability, yet remains underserved by current benchmarks. We introduce SWE-Bench ProMax, an expert-curated, multilingual code refactoring benchmark of 170 instances drawn from real commits across seven programming languages (Python, Java, TypeScript, Go, C, C++, and Rust). Every instance undergoes rigorous, multi-stage curation that directly addresses the quality problems identified in prior benchmarks: issue descriptions are rewritten from scratch to provide precise, unambiguous specifications, and test suites are manually reviewed to remove overly narrow and overly broad tests. Tasks with insufficient complexity or limited cross-file scope are filtered out, yielding a benchmark of challenging, large-scale refactoring tasks that average 11.4 modified files and 261.6 lines of code per instance, substantially exceeding the scale of existing benchmarks. Experiments with frontier models under two agent scaffolds show that the best model achieves only 41.2% resolve rate, confirming that SWE-Bench ProMax presents a meaningful and unsaturated challenge for current AI coding agents. Our benchmark is available at https://huggingface.co/datasets/swe-bench-promax/SWE-Bench-ProMax.

We present Ouroboros, a self-developing agent harness whose tools, prompts, context assembly, and core implementation improve through reviewed commits that become the runtime for later work. Core evolution proceeds in two modes. In recursive free evolution, improvement is itself a task, and completing one evolution cycle can schedule the next. In experience-driven core evolution, ordinary work and social interaction expose bugs, rough edges, and inefficient context construction that lead to reviewed structural changes. On Terminal-Bench 2.1, an Opus 5 run scores 86.74%, the best result reported on the benchmark. On OSWorld-Verified, an Opus 5 run reaches 90.69%, exceeding the best previously reported score. A five-rollout CL-Bench campaign achieves a normalized reward of 0.2301, setting a new state of the art. Hope is the longest-running publicly documented Ouroboros deployment. It is a 161-day living agent experiment in free evolution under governed human communication across seven surfaces. Human interaction surfaces faults and generates proposals, but the agent decides which changes to pursue. Because a self-developing agent may rewrite its own code and select new model APIs, operational safety becomes a primary design problem: guardrails must remain authoritative under evolutionary and public social pressure. Benchmark campaigns use frozen system snapshots, while Hope continues live evolution on a separate lineage.

On-policy (Self-)Distillation (OPD / OPSD) has shown strong potential for post-training large language models (LLMs). However, existing methods still rely heavily on external supervision, including ground-truth signals, environmental feedback, or guidance from larger models, and therefore fall short of genuine "self"-distillation. In this study, we show that on-policy self-distillation can be achieved using only a model's own generations via internal consistency. We propose unsupervised on-policy self-distillation (U-OPSD). U-OPSD first samples multiple rollouts and constructs a pseudo solution by majority vote under a self-consistency threshold. It then conditions the model's distribution on the pseudo-solution and distills itself on the disagreeing completions, allowing the model to correct itself precisely where it is confidently wrong. Across diverse benchmarks, base models, and training settings, U-OPSD consistently improves over the base models and matches or surpasses supervised methods with ground truth (GT) such as OPSD and GRPO. On five mathematical reasoning benchmarks, i.e., AIME24, AIME25, HMMT25, MATH500, and AMC23, U-OPSD improves over the base model by 8.5% and 10.7% on Qwen3 non-thinking mode at 4B and 8B scales, and outperforms OPSD by 3.2% and 2.3% on average, respectively. In thinking mode, U-OPSD stays on par with OPSD, ahead by 0.9% at 4B and level at 8B and surpassing GRPO by 0.7% and 1.1%, respectively. Code is available at [https://github.com/williamium3000/u-opsd](https://github.com/williamium3000/u-opsd).

Memory systems have shown promise for improving agent performance, but their potential remains largely unexplored for small language models, which struggle to generate sufficient successful trajectories on their own. We propose Agent Memory Distillation (AMD), a training-free framework that transfers structured knowledge from a large teacher agent to a small student agent through hierarchical memory. AMD constructs three complementary memory types from successful teacher trajectories: Workflow memory encodes task-level strategies, Subtask memory provides concrete behavioral examples at an intermediate granularity, and Function memory captures per-function calling conventions and common pitfalls. Workflow and Subtask memories are injected proactively at the start of each task, while Function memory is retrieved reactively upon tool-calling errors. We evaluate AMD on three tool-use benchmarks using four student models (4B-8B parameters) with GPT-5-mini as the teacher, achieving average accuracy gains of 27.2%p, 11.2%p, and 3.4%p on AppWorld, BFCL V3, and ToolSandbox, while consistently outperforming existing memory-based baselines. Further analysis shows that Subtask memory contributes the largest gains, teacher effectiveness depends on both teacher capability and student compatibility, and 4B-sized students benefit most from AMD.

M.G. Siegler / Spyglass : With his long essay, Zuckerberg returns to his “open” AI arguments at an opportune time, as Chinese open models narrow the performance gap at much lower cost —  Zuck returns to his “open” AI arguments at an opportune time...  I'm old enough to remember the last time Mark Zuckerberg …

Bloomberg : Sources: OpenAI bought back ~$7B in shares from current and former employees in a tender offer valuing it at $852B, unchanged from its most recent funding round —  OpenAI has completed a deal to help employees sell roughly $7 billion worth of shares in the company ahead of a possible Wall Street debut …

The Information : Source: Applied Compute, which helps companies customize models with their data, is in talks to raise “hundreds of millions” led by Elad Gil at a ~$3B valuation —  Applied Compute, a year-old startup that helps companies run and customize open-source models with their own data …

Alicia Clanton / Bloomberg : Filing: Sergey Brin donated an additional $20M to Building a Better California, which fights California's wealth tax, bringing his total contributions to $102M —  Billionaire Sergey Brin has now poured over $100 million into his fight to stop a wealth tax in California.

Anthropic : Anthropic details an unreleased Claude model's attempt to solve the Riemann hypothesis; it didn't solve it but “unexpectedly” made strides on a related problem —  Recently, a member of staff at Anthropic gave Claude an unreasonable challenge.  It was about one of the most famous …

Isaiah Poritz / Bloomberg : The Ninth Circuit dismisses Meta's bid to pause a landmark trial by state AGs alleging its products addict children and have caused a youth mental health crisis —  Meta Platforms Inc. failed to obtain an emergency order pausing a landmark trial slated to begin this week involving claims …

Yueqi Yang / The Information : Source: Kalshi's annualized revenue topped $4B in July, up from $2B+ two months earlier, and its operating expenses totaled $300M in June —  World Cup wagers helped double Kalshi's revenue to a more than $4 billion annualized rate in July, from a pace of over $2 billion two months earlier, a person familiar with the matter said.

Natasha Mascarenhas / Bloomberg : Accel raised $3.5B in funds for early-stage investments, including a $1.35B global expansion fund for larger early-stage rounds and rapid follow-on investments —  Accel, a global venture capital firm with offices in Silicon Valley, London and Bangalore, India, has raised $3.5 billion …

Will Knight / Wired : Researchers find that feeding a frontier model's encrypted reasoning traces to a weaker model from the same provider can make it output the traces in plaintext —  Researchers devised a way to extract “reasoning traces” from Claude, GPT, and Gemini.  What they found, they say …

Business Insider : Sources: former OpenAI Chief Product Officer Kevin Weil is aiming to raise $150M for a new AI science startup, seeking a valuation of at least $750M —  Kevin Weil, the former chief product officer of OpenAI, has sought a valuation of at least $750 million for a new AI science startup …

Jordan Novet / CNBC : CoreWeave reports Q2 revenue up 112% YoY to $2.58B, vs. $2.56B est., and a $104B revenue backlog, with 1.5 GW of contracted power; CRWV jumps 9%+ after hours —  CoreWeave shares jumped 14% in extended trading on Tuesday after the AI infrastructure provider reported revenue than topped Wall Street expectations.

Dina Bass / Bloomberg : Super Micro reports Q4 revenue up 93% YoY to $11.1B, vs. $11.3B est., and forecasts Q1 and FY 2027 revenue above estimates; SMCI jumps 9%+ after hours —  Super Micro Computer Inc. gave a revenue forecast for the current quarter that topped analysts' estimates, a sign the booming artificial …

国际团队在 2025 年 3 月至 4 月间，结合 NASA 的 IXPE (Imaging X-ray Polarimetry Explorer、NICER (Neutron Star Interior Composition Explorer) 以及澳洲联邦科学与工业研究组织（CSIRO）的 Murriyang 无线电望远镜，对磁星 1E 1547-5408 进行了超过 140 小时的联合观测。这是人类史上首次针对磁星进行同步的无线电与 X 射线偏振测量。这颗每 2 秒自转一周的磁星能持续发出明亮的无线电与 X 射线。观测显示，其光子偏振度是类似天体的三倍之多，远高于标准表面辐射模型的预期。为了寻找原因，科学家将目光转向量子电动力学中拥有 90 年历史的真空双折射理论。该理论于 1936 年首次提出，认为在极端强烈的磁场下，太空的真空会如同透镜或稜鏡般过滤光线，进而提升总体偏振度。 IXPE 的 X 射线偏振测量能力成为检验此理论的关键。研究团队的模拟结果支持了真空双折射造成此信号的可能性。

Meta 在 Apache 2.0 许可证下推出了 300 亿参数的开放权重模型 Muse Glimmer。Muse Glimmer 为本地运行的智能体工作流程进行了优化，能在配备了单个消费级 GPU 的 Mac 或 PC 上运行，支持从本地智能体和函数调用到本地编程和 LLM 自动化评估等多种应用场景。Meta 将在未来几天推出针对 llama.cpp、MLX 和 ExecuTorch 的集成。

同行审议是学术出版流程的重要一环，然而随着在 AI 时代论文数稳步增长，审稿人数量逐渐跟不上了。以前一篇论文通常能找到两到三位审稿人，如今很多论文只能找到一名审稿人。研究人员表示，这种情况越来越普遍。同行审议成为科学的基石其实只有约 50 年历史，是 1970 年代美国科学基金会为了避免将科研项目的拨款决定置于国会议员手中。英国的情况类似。如今同行审议已成为标准流程。加快审稿的一种方法是引入 AI 审稿。未来 AI 生成的审稿意见可能会日益常见，今天的很多审稿人已经在使用 AI 工具。

2024 年的一项研究分析了美国在 2020 年 1 月到 2022 年 12 月之间的 900 万份死亡证明，发现出租车和救护车司机死于阿尔茨海默病的可能性低于其他任何职业。在 443 种职业中，出租车和救护车司机死于阿尔茨海默病的风险最低。在调整了年龄、性别、种族、民族和教育程度等因素后，大约每 100 名出租车和救护车司机中有 1 人死于阿尔茨海默病，而其整体比例为 1/60。这一模式无法推广到其他驾驶类职业，意味着降低阿尔茨海默病风险并非驾驶而是持续的实时导航：不断确定自身在空间中的位置，追踪目的地，随情况变化更新心智地图（mental maps）。工作依赖于固定或预定路线的驾驶者，如公交车司机和飞机飞行员，没有获得类似的优势。 研究人员认为，导航密集型工作与低阿尔茨海默病风险之间的关联集中在海马体上。海马体是大脑中负责记忆和空间导航的部分，也是阿尔茨海默病最早受损的脑区：空间导航和方向感障碍是最早出现的症状之一。2023 年，研究人员在一个包含逾 22,500 人的数据集上运行了一个机器学习模型，根据环境的复杂程度以 84% 的准确率预测哪些区域的阿尔茨海默病发病率较高。生活在高空间复杂环境中的居民患上阿尔茨海默病的可能性较低。研究人员认为，定期构建认知地图能锻炼阿尔茨海默病最先攻击的神经回路。反复激活与空间导航相关的认知系统有助于延缓症状出现。

最新行业数据显示，2026 年上半年，中国人形机器人制造商占全球出货量的 97% 以上，显示中国在这一新兴领域已领先美国。Smart Analytics Global 的数据显示，2026 年上半年，全球人形机器人出货量约 1 万 9100 台，是去年同期 5100 台的三倍多。总部位于上海的智元机器人超越总部位于杭州的宇树科技，跃居市场份额榜首。今年上半年，智元机器人出货量达 8400 台，占全球出货量的 44%，宇树科技则为 5900 台。

苹果正在 iPhone 和 MacBook 等产品上测试长鑫的内存芯片，以缓解 AI 热导致的内存短缺。苹果已与长鑫就供应内存芯片进行了初步磋商，目标是将其用于部分在中国销售的苹果设备。此前有报道称，PC 制造商惠普和宏碁已开始在美国以外地区销售的笔记本产品中少量使用长鑫内存。长鑫目前占据了约 7% 的内存市场份额，三星、SK海力士和美光占据了九成的市场份额。

四年前 Google 的一位工程总监宣称到 2025 年 AI 将推动企业实行四天工作制。OpenAI 今年早些时候也鼓励企业试行四天工作制，声称 AI 将能大幅提升人类劳动效率，业界应为此做好准备。OpenAI 的一位前员工称，该公司从未真正试行过四天工作制。前员工称 OpenAI 的工作文化令人精疲力竭，频繁举行危机会议，周末要加班，绩效考核残酷无情，同事会被突然解雇。科技公司员工称，他们的工作时长远超典型的每周五天 40 小时。前 OpenAI 员工表示，他们每周至少工作 70 小时。他现在在一家 AI 初创公司工作，每周工作时间接近 50-60 小时，在工作生活平衡上比 OpenAI 好点。OpenAI 和 Anthropic 员工称，公司的冲刺开发通常持续数周，一周的工作时长甚至超过 90 小时。前 Google 员工 Amin Shali 称在离职后其睡眠和整体健康状况都改善了。

微软据报道大幅上涨了 PC OEM 厂商的 Windows 11 授权费。价格上涨于 7 月生效，平均上涨了 7% 至 10%。过去 Windows OEM 授权费上涨通常是个位数，这次上涨幅度更大。具体涨幅将基于 PC 处理器而存在差异，高端 CPU 系统需要支付不同的价格。微软此举无疑是雪上加霜，PC 市场已因内存和存储器涨价而苦苦挣扎。

Anthropic 为其 AI 工具 Claude 生成的文本添加看不见但能被机器读取的水印。此举是为了遵守欧盟 AI Act 第 50 条关于 AI 生成文本透明度要求。Anthropic 表示，水印将适用于全球用户，而不仅仅是欧洲用户。Anthropic 的系统将对文本和支持的文件使用不同的技术，文本添加不可见水印，文件则包含签名的来源信息。用户在阅读文本时不会看到水印。当用户复制粘贴文本内容时，水印仍将保留。Anthropic 还预计部分水印在编辑后仍然有效。

因 Google 与 Epic 诉讼案法官的裁决，Google 开始向第三方应用商店开放其 Play Store。Play Store 美区上架的第一个替代应用商店是 Aptoide，Epic、亚马逊、三星、微软等公司上架各自的应用商店可能只是时间问题。Android 生态系统有大量第三方应用商店，但大部分应用商店都由 Android 厂商捆绑在各自的产品中，如亚马逊的 Fire 平板捆绑了亚马逊的应用商店，三星 Galaxy 智能手机捆绑了其 Galaxy Store。

贫困问题的一个潜在解决方案是：直接给穷人现金。越来越多的研究表明，现金转移支付能使贫困家庭受益。加州圣克鲁兹经济学家 Jonathan Robinson 的团队在美国国际开发署（USAID）的支持下分别在利比里亚和马拉维选择了 150 个村庄，向每户家庭提供一笔一次性的现金（250 美元或 500 美元、750 美元），在每个国家再选择 150 个对照村庄，不提供现金。在之后的两年时间里，研究人员每月实地调查和电话调查了数百名女性。结果显示，贫困家庭通常会很快花光钱，但这笔钱通常花在改善粮食安全等方面的投资上。在两年后，实验组家庭的粮食安全状况仍然高于对照组家庭，显示现金给贫困家庭带来了至少两年的益处。研究人员本想要长期追踪这 600 个村庄，但所有资金都来自 USAID，而 USAID 在 2025 年初被特朗普政府/马斯克的政府效率部解散了，研究没法继续下去了。

8 月 3 日 9:32 pm，一艘搁浅船只在阿拉斯加 Point Highland 附近发出求助信号。海岸警卫队确认该船未遇险后发布了海上援助请求广播，呼吁附近海员提供自愿帮助。这意味着附近船只没有法律义务提供援助。如果海岸警卫队认定该搁浅船只遇险，那么根据国际海事法和美国法律，附近船只的船长有义务提供援助。扎克伯格的超级游轮 Launchpad 当时就在附近，但它没有参与自愿营救，而是减速让身后的一艘船 Wilderness Legacy 驶过它去营救搁浅船只。Wilderness Legacy 将船拖至 Farragut Bay，为其补充了燃料，这艘船上有两名女子、一名儿童和一只狗。扎克伯格发言人称当时扎克伯格及其家人并不在游轮上，称船员未及时收听救援信息，当他们查看到救援信息时救援行动已经开始了。