AI Daily Digest — August 1, 2026: ARC-AGI-3 Harness Discovery, EU AI Gigafactories, Devin SWE-1.7
AI Daily Digest — August 1, 2026: ARC-AGI-3 Harness Discovery, EU AI Gigafactories, Devin SWE-1.7
OpenAI Shows How Two Harness Settings Tripled ARC-AGI-3 Scores
OpenAI published a rare technical deep-dive on July 29 explaining why GPT-5.6 Sol looked weak on ARC-AGI-3, a benchmark where agents explore unfamiliar 2D puzzle games and infer the rules through trial and error. With the official evaluation harness, Sol scored just 7.8%, and GPT-5.5 could barely play at all (0.4%). The team discovered the problem was not the model but the harness: after every game action, all private reasoning was discarded, and a rolling truncation window deleted older actions as history grew. Sol was effectively being asked to figure out each game anew on every single move.
OpenAI 展示了两个评估框架设置如何使 ARC-AGI-3 分数翻了三倍 OpenAI 于 7 月 29 日发布了一篇罕见的技术深度解析，解释了为何 GPT-5.6 Sol 在 ARC-AGI-3 基准测试中表现不佳。该基准测试要求智能体探索陌生的 2D 益智游戏，并通过反复试验推断规则。在使用官方评估框架时，Sol 的得分仅为 7.8%，而 GPT-5.5 几乎无法进行游戏（0.4%）。团队发现问题不在于模型，而在于评估框架：在每次游戏操作后，所有的私有推理过程都会被丢弃，且随着历史记录的增长，滚动截断窗口会删除旧的操作记录。实际上，Sol 在每一步操作时都被迫重新理解整个游戏。
Enabling two Responses API settings — retained reasoning and context compaction — tripled the score on the public task set from 13.3% to 38.3% while cutting output tokens by roughly 6x. With that harness, GPT-5.6 Sol solves all six levels of the benchmark, whereas no frontier model solves any level beyond the first on the official leaderboard. Human testers average around 48% on the RHAE metric the benchmark uses. The finding reframes how benchmark leaderboards should be read: they measure model × harness × settings, not the model in isolation. For anyone building long-running agents, the lesson is concrete — if your harness throws away chain-of-thought between tool calls, you are not measuring your model, you are measuring a model with amnesia.
启用两个 Responses API 设置（保留推理和上下文压缩）后，公共任务集的得分从 13.3% 提升至 38.3%，同时输出 Token 数量减少了约 6 倍。使用该框架，GPT-5.6 Sol 能够解决该基准测试的所有六个关卡，而官方排行榜上没有任何前沿模型能解决第一关之后的任务。人类测试者在该基准测试使用的 RHAE 指标上平均得分约为 48%。这一发现重新定义了如何解读基准测试排行榜：它们衡量的是“模型 × 评估框架 × 设置”的综合表现，而非孤立的模型能力。对于任何构建长周期智能体的开发者来说，教训很明确——如果你的评估框架在工具调用之间丢弃了思维链，那么你测量的不是模型本身，而是一个患有“失忆症”的模型。
EU Opens €10B Tender for Seven AI Gigafactories
The European Commission opened the formal call for tenders on July 30 to build up to seven AI gigafactories across the EU, backed by €10 billion in EU and national public funding with a target of more than €30 billion total once private capital is included. The initiative runs through the EuroHPC Joint Undertaking. Four smaller facilities (at least 75,000 AI chips each) are eligible for up to €500 million; three larger gigafactories (at least 100,000 chips each) can receive up to €1 billion across two development phases. Each gigafactory must pack roughly four times the compute of Europe’s current largest AI data centers.
欧盟启动 100 亿欧元招标，建设七座 AI 超级工厂 欧盟委员会于 7 月 30 日正式启动招标，计划在欧盟境内建设最多七座 AI 超级工厂。该项目由 100 亿欧元的欧盟及成员国公共资金支持，目标是在引入私人资本后总投资额超过 300 亿欧元。该倡议由欧洲高性能计算共同计划（EuroHPC JU）负责执行。四座较小的设施（每座至少配备 7.5 万枚 AI 芯片）最高可获得 5 亿欧元资助；三座较大的超级工厂（每座至少配备 10 万枚芯片）可在两个开发阶段中获得最高 10 亿欧元的资助。每座超级工厂的算力必须达到欧洲目前最大 AI 数据中心的约四倍。
The Commission has signed letters of intent with AMD, NVIDIA and Qualcomm to ease hardware access for winning consortia. Bidding closes on November 12, awards are expected in early 2027, and construction is slated for the same year. An earlier non-binding call for expressions of interest drew 76 submissions covering 60 sites in 16 member states — far more than Brussels anticipated. This is Europe’s most explicit statement yet that AI compute is critical infrastructure, on par with electricity grids and broadband. It responds directly to dependency on US and Chinese compute, following June’s draft laws aimed at reducing reliance on American cloud and chip suppliers. Non-EU entities are barred from gigafactory consortiums, and majority owners must be European.
欧盟委员会已与 AMD、NVIDIA 和高通签署意向书，以简化中标财团获取硬件的流程。投标截止日期为 11 月 12 日，预计于 2027 年初公布中标结果，并计划于同年动工。此前的一项非约束性意向征集吸引了来自 16 个成员国的 76 份申请，涵盖 60 个选址，远超布鲁塞尔的预期。这是欧洲迄今为止最明确的表态：AI 算力是与电网和宽带同等重要的关键基础设施。此举直接回应了对美中算力的依赖，并紧随 6 月份旨在减少对美国云服务和芯片供应商依赖的法律草案。非欧盟实体被禁止参与超级工厂财团，且多数所有权必须由欧洲方持有。
Cognition’s July Blitz: SWE-1.7, Security Swarm, and a Federal Push
Cognition had one of the densest product months of any AI lab in July. On July 8 it launched SWE-1.7, its own frontier coding model trained from a Kimi K2.7 base via reinforcement learning, with four notable innovations: entropy-preserving top-p sampling to stabilize long RL runs, multi-cluster training spanning three continents with fault tolerance, a data-quality pipeline that filters tasks through automated execution tests, and self-compaction that lets the model summarize its working state to extend task horizons past the raw context window. SWE-1.7 scores 42.3% on FrontierCode 1.1 Main, 81.5% on Terminal-Bench 2.1 and 77.8% on SWE-Bench Multilingual, served at roughly 1,000 tokens/sec via Cerebras and free within Devin Pro.
Cognition 的七月攻势：SWE-1.7、安全集群与联邦政府布局 Cognition 在 7 月经历了 AI 实验室中最密集的产品发布月之一。7 月 8 日，该公司推出了 SWE-1.7，这是其基于 Kimi K2.7 基础模型通过强化学习训练的前沿编码模型，具有四大显著创新：用于稳定长周期强化学习运行的“熵保持 top-p 采样”、跨越三大洲且具备容错能力的“多集群训练”、通过自动化执行测试过滤任务的“数据质量流水线”，以及允许模型总结工作状态以突破原始上下文窗口限制的“自我压缩”技术。SWE-1.7 在 FrontierCode 1.1 Main 上得分 42.3%，在 Terminal-Bench 2.1 上得分 81.5%，在 SWE-Bench Multilingual 上得分 77.8%，通过 Cerebras 提供约 1,000 token/秒的推理速度，并在 Devin Pro 中免费提供。
Around the model, Cognition shipped Devin Security Swarm (July 1), an agentic MapReduce approach where multiple agents investigate different code regions and reason across files to find multi-step attack chains — it found 72% of real vulnerabilities in a 50-CVE test set at $90.23 per run. The company also reached FedRAMP High In-Process status (July 13) and signed an MOU with the US Department of Energy to join the Genesis Mission, positioning autonomous software engineering for the federal market. The TierZero acquisition (July 20) extends Devin into incident detection and system reliability. Cognition discloses that more than 90% of its own code is now written by Devin. With over $1 billion raised in May at a $26 billion valuation and annualized revenue up from $37 million to $492 million in twelve months, Cognition’s bet is that the winning stack is model + agent harness + compliance, not a single model.
围绕该模型，Cognition 于 7 月 1 日发布了 Devin Security Swarm，这是一种智能体 MapReduce 方法，多个智能体调查不同的代码区域并跨文件推理，以发现多步攻击链——它在 50 个 CVE 的测试集中发现了 72% 的真实漏洞，单次运行成本为 90.23 美元。该公司还于 7 月 13 日获得了 FedRAMP High 认证流程资格，并与美国能源部签署了加入“创世纪任务”（Genesis Mission）的谅解备忘录，旨在将自主软件工程引入联邦市场。7 月 20 日对 TierZero 的收购将 Devin 的能力扩展到了事件检测和系统可靠性领域。Cognition 透露，其自身超过 90% 的代码现在由 Devin 编写。凭借 5 月份以 260 亿美元估值筹集的超过 10 亿美元资金，以及 12 个月内从 3,700 万美元增长至 4.92 亿美元的年化收入，Cognition 的赌注是：成功的技术栈是“模型 + 智能体框架 + 合规性”，而非单一模型。