与「Author」相关的搜索结果

Every gun, mine, and guard tower on the Berlin Wall was pointed at East Germans. In 1957, the communist regime in East Germany made unauthorized emigration treason after an estimated 2.5 million people voted with their feet, fleeing the communist East towards the West. So in 1961, the authorities closed the border, called the wall “the Anti-Fascist Protection Rampart”, and gave border guards shoot to kill orders. People only ever risked their lives escaping in one direction.

开发系统最极致高效的Agents.md，没有之一： # AGENTS.md ## Core Principles - Choose the simplest implementation that fully satisfies the current requirements. Avoid unnecessary abstraction, configuration, indirection, or speculative extensibility. - Make the smallest necessary change that fixes the root cause. Do not refactor unrelated modules or change strategy semantics unless explicitly requested. - Grow the system in layers. Start from the smallest working end-to-end version and add new capabilities incrementally. Never replace a working system with unfinished complexity. - Reuse existing project components before creating new ones. Prefer extending proven modules over introducing parallel implementations. - Prefer well-maintained libraries when they reduce overall complexity or improve reliability. Do not reimplement common functionality without a clear benefit. - Keep components modular with clearly defined responsibilities. Avoid unnecessary coupling between strategy logic, execution, accounting, replay, and infrastructure. - Design for long-term maintainability once a feature or strategy has been validated. Do not over-engineer speculative ideas before evidence exists. --- ## Strategy Development - Validate hypotheses with historical replay before introducing forward-only logic whenever historical validation is possible. - Every trading strategy must progress through Replay → Shadow → Canary → Live. Do not skip validation stages. - Base design decisions on measurable evidence rather than intuition. Optimize only after demonstrating that an edge exists. - Treat every strategy as an independent contract. Do not silently alter frozen behavior without explicit authorization. --- ## Existing Systems - Do not break running Shadow or Live systems for unrelated work. - Preserve compatibility only when required by active production or validation workflows. Otherwise, remove obsolete code instead of accumulating compatibility layers. - Reuse existing infrastructure whenever possible, including replay engines, accounting, execution, wallet management, order book handling, logging, monitoring, and daemon frameworks. --- ## Engineering Standards - Prefer deterministic behavior over hidden automation. - Fail loudly when assumptions are violated. Do not silently ignore errors or fall back to unexpected behavior. - Keep configuration minimal. Introduce new configuration only when behavior genuinely needs to vary. - Remove dead code instead of leaving unused paths behind. - Write code that is easy to inspect, replay, test, and reason about. - Keep implementation consistent with existing project architecture unless an architectural change is explicitly requested. --- ## Scope Discipline - Implement only the requested scope. - Do not introduce unrelated optimizations, redesigns, migrations, or feature expansions. - Non-blocking findings outside the requested scope may be noted separately but must not be merged into the current task. - Consider a task complete once its agreed acceptance criteria are satisfied. Treat subsequent improvements as separate work items.

My triggered podcast coming up 6 pm et @rumblevideo with AJ Rice, author of new book “the curse of the bearded lady”, all about the trans mafia and woke insanity taking over the Democrat party, don’t miss it!!!

🚨 Here is the truth from Ceuta, Spain: 60,000+ migrants from Morocco’s border stormed the small town and are now hiding in the mountains and city to reach mainland Europe. Moroccan migrants said Moroccan authorities let them through their border and Spain literally opened their gates upon arrival. Many migrants thank PM Pedro Sánchez for opening the border... Ceuta is now effectively shut down due to looting concerns, and the military is losing power as migrants outnumber them and continue to escape into the mountains. Now thousands of military age men are set to spread throughout Europe due to one country’s “failure.”

RAG system that skips HTML parsing entirely! PixelRAG is an open-source visual RAG framework that renders documents as screenshots instead of parsing them into text. Most RAG pipelines start by converting HTML to text. Tables flatten into unstructured rows. Charts disappear. Layout context is gone before the LLM ever sees it. The paper measured this directly: HTML-to-text conversion accounts for 36.6% of retrieval failures on SimpleQA. PixelRAG skips that step entirely. It renders pages as screenshot tiles using Playwright, embeds those tiles with a fine-tuned Qwen3-VL-Embedding model, builds a FAISS index, and passes retrieved images directly to a VLM reader. No text abstraction in between. Benchmarked across six datasets against the strongest text-based baselines: - SimpleQA: 78.8% vs 71.6% (+7.1 points) - NQ-Tables: 48.8% vs 42.5% (+6.3 points) - EVQA: +15.5 points - LiveVQA: +11.3 points One honest caveat from the authors: this requires Qwen3-VL-4B class models or larger to see the benefit. Smaller models trail text retrieval. The authors also recommend using PixelRAG as an enhancement layer alongside existing text systems rather than a full replacement. Ships with a pre-built Wikipedia index covering 8.28M articles across 28.1M screenshot tiles. A Claude Code plugin lets Claude take screenshots of any URL and reason over the visual content directly. Key capabilities: • Renders web pages, PDFs, and images as screenshot tiles via Playwright • Fine-tuned Qwen3-VL-Embedding model for visual retrieval • FAISS index for fast vector search • Pre-built Wikipedia index: 8.28M articles, 28.1M tiles • 3x token cost reduction via image compression • Claude Code plugin for direct URL screenshot and visual reasoning • LoRA fine-tuning support via pixelrag-train 100% open source. I've shared the link in the replies!

AI模型评分都是被专项攻坚创造出来的，于是我对比了Fable5，Grok4.5， Kimi K3针对同一个交易系统审计结果进行了对比。 先说结论： Fable5：最适合作为系统级主审核模型 Kimi：最适合作为代码缺陷与一致性专项审核模型 Grok：最适合作为代码梳理和方案发散模型，不适合单独决定策略修改 最佳组合：Fable5全面审核+Grok 4.5代码梳理+K3代码审核 具体细节： 1. Fable5：系统级判断能力最强 Fable5 最大的优势不是代码读得比另外两个模型更多，而是它能把： 代码规则； sizing snapshot； intent ledger； 实际 block 统计； 当前资产 headroom； SELL/REDEEM 回流路径； 放进同一个因果框架。 它使用了几个非常关键的实盘指标： ADD 近 7 天约占新增资金 43%； 84% 资金已经部署； ETH、SOL、XRP headroom 为 0； 近 40 个周期中主要阻塞是：blocked_capital_efficiency=47 blocked_asset_cap=28 deployment cap=0 runway=0 这让它能够区分： “某个机制理论上可能限制资金” 和 “当前实盘真正正在限制资金的机制”。 最终它得出： ADD 对资金流向重要，但当前周转主因在回收端、资产 cap 和效率过滤，不在 ADD 准入本身。 这是三个模型中最接近生产系统审核要求的判断。 弱点 Fable5 仍有一些过度推断： 把 ADD 描述为让资金“锁得更久”，实际上 ADD 的剩余 TTE 通常比 ENTRY 短； 把超 cap 资产总持仓约 $382 说成可以“直接解锁 $382”，没有区分总持仓、超额部分和可成交部分； 把模型中的 redeem_lag_days=2 一度当作实际回款延迟； “$5 仓位几乎不受每美元每日利润门约束”的推理不正确，因为该指标已经按资金归一化； 2-lot 最低 ENTRY 建议可能系统性损失覆盖率。 因此，Fable5 的系统方向判断最好，但具体数字和金融指标仍需二次校验。 最适合的角色 PRIMARY_SYSTEM_REVIEWER LIVE_OPERATIONAL_DIAGNOSIS CHANGE_PRIORITY_DECISION CROSS_MODULE_ROOT_CAUSE_ANALYSIS 2. Kimi：代码缺陷侦测能力最强 Kimi 对代码结构的还原比较准确： 固定 ADD 次数和 interval 已退役； ADD 采用 target-gap 模型； ENTRY 60%，ADD 补到 100%； allocator 是最终数量权威； style 仅作诊断； 现金、集中度、shock、深度共同限制订单。 更重要的是，Kimi 找出了其他两个模型没有明确指出的具体问题： shared_deployable_pool() 读取 account_snap["capital"]["deployable_cash"] 但该字段可能没有实际写入 → 回退到 free_cash → 策略层与 allocator 层资金口径可能不一致 它还发现了： 合同写 debounce 60 秒，代码/配置为 30 秒； 注释周期 16 分钟，实际 loop 600 秒。 这些是典型的静态审核、字段追踪和合同一致性检查优势。 弱点 Kimi 在资本效率和交易语义上的推理弱于它的代码检查能力。 典型错误是： ADD 价格更高，所以边际 edge/day 必然更差。 这忽略了剩余持有时间也缩短。更高 ask 并不必然意味着更低 edge/day。 它还认为： 60/40 会让剩余资金长期闲置； 提高 entry share 会改善周转； CONFIRMATION_NO 应收紧； 增加单市场软 cap 会改善组合周转。 这些结论缺少真实候选竞争、实际 block attribution 和反事实分配数据支持。 最适合的角色 STATIC_CODE_AUDITOR SCHEMA_AND_FIELD_FLOW_CHECKER CONTRACT_IMPLEMENTATION_DIFF LOCALIZED_BUG_DISCOVERY Kimi 很适合回答： “代码是否存在字段没有写入、默认值回退、文档与实现不一致、某个 gate 实际是否生效？” 但不适合单独回答： “应该如何改变交易策略和资本分配？” 3. Grok：代码梳理最完整，但最容易过度设计 Grok 对整个 ADD 路径的整理最详尽： 各层准入条件； risk latch； REDUCE reentry cooldown； 价格带； fingerprint； emergency cap； market target； ENTRY/ADD gap； allocator 的现金、集中度、shock 和深度约束； ADD 与 ENTRY 的评分和 continuity； SELL/REDEEM 对现金回收的影响。 它对当前代码执行模型的概括非常清楚： 能不能加由 headroom 决定；加多少由 target gap 离散为 lot；ADD style 只是解释标签。 因此，在“快速理解一个陌生复杂系统”方面，Grok 表现很好。 弱点 Grok 最大的问题是： 从“发现一个可能的机制副作用”快速跳到“建议修改策略”。 它提出了大量未经实盘证明的改动： TIME_TOPUP 冷却； ADD 1.5 倍 edge/day 门槛； ask≥0.97 限制为 1 lot； 降低 peak target； 提高 entry share； 单次仅补部分 gap； 弱化 continuity； 降低 TTE confirmation 权重。 这些建议表面上都很合理，但存在三个问题： 没有先证明这些机制实际造成了损失； 没有量化被 ADD 挤出的 ENTRY 是否更优； 可能重新引入此前已经修复的低 ADD recall 和 leader fidelity 偏差。 Grok很擅长生成完整优化空间，但容易把： POSSIBLE SIDE EFFECT 升级成： CONFIRMED ROOT CAUSE 再进一步升级成： SHOULD CHANGE PRODUCTION LOGIC 这是生产交易系统审核中最危险的倾向。 最适合的角色 SYSTEM_MAPPING CODE_AND_CONFIG_EXPLANATION HYPOTHESIS_GENERATION DESIGN_OPTION_ENUMERATION 不适合作为唯一的： PRODUCTION_CHANGE_APPROVER ROOT_CAUSE_FINAL_AUTHORITY STRATEGY_SEMANTICS_GATEKEEPER 三个模型的典型思维模式 Grok 发现机制 → 推演可能副作用 → 生成多种优化 → 倾向建议修改 优点：覆盖广、思路多。 风险：过度设计、假设升级过快。 Kimi 追踪代码和字段 → 找实现不一致 → 找局部缺陷 → 尝试从缺陷推导策略改进 优点：代码问题定位强。 风险：局部正确不等于系统结论正确。 Fable5 理解代码 → 读取运行数据 → 找实际 binding constraint → 区分主因和次因 → 按实盘收益排序 优点：最接近生产运营思维。 风险：仍会在个别指标含义和金额口径上过度断言。

开源项目 LoopX：超长程 Agent 自主运行 200+ hours，状态不漂移。 我的技术主张是：LLM 上下文有限，长程 Agent 需要外置状态，通过完备的状态管理、监督和规划，让 Agent 无人干预时跑得稳、持续有产出；有人干预时跑得更好，能吸收反馈继续演进。 两条真实 trajectory 分别跨越 220.7 / 272.9 小时，跨多轮执行、等待、人工决策、writeback 与 resume 后，整个 loop 仍能找回目标、证据和下一步。 目前 LoopX 已有 3 个 showcase：auto PR issue fix、AutoML experiment 和 auto coredump fix。 以 OpenViking 开源仓库的 PR issue fix 为例，Agent 不只是循环写代码。它需要持续理解 issue 的不同状态，判断何时开发、何时等待、何时请求 review，处理 CI、冲突和上游变化，并连续交付多个 PR。 这对应 LoopX 的 domain state 管理：领域系统决定真实状态，LoopX 负责把状态投影成下一步可执行的工作。 与此同时，Agent 还可以在干活过程中实现能力自进化。当它发现现有系统缺少某项能力时，可以提出 feature、完成开发与验证、发布新的离线或在线版本，再使用新能力继续原来的任务。 长程 Agent 天然适合自进化，“完成工作”和“升级完成工作的系统”可以在同一条长程轨迹中发生。 LoopX 把这些信息外置成结构化控制面： • Goal / Vision：目标是什么，什么不能被局部优化牺牲 • Todo / Gate：当前执行的 frontier，以及必须留给人的关键判断 • Identity / Authority：谁能 claim、writeback、approve • Evidence / Receipt：每次推进留下什么可回读证据 • Quota / Scheduler：何时继续执行，何时安静等待 • Handoff / Recovery：换模型、换会话、换 host 后如何恢复 你也可以把它理解为一块专门给 Agent 设计的可执行 Kanban。 普通看板只展示“谁在做什么”；LoopX 的状态会直接约束和驱动下一次 bounded turn，让看板本身成为执行系统的一部分。 这套系统最强的地方是通用性。它不只可以修 PR，还可以做 auto research、长期实验、自媒体运营、复杂 feature 开发和办公任务。 Agent 不再只是一次性的回答机器，而可以围绕一个人的 vision，长期工作、等待、吸收反馈、积累证据并持续演进。 LoopX 从一开始 build in public：状态协议、CLI、控制面实现和真实运行轨迹都进入了开源仓库；它也已经和 OpenViking、NoKV 等 agent infra 项目形成了开源合作伙伴关系。 我希望 LoopX 最终能放大每个人的 vision 和想象力。只要你有自己的目标、想法或技术主张，就能拥有一个全天候继续工作和探索的 Agent 系统，帮助你把愿景一点点变成现实。 欢迎试用、提 issue、贡献代码，或者用一个真实的 multi-day task 跑 LoopX。

After a supposedly 111-year-old man was found to have been dead since 81 and unreported so his daughter could keep collecting his pension checks, Japanese authorities began investigating the country's oldest old. They found an insane situation.

A wallet address was previously created by a former employee, which they then used to generate a token, as part of a video tutorial. That individual is no longer with the company as part of this incident. The individual retained unauthorised access to the associated seed phrase after their departure and used it to generate a new private key. We are now aware that the same address is being used independently in connection with a new meme token. BNB Chain did not create, authorise, promote or participate in the creation of this token and has no control over the token or wallet address. These are not affiliated with or endorsed by BNB Chain. We are pursuing legal action against the former employee and cooperating with relevant authorities on this matter.

Indian authorities have suspended a policeman for firing rounds in the air from an assault rifle to disperse student protesters in an eastern state, the latest flashpoint in a wave of youth protests that forced the federal education minister to resign