2026年7月16日 技术热点总结

项目简介：SpaceXAI’s coding agent harness and TUI. Fullscreen, mouse interactive, extensible.

技术栈：Rust

核心介绍：Grok Build (grok) full-screen TUI that understands your codebase, edits files, executes shell commands, searches the web, and manages long-running tasks — interactively, headlessly for scripting/CI, or embedded in editors via the Agent Client Installing the released binary · Building from source · Documentation · Repository layout · Development · Contributing · This repository contains the Rust source for the grok CLI/TUI and its agent

项目数据：⭐ 2,115 Stars，🍴 297 Forks

项目简介：JZSub — 一条视频链接，自动交付最高画质、封面和 GPT 双语字幕 MP4。

技术栈：Python、bilingual、codex、subtitles、video、yt-dlp

核心介绍：给 JZSub 一个视频链接，拿回最高画质视频、封面、双语字幕和烧录完成的 MP4。 YouTube Bilibili Vimeo Twitch Dailymotion TikTok 以及 yt-dlp 支持的其他站点。实际可用格式、字幕和登录要求由平台决定。

关键特性：最高画质：下载最佳视频与音频，尽可能无损封装 MP4；按需交付：–deliver 可选完整烧录、仅视频（含原始字幕文件）、仅原始字幕、或双语字幕文件；双语字幕：紧凑分批翻译、批间上下文衔接，再按句子边界重新切分；原文保持不变。默认译成中文，–target-lang 可指定日语、法语等任意目标语言；自动烧录：MiSans 字幕、底部双语堆叠、自适应竖屏、精确背景、libass 渲染；低耗运行：只读取紧凑字幕文档，不读取完整清单；烧录每 5% 显示一次进度。

项目数据：⭐ 674 Stars，🍴 74 Forks

项目简介：Design clear, theme-specific GitHub README homepages with SVG titles, real proof, and maintainable Markdown

技术栈：Python、agent-skill、codex-skill、github-readme、readme-design、svg

核心介绍：English · 简体中文 These are not hypothetical templates. The method is already used by four public repositories, each with its own visual language and content structure: If this Skill helped you create a public README you are proud of, you are welcome to propose it for this list in a PR. This is completely optional: the footer signature is appreciated but never required, and showcase submissions …

技术栈：Rust

核心介绍：Telegram: https://t.me/CluvexStudio Aether is a censorship circumvention client designed for heavily restricted networks. It automatically discovers reachable routes, establishes an encrypted tunnel, and exposes a local SOCKS5 proxy for your applications. Unlike traditional VPN clients, Aether is built for environments where Deep Packet Inspection (DPI), protocol fingerprinting, UDP throttling, and endpoint blocking are common. Prebuilt binaries are available on the Releases page for:

关键特性：Automatic endpoint discovery, with end-to-end data-plane validation so a g…

项目简介：A new open-source format for interactive video on the web, with a built-in state machine, frame-accurate transitions, and packed-alpha transparency.

技术栈：TypeScript

核心介绍：AVAL is a web-only format and runtime for short prerendered animation with continuous partial loops, user-defined states, authored triggers, bounded transitions, packed transparency, and host-owned fallback markup. The central idea is simple: encode independently decodable motion units and a small deterministic state graph in one .avl asset. The browser keeps a decoder timeline moving forward across a loop…

Coding agents must integrate external tool returns into ongoing reasoning – a capability that standard left-to-right pretraining on code exposes only in its forward direction. We observe that the action-observation-continuation loop of a coding agent is structurally isomorphic to a function call site, where a caller binds arguments, a callee returns a value computed elsewhere, and downstream code consumes that value. This conditioning structure exists at internet scale in ordinary code. We exploit it through function-aware fill-in-the-middle (FIM) mid-training: a self-supervised objective t…

Large language models fine-tuned for forecasting can be accurate yet poorly calibrated, and their chain-of-thought (CoT) reasoning may not faithfully reflect the evidence behind a forecast. We ask whether internal representations offer a more direct window into both. Working with Eternis-Forecaster 8B on OpenForesight, we train representation-pooling probes on intermediate activations and find they achieve substantially better calibration; a result that also holds for GLM-4.7-Flash and GLM-4.5-Air. We then assess CoT faithfulness through evidence ablation and diversionary injection: removin…

Vision language models (VLMs) have achieved strong performance on visual document understanding benchmarks such as DocVQA, ChartQA, and MMLongBench-Doc. However, real-world documents combine multiple factors such as length, layout complexity, modality, and question difficulty, which makes it difficult to attribute model failures to specific causes. We introduce SynthDocBench, a fully synthetic benchmark for long-context visual document understanding that systematically controls factors including document length, layout structure, modality composition, and question type. The benchmark is con…

Visual generators excel at rendering, but they confidently fabricate what they do not know. User requests are unbounded, evolving, and deeply long-tailed: new characters, trending entities, post-cutoff events, and more. This world-knowledge bottleneck is structural: generators are trained on fixed corpora, but the visual world is open-ended. We construct SearchGen-20K and SearchGen-Bench, with 20,839 prompts spanning twelve failure categories and twenty-two domains, paired with a pre-executed multimodal SearchGen-Corpus-1M to support offline, reproducible research. On SearchGen-Bench, front…

Mainstream visual encoders are pretrained on natural images and cannot be effectively applied to document images without document-oriented adaptation, as dense text and fine-grained character strokes demand character-level visual perception. We present MonkeyOCRv2, a visual-text pretrained model for document AI. First, we construct MonkeyDoc v2, to our knowledge the largest document-image pretraining corpus, comprising 113 million images spanning 17 languages. Second, we propose a pretraining strategy that jointly learns image-to-text generation and pixel-level document reconstruction: the …

This work introduces a unified formulation for vision models, where diverse forms of visual information beyond natural images, such as masks, depth maps, and other structured visual signals, are all represented as RGB images, while general visual tasks can be converted into a common RGB-to-RGB image editing problem. In this paradigm, different types of visual information internally share the same encoding and decoding architecture and parameters as natural images, enabling a single model to transfer across tasks through a unified visual interface, in a way analogous to how language models o…