每周科技速递 - 2026-06-08
每周科技速递 - 2026-06-08
本周科技圈焦点：系统提示词大规模泄露项目引爆 GitHub，突破 4.1 万星标；自进化 Agent 框架 GenericAgent 以 6 倍 Token 效率降低持续高热；64M 参数小模型 minimind 实现 2 小时从零训练；Gary Marcus 尖锐批评 AI slop 侵蚀生产力；比特币逼近 $60,000 关口，机构情绪显著转向谨慎。
🧠 LLM / 大模型
- 🔥12 | jingyaogong/minimind — 🧠「大模型」2小时完全从0训练64M的小参数LLM！Train a 64M-parameter LLM from scratch in just 2h!
- 🔥8 | sgl-project/sglang — SGLang is a high-performance serving framework for large language models and multimodal models.
- 🔥8 | ScrapeGraphAI/Scrapegraph-ai — Python scraper based on AI
- 🔥8 | yamadashy/repomix — 📦 Repomix is a powerful tool that packs your entire repository into a single, AI-friendly file. Perfect for when you nee
- 🔥6 | NVIDIA/TensorRT-LLM — TensorRT LLM provides users with an easy-to-use Python API to define Large Language Models (LLMs) and supports state-of-
- 🔥6 | GreyDGL/PentestGPT — Automated Penetration Testing Agentic Framework Powered by Large Language Models
- 🔥6 | 0xPlaygrounds/rig — ⚙️🦀 Build modular and scalable LLM Applications in Rust
- 🔥6 | skyzh/tiny-llm — A course of learning LLM inference serving on Apple Silicon for systems engineers: build a tiny vLLM + Qwen.
- 🔥5 | ollama v0.30.7-rc1 — What's Changed docs: update Zod example to use native toJSONSchema by @fuleinist in https://github.com/ollama/ollama/pul
- 🔥5 | bitsandbytes-foundation/bitsandbytes — Accessible large language models via k-bit quantization for PyTorch.
- 🔥5 | open-compass/opencompass — OpenCompass is an LLM evaluation platform, supporting a wide range of models (Llama3, Mistral, InternLM2,GPT-4,LLaMa2, Q
- 🔥5 | guardrails-ai/guardrails — Adding guardrails to large language models.
- 🔥5 | ModelTC/LightLLM — LightLLM is a Python-based LLM (Large Language Model) inference and serving framework, notable for its lightweight desig
- 🔥5 | thu-pacman/chitu — High-performance inference framework for large language models, focusing on efficiency, flexibility, and availability.
- 🔥5 | stanford-crfm/helm — Holistic Evaluation of Language Models (HELM) is an open source Python framework created by the Center for Research on F
🤖 AI Agent
- 🔥13 | lsdefine/GenericAgent — Self-evolving agent: grows skill tree from 3.3K-line seed, achieving full system control with 6x less token consumption
- 🔥11 | CherryHQ/cherry-studio — AI productivity studio with smart chat, autonomous agents, and 300+ assistants. Unified access to frontier LLMs
- 🔥11 | EvoMap/evolver — The GEP-powered self-evolving engine for AI agents. Auditable evolution with Genes, Capsules, and Events. | evomap.ai
- 🔥10 | openclaw
v2026.6.5-beta.2 — Highlights - QQBot now strips model
reasoning/thinking scaffolding before native delivery, preventing raw
con - 🔥8 | openclaw v2026.6.1 — Highlights - Agents and CLI-backed runtimes recover more cleanly from interrupted tool calls, stale session bindings, co
- 🔥7 | Xiangyue-Zhang/auto-deep-researcher-24x7 — 🔥 An autonomous AI agent that runs your deep learning experiments 24/7 while you sleep. Zero-cost monitoring, Leader-Wor
- 🔥6 | Upsonic/Upsonic — Build autonomous AI agents in Python.
- 🔥6 | skalesapp/skales — Local-first AI desktop agent for Windows, macOS, Linux & Android. Set a goal, it works on its own. Teams (pair two deskt
- 🔥5 | Build tools, to build more —
- 🔥5 | Opus 4.8 —
- 🔥5 | stakpak/agent — Ship your code, on autopilot. An open source agent that lives on your machines 24/7 and keeps your apps running. 🦀
- 🔥5 | grinev/opencode-telegram-bot — OpenCode mobile client via Telegram: run and monitor AI coding tasks from your phone while everything runs locally on yo
- 🔥5 | BlockRunAI/Franklin — The AI agent with a wallet — spends USDC autonomously to get real work done. Apache-2.0, TypeScript.
- 🔥5 | gmickel/flow-next — Spec-driven AI workflow plugin for Claude Code, OpenAI Codex, and Factory Droid. Zero-dep task tracking, worker subagent
- 🔥5 | swarmclawai/swarmclaw — Open-source self-hosted AI agent runtime and multi-agent framework for autonomous agent swarms. Agent memory, MCP tools,
💰 加密货币
- 🔥7 | A quick review of the Ways and Means tax bills: State of Crypto —
- 🔥7 | Michael Saylor revives bitcoin-buy speculation as scrutiny over Strategy grows —
- 🔥7 | Bitcoin near $60,000 today vs February: Institutional sentiment has flipped —
- 🔥7 | foundry nightly-43ed279745c5decd51dfc60eaaba4b0911a30e9c — Other - fix: redact private keys from Debug (#15091) by @stevencartavia - refactor(script): isolate Tempo session signer
- 🔥5 | OffcierCia/DeFi-Developer-Road-Map — DeFi Developer roadmap is a curated Developer handbook which includes a list of the best tools for DApps development, re
- 🔥5 | lifinance/contracts — LI.FI Smart Contracts
- 🔥5 | exactly/protocol — exactly is a decentralized, non-custodial and autonomous marketplace for depositing and borrowing crypto assets at a var
🔬 前沿科技
- 🔥15 | asgeirtj/system_prompts_leaks — Extracted system prompts from Anthropic - Claude Code, Claude Design, Opus 4.8, Sonnet 4.6. OpenAI - ChatGPT 5.5 Thinkin
- 🔥14 | rasbt/LLMs-from-scratch — Implement a ChatGPT-like LLM in PyTorch from scratch, step by step
- 🔥10 | tensorflow/tensorflow — An Open Source Machine Learning Framework for Everyone
- 🔥10 | huggingface/transformers — 🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and
- 🔥9 | ultralytics/ultralytics — Ultralytics YOLO 🚀
- 🔥8 | roboflow/supervision — We write your reusable computer vision tools. 💜
- 🔥7 | Firefox Merges Support for Vulkan Video Decoding —
- 🔥7 | VibeOS: First ever AI-native operating system —
- 🔥7 | Show HN: Nightwatch, The open-source, read-only AI SRE —
- 🔥7 | How's Linear so fast? A technical breakdown —
- 🔥7 | If LLMs Have Human-Like Attributes, Then So Does Age of Empires II —