Scoutari

AI builder 的決策情報台。

瀏覽

  • 今日情報
  • Skill 榜單
  • AI 週報
  • 關於與編輯方針
  • RSS

主題

  • 模型發布
  • API 與價格
  • MCP 與 Skills
  • AI Coding
  • 中國模型
  • 開源模型
  • Agent
  • Infra / 成本
  • 值得試用
© 2026 Scoutari · scoutari.com摘要由 AI 輔助生成,並始終附上原文連結。
情報收藏週報設定
SScoutariAI Builder Intel · decision desk
Skill RadarSavedPersonalizeSubmit sourceLog inSign up
Sign up
Skill RadarSavedPersonalizeSubmit sourceLog in

Research & papers

Notable papers, research breakthroughs, and model benchmarks.

Model releasesAPI & pricingMCP & SkillsAI codingChinese modelsOpen sourceAgentsInfra & costMultimodal & imageAI videoVoice AIAI hardwareRobotics & embodiedAI safetyEnterprise AIAI fundingPolicy & regulationWorth trying
  • Mon, July 6Open Source

    ECC Launches Agent Performance Optimization Framework for Claude Code, Codex, Cursor, etc.

    GitHub trending project ECC launches as an agent performance optimization system, offering skills, instincts, memory, and security.

  • Fri, July 3Research

    Using DSPy to Evaluate and Improve Datasette Agent SQL System Prompt

    A study uses DSPy to evaluate and improve Datasette Agent's SQL system prompt.

  • Thu, July 2Research

    AIEWF Conference: Autonomous AI Research Tools Spark Debate on Human Agency

    Speakers at AIEWF opposed the software factory vision, emphasizing the need to preserve human understanding and control.

  • Thu, July 2Agent

    Autoresearch: Feedback Loop Behind Self-Improving Agents

    Introspection co-founder Roland Gavrilescu explains autoresearch, agent 'recipes', self-improvement loops, and the central role of humans in the software factory.

  • Wed, July 1Model/API

    Google AI Launches TabFM: Hybrid Attention Table Foundation Model for Zero-Shot Classification and Regression

    Google Research released TabFM, a foundation model for tabular data that enables zero-shot classification and regression via in-context learning.

  • Wed, July 1Model/API

    Anthropic Claude Sonnet 5 vs Sonnet 4.6 vs Opus 4.8: Agentic Coding Benchmarks, API Pricing & Value

    Anthropic's Claude Sonnet 5 narrows the gap with Opus 4.8 on agentic coding tasks while maintaining lower Sonnet-tier pricing.

  • Wed, July 1Tools

    ScarfBench: Benchmarking AI Agents for Enterprise Java Framework Migration

    ScarfBench is a new tool for benchmarking AI agents on enterprise Java framework migration tasks.

  • Wed, July 1Tools

    shot-scraper video: Let coding agents auto-record web operation demos

    shot-scraper 1.10 introduces the 'shot-scraper video' command to automatically record web operation demo videos from a storyboard.yml file.

  • Tue, June 30Research

    Together AI Showcases Full-Stack Research at ICML 2026

    Together AI publishes eight full-stack papers at ICML 2026, exhibiting at booth B714 in Seoul.

  • Tue, June 30Research

    OpenAI Launches GeneBench-Pro Benchmark for AI in Genomics and Science

    OpenAI releases GeneBench-Pro, a new benchmark using complex real-world datasets to test AI performance in genomics, biology, and scientific research.

  • Wed, June 24Tools

    Nous Research Adds /learn to Hermes Agent Skill System: Capture Workflows as Slash Commands Without Hand-Writing SKILL.md

    Nous Research adds /learn to Hermes Agent, auto-generating compliant SKILL.md files from local directories, doc URLs, past conversations, or notes—no manual writing or ingestion engine needed.

  • Wed, June 24Research

    DFlash Speculative Decoding: Parallel Token Blocks Boost Throughput Up to 15x on NVIDIA Blackwell

    UC San Diego's DFlash replaces autoregressive draft generation with a lightweight block diffusion model, generating entire token blocks in a single forward pass for accelerated speculative decoding.

  • Wed, June 24Tools

    OPFS + Pyodide Test Tool: Edit Persistent SQLite Files in Browser

    Simon Willison's test tool explores using the browser's Origin Private File System to edit local SQLite files via Datasette Lite.

  • Tue, June 23Tools

    Sony's AI Camera Assistant Produces Terrible Photos

    Sony Xperia 1 VIII's AI Camera Assistant delivers some of the worst photos in years.

  • Sun, June 21Tools

    Nous Research Adds Blank Slate Mode to Hermes Agent via platform_toolset CLI and disabled_toolset

    Nous Research adds Blank Slate mode to its open-source Hermes Agent, starting with only provider, model, file operations, and terminal, with other toolsets requiring user opt-in.

  • Fri, June 19Tools

    Datasette Apps: Host Custom HTML Apps Inside Datasette

    New plugin lets users host sandboxed HTML+JavaScript apps within Datasette instances.

  • Thu, June 18Model/API

    OpenAI Launches LifeSciBench: 750 Biology Tasks to Evaluate AI Models

    OpenAI introduces LifeSciBench, a benchmark assessing AI models in real-world life science research.

  • Thu, June 18Model/API

    China Z.ai Lab Releases GLM-5.2, the Strongest Open-Source Text LLM

    Z.ai Lab open-sourced GLM-5.2, a 753B parameter text-only LLM, on June 16.

  • Tue, June 16Agent

    Hermes Agent Adds Async Sub-Agents, Delegation No Longer Blocks Parent Dialog

    Nous Research's Hermes Agent adds async sub-agent feature, delegation tool spawns background agents without blocking parent dialog.