2026年8月3日 技术热点总结

项目简介：An open-source, agentic-first CRM.

技术栈：TypeScript

核心介绍：CRM An open-source, agentic-first CRM. A durable research agent is the product. The database is just where it writes things down. The agent · Stack · Quick start · Configuration · Deploying · Contributing Most CRMs are a database with a form in front of it. The AI ones bolt a chat box onto the side of that form. Both leave the actual work — finding out what is true, and writing it down — to a human who has better things to do.

项目数据：⭐ 1,794 Stars，🍴 214 Forks

项目简介：Run the full 2.78-trillion-parameter Kimi K3 model beyond available RAM by streaming activated weights directly from NVMe. A dependency-free, embeddable C inference engine.

技术栈：C

核心介绍：WASTE is an embeddable inference engine written in C, with no third-party runtime dependencies. It keeps the model trunk in memory, streams selected experts directly from disk, and uses the remaining RAM as a bounded expert cache. The project is driven by humans: the ideas, hypotheses, priorities, tests, and decisions are human. The code is written by LLMs. At this scale, that is the only wa…

项目简介：Desktop app that records your on-screen work session and uses the GitHub Copilot CLI to reconstruct it as an intent + ordered steps, then builds a reusable Skill or Automation for Microsoft Scout, Microsoft Copilot Cowork, or Copilot Studio.

技术栈：TypeScript、agent-skills、ai-agents、automation、copilot、copilot-cli、copilot-cowork、copilot-studio、electron

核心介绍：Skill Recorder captures a real work session on your screen: the clicks, the app and window switches, the pages you visit, and (if you want) your spoken narration. It then uses the GitHub Copilot CLI to reconstruct *what…

项目简介：Codex-native architect orchestration with Luna and Terra implementation lanes and mandatory fresh Sol review.

技术栈：Shell

核心介绍：with a requested read-only profile stands between the diff and done.** Sol Advisor is a Codex-native architect workflow for capability-routed software delivery. The primary session stays focused on requirements, architecture, specs, and verification while native Codex custom-agent threads handle implementation and review.

项目数据：⭐ 647 Stars，🍴 41 Forks

项目简介：33 ways to spot AI-written text, right in your terminal. Before/after examples, draft checker, zero dependencies.

技术栈：JavaScript、ai-slop、ai-slop-detection、ai-slop-detector、ai-slop-fixer、ai-slop-remover、claude-code-plugin、claude-skills、codex

核心介绍：No network, no API key, no dependencies. One 87 KB C program. windows x64 · unpack and run · all releases A terminal reference for the humanizer skill, which collects the patterns catalogued in Wikipedia’s Signs of AI writing. Thirty-three habits give away text produced by a language model: dashes where a comm…

Existing token compression methods for omnimodal large language models typically rely on one modality to determine what to retain in the other. We show that this assumption often breaks down: for the same query, audio and video relevance often peaks at different moments. This cross-modal salience mismatch makes unidirectional guidance prone to discarding answer-critical cues under aggressive compression. We propose OmniScope, a training-free token compression framework that uses the query as a shared semantic anchor while estimating relevance separately for audio and video. OmniScope alloca…

On-policy self-distillation (OPSD) is a promising approach to improve reasoning language models, but it remains brittle in practice: making it work reliably often requires substantial engineering effort. We identify a structural source of this difficulty: vanilla OPSD is precisely the β=1 member of a broader policy-optimization family, where β weights the KL penalty anchoring the student to a reference policy. This equivalence turns β from an implicit value fixed at one into a controllable regularization parameter, yielding a more general formulation that trades off proximity to a reference…

Sparse mixture-of-experts (MoE) language models route each token to multiple experts, suggesting a geometric account of their benefit: co-selected experts should contribute distinct representation directions. Existing evidence often conflates route coherence, candidate quality, and candidate-by-context interaction. We distinguish these quantities using an Expert Subspace Separation Index (ESSI), matched-route residuals, and a prefix-controlled 2times2 factorial; frozen-route interventions and a controlled Top-k study assess functional value. Three paired contrasts organize the findings. Fir…

This work presents Fairness Pruning, a lightweight structural intervention method designed for the management and future mitigation of demographic bias in large language models (LLMs). As a foundational empirical validation of this method, this work focuses on causal bias localization. Using minimally contrastive prompt pairs and inference-time activation capture, the method identifies neurons that react differentially when processing demographic attributes in GLU architectures, evaluating the signal at the down_proj input. Empirical evaluation was conducted on models of up to 3 billion par…

Memory is central to long-horizon LLM agents, yet existing memory systems primarily preserve interaction content rather than modeling which agents can be trusted and under what conditions. This limitation is particularly important in multi-agent systems, where a central model may be unable to directly verify plausible or correlated peer responses. We introduce Σ-Mem, an online reliability memory that records historical competence evidence for individual peers and peer relationship evidence across the peer set. Both forms of evidence are maintained as real symmetric states and updated from p…

Multimodal large language models increasingly use sketches, annotations, tools, and intermediate images during reasoning, but it remains unclear whether they truly rely on these visual states. Existing benchmarks are limited both by task collections with narrow coverage or partially text-solvable samples and by evaluations that emphasize final answers without diagnosing how intermediate visual states are generated, rendered, and used. We introduce See2Think, a unified evaluation framework comprising See2ThinkBench and Visual Action-of-Thought (VAoT). See2ThinkBench contains 1,200 open-ended…