2026年7月31日 技术热点总结

项目简介：Open-source, self-hosted revenue-first analytics for founders: web analytics, Session Replay, revenue attribution, and customer revenue integrations. datafast alternative

技术栈：TypeScript、analytics、bing、datafast、dodopayments、google-search、lemonsqueezy、polar、product-analytics

核心介绍：This repository contains the focused open-source edition of Talivia, the revenue-first analytics platform available at talivia.com. The self-hosted edition combines core web analytics, Session Replay, website collaborators, shared analytics, import/export, and customer revenue from Stripe, LemonSqu…

项目简介：Run full Kimi K3 on a single device. And an OpenAI-compatible API server for local chat and coding agents.

技术栈：Python、kimi、kimi-k3、local-ai、local-llm

核心介绍：____ _ _ __ _ Deltafin is a small research project that runs a Mixture-of-Experts model far larger than the machine it sits on. It supports Apple Silicon macOS and x86-64/aarch64 Linux, automatically selecting MPS, CUDA or CPU for the resident model path and a qualified Metal, CUDA or CPU MXFP4 expert kernel. The maintainer-run exact-oracle reference is now 0.2660 token/s (3.76 seconds/token median) on…

项目简介：An explorable 3D city that shows how Postgres actually works

技术栈：TypeScript

核心介绍：PGSimCity turns a PostgreSQL cluster into a city you can inspect, walk through, and break. It is for engineers who are good at their job and have never had to operate a database — the people who need to understand why a checkpoint spikes latency, why one forgotten transaction bloats a table forever, and what synchronous_commit is really charging them. PGSimCity is an independent, non-commercial educational visualization of PostgreSQL internals. It is not affiliated with, sponsored, endorsed, …

项目简介：🎬 Open source, transcript-based video/audio editor that lives in the browser.

技术栈：TypeScript、audio、media、media-editing、media-editor、transcribe、transcription、video、video-editing

核心介绍：Edit video and audio like you edit text — fully offline, on your device. Rescript is an open-source, transcript-based media editor. Drop in a video or audio file and it is transcribed locally with per-word timestamps and speaker labels. Delete words in the transcript and the corresponding clip is cut from the media. Export the final cut — without your file ever leaving your device. See the Rel…

技术栈：Lua

核心介绍：A mod for the [Pokémon Gen 1 Recompilation Project](https://github.com/bryanthaboi/pokemon-gen1-recomp-project). The overworld as a 3D diorama. Terrain is extruded into real geometry, occlusion comes from a depth buffer rather than a y-sort, characters stand as leaning sprite slabs, a shadow map throws real cast shadows across whatever they land on, and an optional tilt-shift pass sells the miniature-model look. And battles fought on that world rather than on a white field. When

项目数据：⭐ 389 Stars，🍴 13 Forks

LLM-based agents excel at software engineering tasks where an existing codebase provides context, but constructing a program from scratch remains fundamentally harder. Recent benchmarks such as ProgramBench quantify this gap: given only natural-language documentation and an execute-only binary as a behavioral oracle, even frontier models solve fewer than 1% of instances. Existing frameworks conflate documentation reading, behavioral exploration, and code synthesis into a single pass, causing agents to probe insufficiently, lose behavioral intent as context drifts, and propagate early misint…

Coding agents have made substantial progress on software engineering tasks that modify existing codebases, including bug fixing and feature implementation. However, constructing a complete program from scratch remains a major challenge: even the frontier models evaluated on ProgramBench fully resolve fewer than 1% of tasks. One obstacle is the lack of scalable training environments for this from-scratch setting, spanning the whole software engineering life cycle, as existing environment-construction frameworks focus only on a single phase in software development. To address this gap, we int…

Existing autoregressive video distillation methods commonly adopt a Distribution Matching Distillation (DMD)-based multi-stage pipeline. However, they typically decouple the initialization and DMD stages — which then pursue different target distributions — and judge the intermediate student mainly by visual scores such as VBench. In this paper, we revisit this design from a distributional perspective. Given the mode-seeking nature of the distribution matching loss, a good initialization should match the mode coverage of the target DMD teacher, rather than merely pursuing high quality. To …

We present Voice Memory, a inference-only scheme for agentic speech recognition: at stream time, a frozen corrector reads a single per-domain memory.md and decides per utterance whether to act on the hypothesis or abstain and keep the 1-best. Asynchronously, a score-gated optimizer revises that file through bounded edits, accepting an edit only when it strictly improves a held-out score. Extended from classical ASR-LM framework, we refer this split the listener-thinker architecture; the two roles are coupled only through the memory, so no weights change and the learned skill stays auditable…

Generalist manipulation policies increasingly take the form of action-chunking flow policies built on large pretrained backbones. Such chunks run open-loop, so the policy cannot react to sensory input arriving mid-execution, sacrificing reactivity. Replanning more often would restore it, but the perception-to-action pipeline (a large backbone plus multiple denoising steps) is too slow: this latency forbids frequent replanning and leaves committed actions stale, making such policies ill-suited for dynamic, closed-loop control. We present πR^2, which makes these policies reactive and real-tim…

Memory has evolved into a foundational architectural dimension in large language models (LLMs), shifting from an implicit byproduct of computation to a spectrum of explicit, controllable mechanisms. While recent advances introduce diverse strategies—spanning transient attention, recurrent state dynamics, parameter-efficient adaptations, and scalable lookup storage—this rapid evolution has led to a highly fragmented research landscape. In this survey, we present a systematic, architecture-centric taxonomy of memory in LLMs. Our framework characterizes memory along three orthogonal axes: …