2026年9月29日 技术热点总结

项目简介：Find code by asking what it does. A CLI for coding agents that uses Jev to discover relevant files and source context.

技术栈：TypeScript、ai-sdk、claude-code、cli、code-search、codex、coding-agents、context-retrieval、developer-tools

核心介绍：Find code by asking what it does. In our ten-task SWE-bench comparison, Jevgrep successfully completed the same 8 of 10 tasks as the baseline, at lower cost. Coding agents spend part of every unfamiliar task finding the right files. Jevgrep gives them a place to start: ask a repository question, and jg returns relevant files, reading leads, and ver…

项目简介：Qwen3.8-Flash-Next (125B MoE) on a 8GB+ NVIDIA GPU: one-click install for Windows / Linux. Strata inference engine, OpenAI/Anthropic API on localhost, optional image input.

技术栈：C++

核心介绍：Strata Run a 125-billion-parameter AI model on a normal gaming PC one NVIDIA card (12-24 GB) + 64 GB of RAM · Windows or Linux · one click to install A voxel pagoda garden, 1 shot prompt running on an RTX 5070 with Strata (IQ3_S, 128K context) · full video (49 s) Strata runs Qwen3.8-Flash-Next – a large, smart AI model that normally needs a server – on your own PC. It writes its answe…

项目简介：Run open decision models locally: pull and serve Laya, decider, NLI and GLiClass behind a TypeSafe-compatible API. Ollama for decision models.

技术栈：Rust、calibration、classification、decision-models、gliclass、jev、laya、llm-routing、local-inference

核心介绍：ollaya Run open decision models locally, the way Ollama runs LLMs. Website · Models · Docs · Releases · Hugging Face A decision model reads a *state* (a message, an email, a ticket, any JSON) plus typed questions (choice, score, noul) and returns calibrated probabilities in a single forward pass, in milliseconds. It never generate…

项目简介：Motion films that never cut to the next slide: every beat grows out of the one before, one continuous camera, continuity measured by an oracle. A Claude Agent Skill for product launch films and feature demos.

技术栈：Python、agent-skill、animation、canvas、claude-skill、launch-video、motion-graphics、video

核心介绍：> Motion films that never cut to the next slide. A Claude Agent Skill that makes product launch films, teasers and feature demos where every beat grows out of the one before — one continuous take, not a stack of scenes. > 一镜到底的连贯动效。 做产品发布片、预告片、功能演示：每一个画面都从上一个画面里长出来，是一…

项目简介：Open-sourcing our company brain – A teammate in your Slack that remembers everything your team says, and can go do the work.

技术栈：TypeScript

核心介绍：Company Brain A teammate in your Slack that truly knows and understands your company, and can do anything. Deploy · User guide · Permissions · Use cases · Discord Used to be a paid product with thousands of users. Now it’s free and open source. Read the announcement → A few weeks ago, we discontinued our Company Brain product at supermemory. This is that product, the whole thing, open sourced and rebuilt to run on your own Cl…

On-policy distillation (OPD) corrects a student on the responses it writes, but its signal is the teacher's next-token distribution: it tells the student what the teacher says but misses how it thinks. Latent supervision promises the missing part by aligning the student's latent states to the teacher's. Recent methods such as OPRD bring this signal into on-policy distillation. However, we observe two failures of this recipe when distilling Qwen3-4B and Qwen3-8B into Qwen3-1.7B-Base. Early gain, late collapse: latent supervision alone lifts MATH-500 accuracy from 25 to 46 in 10 steps, but su…

Time series agents answer analytical questions by calling external tools, and which tools they carry is decided by people before the agent runs. However, we identify two failures in this setup. Human-Agent Tool Misalignment: a library of 21 expert-curated tools helps on some tasks and hurts on others, dropping anomaly accuracy under every backbone we test. Silent Harm: one round of generic self-revision changes 147 answers and breaks 56 of them, while the final score moves by less than a point. Both follow from the same gap: whether a tool helps is decided question by question at runtime, w…

Merging low-rank adapters (LoRAs) promises to eliminate the overhead of swapping task-specific weights at inference time. However, existing merging methods assume every layer needs the same rank budget. Further, some methods assume that rank budget needs to be split equally among the tasks too. We show this uniform-budget assumption is a major source of the performance gap between merged and per-task LoRAs. However, rank selection is an NP hard problem. To this end, we introduce Net Utility, a data free metric that first decomposes every task LoRA by its Singular Value Decomposition (SVD) a…

Latent world models predict the consequences of actions, but accurate prediction does not guarantee that latent distance reflects which candidate will execute successfully. We identify a decision-local prediction gap: among the few futures competing for execution, a candidate predicted closer to the goal can produce a worse realized outcome than an available alternative. We introduce D-JEPA, a decision-aligned latent world model that learns decision-relevant relations among candidate futures from executed outcomes. A bounded, permutation-equivariant operator jointly reasons over goal-relati…

Prefill and decode reward different approaches to quantization: low-precision arithmetic accelerates prompt processing, while compact weights reduce memory traffic during generation. We propose "disaggregated quantization" (DQ), which specializes computation formats, weights and storage placement to both of these phases. On Qwen 3 and Gemma 3, removing activation quantization specifically on decode improves accuracy on decode-heavy tasks without increasing inference cost. Training separate compute-native prefill weights accelerates prompt processing relative to weight-only inference while m…

We present LightMIS, a scalable family of ultra-lightweight convolutional networks for 2D binary medical image segmentation without a learned stage-wise decoder. LightMIS aligns the outputs of a five-level encoder to a common resolution using Scale-Aligned Projection blocks, aggregates them once, and refines the fused representation with an Adaptive Fusion Cascade. The cascade combines Adaptive Kernel Fusion with the proposed Progressive Receptive Fusion module, which uses temporary channel expansion, complementary depthwise receptive fields, and progressive cross-branch information transfe…