Field notes · research archive · daily briefings Mud's Blog

News

Daily AI News Digest — 2026-03-29

GitHub velocity is led by framersai/agentos; paper attention is clustering around SlopCodeBench: Benchmarking How Coding Agents Degrade Over Long-Horizon Iterative Tasks; social attention is tilting toward The incident that blamed AI for the Iranian school bombing is a more fundamental problem; biggest mover: AVO: Agentic Variation Operators for Autonomous Evolutionary Search (+49). 10 repo signals, 10 paper picks, and 10 community items made today's cut.

Raw feed JSON Digest archive Monthly archive
Issue date
Generated
Sections 4

Signal Board

Repo momentum board

Local signal score blends freshness, feed rank, keyword relevance, and GitHub star velocity.

01 framersai/agentos GitHub · 37 stars · 37/day · 0h ago · signal 9.40
9.40
02 GeminiLight/MindOS GitHub · 113 stars · 113/day · 0h ago · signal 9.34
9.34
03 LearningCircuit/local-deep-research GitHub · 4196 stars · 4196/day · 0h ago · signal 9.26
9.26
04 vllm-project/guidellm GitHub · 952 stars · 952/day · 0h ago · signal 8.83
8.83
05 NVIDIA/TensorRT-LLM GitHub · 13200 stars · 13200/day · 0h ago · signal 8.69
8.69
06 cloud-apim/otoroshi-llm-extension GitHub · 16 stars · 16/day · 0h ago · signal 8.69
8.69
07 Tom-Ryder/AgentRunKit GitHub · 13 stars · 13/day · 0h ago · signal 8.66
8.66
08 edwardcapriolo/deliverance GitHub · 13 stars · 13/day · 0h ago · signal 8.65
8.65

Highlights

Top signals

framersai/agentos — TypeScript runtime for autonomous AI agents — multimodal RAG, cognitive memory, streaming guardrails, voice pipeline, and emergent multi-agent orchestration.

framersai/agentos

 SlopCodeBench: Benchmarking How Coding Agents Degrade Over Long-Horizon Iterative Tasks

SlopCodeBench: Benchmarking How Coding Agents Degrade Over Long-Horizon Iterative Tasks

 The incident that blamed AI for the Iranian school bombing is a more fundamental problem

The incident that blamed AI for the Iranian school bombing is a more fundamental problem

 Calibri: Enhancing Diffusion Transformers via Parameter-Efficient Calibration

Calibri: Enhancing Diffusion Transformers via Parameter-Efficient Calibration

Section

Hot in 24 Hours

The fastest-moving items across repos, papers, and community chatter.

framersai/agentos — TypeScript runtime for autonomous AI agents — multimodal RAG, cognitive memory, streaming guardrails, voice pipeline, and emergent multi-agent orchestration.

framersai/agentos

 Demorck/ClairObscurArchipelagoRandomizer — A Clair Obscur : Expedition 33 mod for the Archipelago multi-game randomizer system

Demorck/ClairObscurArchipelagoRandomizer

 SlopCodeBench: Benchmarking How Coding Agents Degrade Over Long-Horizon Iterative Tasks

SlopCodeBench: Benchmarking How Coding Agents Degrade Over Long-Horizon Iterative Tasks

 AVO: Agentic Variation Operators for Autonomous Evolutionary Search

Section

Repository Momentum

Fresh GitHub projects worth scanning before the feed turns over.

framersai/agentos — TypeScript runtime for autonomous AI agents — multimodal RAG, cognitive memory, streaming guardrails, voice pipeline, and emergent multi-agent orchestration.

framersai/agentos

 GeminiLight/MindOS — MindOS is a Human-AI Collaborative Mind System. Globally sync your mind for all agents: transparent, controllable, and evolving symbiotically.

GeminiLight/MindOS

 LearningCircuit/local-deep-research — Local Deep Research achieves ~95% on SimpleQA benchmark (tested with GPT-4.1-mini). Supports local and cloud LLMs (Ollama, Google, Anthropic, ...). Searches 10+ sources - arXiv, PubMed, web, and your private documents. Everything Local & Encrypted.

LearningCircuit/local-deep-research

 vllm-project/guidellm — Evaluate and Enhance Your LLM Deployments for Real-World Inference Needs

vllm-project/guidellm

 NVIDIA/TensorRT-LLM — TensorRT LLM provides users with an easy-to-use Python API to define Large Language Models (LLMs) and supports state-of-the-art optimizations to perform inference efficiently on NVIDIA GPUs. TensorRT LLM also contains components to create Python and C++ runtimes that orchestrate the inference execution in a performant way.

NVIDIA/TensorRT-LLM

 cloud-apim/otoroshi-llm-extension — Connect, setup, secure and seamlessly manage LLM models using an Universal/OpenAI compatible API

cloud-apim/otoroshi-llm-extension

 Tom-Ryder/AgentRunKit — Lightweight Swift 6 framework for building LLM-powered agents — cloud + on-device inference via MLX on Apple Silicon

Tom-Ryder/AgentRunKit

 edwardcapriolo/deliverance — An advanced Java based inference engine for Large Language Models

edwardcapriolo/deliverance

 dcostenco/prism-mcp — 🧠 Prism MCP — Persistent memory for AI coding agents. Session history, semantic search, TurboQuant 10x compression, visual memory, autonomous Web Scholar, CRDT multi-agent merging, and a Mind Palace dashboard. Works with Claude, Cursor, Windsurf, Gemini.

dcostenco/prism-mcp

 Demorck/ClairObscurArchipelagoRandomizer — A Clair Obscur : Expedition 33 mod for the Archipelago multi-game randomizer system

Demorck/ClairObscurArchipelagoRandomizer

Section

Fresh Papers

New research worth bookmarking for a deeper read.

SlopCodeBench: Benchmarking How Coding Agents Degrade Over Long-Horizon Iterative Tasks

SlopCodeBench: Benchmarking How Coding Agents Degrade Over Long-Horizon Iterative Tasks

 AVO: Agentic Variation Operators for Autonomous Evolutionary Search

AVO: Agentic Variation Operators for Autonomous Evolutionary Search

 GameplayQA: A Benchmarking Framework for Decision-Dense POV-Synced Multi-Video Understanding of 3D Virtual Agents

GameplayQA: A Benchmarking Framework for Decision-Dense POV-Synced Multi-Video Understanding of…

 Calibri: Enhancing Diffusion Transformers via Parameter-Efficient Calibration

Calibri: Enhancing Diffusion Transformers via Parameter-Efficient Calibration

 IQuest-Coder-V1 Technical Report

IQuest-Coder-V1 Technical Report

 The Pulse of Motion: Measuring Physical Frame Rate from Visual Dynamics

The Pulse of Motion: Measuring Physical Frame Rate from Visual Dynamics

 Demographic Fairness in Multimodal LLMs: A Benchmark of Gender and Ethnicity Bias in Face Verification

Demographic Fairness in Multimodal LLMs: A Benchmark of Gender and Ethnicity Bias in Face Verif…

 Training the Knowledge Base through Evidence Distillation and Write-Back Enrichment

Training the Knowledge Base through Evidence Distillation and Write-Back Enrichment

 LanteRn: Latent Visual Structured Reasoning

LanteRn: Latent Visual Structured Reasoning

 RefAlign: Representation Alignment for Reference-to-Video Generation

RefAlign: Representation Alignment for Reference-to-Video Generation

Section

Community Chatter

Directional signals from discussion-heavy sources.

The incident that blamed AI for the Iranian school bombing is a more fundamental problem

The incident that blamed AI for the Iranian school bombing is a more fundamental problem

 The incident that blamed AI for the Iranian school bombing is a more fundamental problem

The incident that blamed AI for the Iranian school bombing is a more fundamental problem

 🚨 BREAKING: CHINA just released a Python framework for building AI agents. 100% OPEN SOURCE. It has visual agent design, MCP tools, memory, RAG, and reasoning. All built in. All working together. It's called AgentScope. You describe your agent syste

🚨 BREAKING: CHINA just released a Python framework for building AI agents.

 Agentic AI rises ⬆️ AI infrastructure shifts 🔀 Agentic AI calls for infrastructure designed for real-world inference and scale, elevating the CPU’s role at the center of AI systems to drive efficiency and unlock performance - and Arm is built for it. - twitter.com

Agentic AI rises ⬆️ AI infrastructure shifts 🔀 Agentic AI calls for infrastructure designed for…

 Announcing ARC-AGI-3 The only unsaturated agentic intelligence benchmark in the world Humans score 100%, AI

Announcing ARC-AGI-3 The only unsaturated agentic intelligence benchmark in the world Humans sc…

 At the moment, ARC-AGI-3 is the only unsaturated agentic AI benchmark. Sub-1% scores from frontier models on the private test set. If you want to be among the first to know when an AGI breakthrough happens, monitor the ARC-AGI-3 leaderboard. Any sudde

At the moment, ARC-AGI-3 is the only unsaturated agentic AI benchmark.

 Anthropic has been testing a new model called "Mythos" with certain customers: - a "step change" in AI capabilities, including "dramatically higher scores" in coding, academic reasoning and cybersecurity - "currently far ahead of any other AI model in cyber cap

Anthropic has been testing a new model called "Mythos" with certain customers: - a "step change…

 I'd just like to interject for a moment. What you're referring to as a "model with no harness", is in fact, a model with a harness, or as I've recently taken to calling it, an agentic language system. An LLM is not an agentic language system unto itself, but rather just o

I'd just like to interject for a moment. What you're referring to as a "model with no harness",…

 The last two years reshaped how we evaluate models and what we expect from them. OpenAI's o1 showed that "thinking" could be a first-class capability, something you train for and expose to users. DeepSeek-R1 proved that reasoning-style post-training coul

The last two years reshaped how we evaluate models and what we expect from them.

 Anthropic’s new model, Capybara: “Compared to Claude Opus 4.6, Capybara achieves dramatically higher scores in software coding, academic reasoning, and cybersecurity.” According to Dario's previous interview, it might be a 10T-parameter model that cos

Anthropic’s new model, Capybara: “Compared to Claude Opus 4.6, Capybara achieves dramatically h…

Archive

Recent Digest Posts

2026-03-29 Daily AI News Digest — 2026-03-29 GitHub velocity is led by framersai/agentos; paper attention is clustering around SlopCodeBench: Benchmarking How Coding Agents Degrade Over Long-Horizon Iterative Tasks; social attention is tilting toward The incident that blamed AI for the Iranian school bombing is a more fundamental problem; biggest mover: AVO: Agentic Variation Operators for Autonomous Evolutionary Search (+49). 10 repo signals, 10 paper picks, and 10 community items made today's cut. 2026-03-28 Daily AI News Digest — 2026-03-28 GitHub velocity is led by smith-horn/skillsmith; paper attention is clustering around SlopCodeBench: Benchmarking How Coding Agents Degrade Over Long-Horizon Iterative Tasks; social attention is tilting toward Show GN: Geas - Claude Code Contract-based governance harness for multi-agent long-term operati…; biggest mover: Demorck/ClairObscurArchipelagoRandomizer (+33). 10 repo signals, 10 paper picks, and 10 community items made today's cut. 2026-03-27 Daily AI News Digest — 2026-03-27 GitHub velocity is led by smith-horn/skillsmith; paper attention is clustering around Why Does Self-Distillation (Sometimes) Degrade the Reasoning Capability of LLMs?; social attention is tilting toward Show GN: Geas - Claude Code Contract-based governance harness for multi-agent long-term operati…; biggest mover: Nudging Hidden States: Training-Free Model Steering for Chain-of-Thought Reasoning in Large Aud… (+15). 10 repo signals, 10 paper picks, and 10 community items made today's cut. 2026-03-26 Daily AI News Digest — 2026-03-26 GitHub velocity is led by comet-ml/opik; paper attention is clustering around SpecEyes: Accelerating Agentic Multimodal LLMs via Speculative Perception and Planning; social attention is tilting toward OpenAI, Sora service terminated. 8 repo signals, 6 paper picks, and 3 community items made today's cut. 2026-03-25 Daily AI News Digest — 2026-03-25 GitHub velocity is led by comet-ml/opik; paper attention is clustering around SpecEyes: Accelerating Agentic Multimodal LLMs via Speculative Perception and Planning; social attention is tilting toward OpenAI, Sora service terminated. 8 repo signals, 6 paper picks, and 3 community items made today's cut. 2026-03-24 Daily AI News Digest — 2026-03-24 Repo momentum is led by ceedaragents/cyrus; paper attention is clustering around OfficeQA Pro: An Enterprise Benchmark for End-to-End Grounded Reasoning; community chatter is moving through GeekNews. 4 repo signals, 4 paper picks, and 4 community items made today's cut.
Browse the monthly archive

Generated from the ranked feed for Mar 29, 2026.

Linked Mentions

No linked mentions yet.