Industry Signals

A field notebook for the moving parts.

The essays, videos, papers and benchmarks that shape how LGI thinks about knowledge systems, agentic work, AI operating models, and governed intelligence infrastructure.

35 signals

EssayFeatured

Agents

Chip Huyen · Agentic Workflows

A practical overview of agents, tool use, planning, workflows, and the design questions that show up when LLMs move from chat into action.

agentstoolsplanningworkflows

Read analysis →

VideoFeatured

Matt Pocock's Agentic Engineering Workflow

David Ondrej · Agentic Coding

A practitioner walkthrough of agentic software engineering habits and AI-assisted development structure.

agentic codingdeveloper workflowClaude Codesoftware engineering

Read analysis →

VideoFeatured

Local AI Coding is Finally Good Enough

ForrestKnight · Local AI

A signal that local and semi-local coding workflows are becoming plausible for real developer use.

local AIcoding agentsdeveloper toolsopen models

Read analysis →

VideoFeatured

The Production AI Playbook: Deploying Agents at Enterprise Scale

AI Engineer · AI Operating Models

A production-focused view of what agents need beyond a demo: evaluation, observability, data, orchestration, and governance.

production agentsobservabilityevaluationgovernance

Read analysis →

VideoFeatured

OpenRAG: An Open-Source Stack for RAG

Phil Nash · Knowledge Systems

A practical signal around open retrieval-augmented generation stacks and source-connected AI systems.

RAGopen sourceretrievalcontext engineering

Read analysis →

VideoFeatured

The Doing Got Cheap. Now What?

AI News and Strategy Daily / Nate B Jones · Strategy

A strategic signal about what happens when AI makes execution cheaper.

AI workexecutionjudgementstrategy

Read analysis →

PaperFeatured

Contextual Memory Virtualisation

arXiv · Memory and Context

Treats LLM session state as a DAG with snapshots, branches, and structurally lossless trimming.

memorycontextDAGssession state

Read analysis →

PaperFeatured

SAGE

arXiv · Memory and Context

Self-evolving graph memory with reader/writer feedback for long-horizon agents.

graph memoryfeedbackretrievallong-horizon agents

Read analysis →

PaperFeatured

HORMA

arXiv · Memory and Context

Hierarchical memory navigation for efficient agents.

hierarchical memoryretrievalagentscontext

Read analysis →

PaperFeatured

Agents-K1

arXiv · Knowledge Graphs

Converts papers into typed scientific knowledge graphs with evidence, claims, citations, and relations.

knowledge graphsscientific papersprovenancecitation roles

Read analysis →

PaperFeatured

Co-Scientist

Nature / Google DeepMind · Research Automation

Multi-agent scientific discovery system using generation, critique, ranking, and hypothesis evolution.

AI sciencemulti-agentdebatehypothesis generation

Read analysis →

PaperFeatured

Life-Harness

arXiv · Harnesses and Determinism

Improves frozen agents by adapting runtime harnesses rather than model weights.

harnessesenvironment feedbackdeterministic interfaces

Read analysis →

Video

OODA Loop + Infinite Brain = the AI System Everyone's Missing

YouTube video · Knowledge Systems

Combines sensemaking loops with persistent memory, positioning AI as part of a larger decision cycle.

OODA loopknowledge systemsmemorydecision loops

Related · Holonic Knowledge Engine

Video

NotebookLM 2.0 Is Here and It's Incredible

YouTube video · Knowledge Systems

A look at source-grounded notebook workflows over collections of user-provided material.

NotebookLMsource groundingknowledge workAI notebooks

Related · Source-Grounded Workflows

Essay

My Workflow for Understanding LLM Architectures

Sebastian Raschka · Technical Learning

A practical research-learning workflow for making sense of model architectures, papers, and technical concepts.

technical learningLLM architectureresearch workflow

Related · Intelligence Library

Video

The End of the GPU Era? 1-Bit LLMs Are Here

YouTube video · Local AI

A local-AI signal about BitNet-style low-bit models and dramatically cheaper inference.

1-bit LLMsmodel efficiencylocal AIinference

Related · Local AI

Video

Run Any AI Model Locally, Even Without a GPU

YouTube video · Local AI

A practical signal around running models locally, including hardware constraints and trade-offs.

local modelshardwareself-hosted AIinference

Related · Local-First AI

Video

Why Nvidia's RTX Spark and Minimax M3 Change Everything for Local AI

YouTube video · Local AI

A hardware/software market signal about local inference becoming more practical.

local AINvidiainference hardwareopen models

Related · AI Infrastructure

Benchmark

ParallelKernelBench: Frontier LLMs Can't Write Fast Multi-GPU Kernels Yet

Builder Signals corpus · Evaluation

A benchmark signal showing that frontier coding models struggle with high-performance multi-GPU kernel work.

benchmarksGPU kernelscoding modelsevaluation

Related · Hard-to-Check Work

Video

AutoResearch Explained

YouTube video · Research Automation

A signal around automated research workflows: finding sources, processing papers, and turning literature into usable outputs.

research automationpapersliterature reviewAI workflows

Related · Intelligence Library

Video

The AI Science Era Just Started

YouTube video · Research Automation

A broad signal about AI's expanding role in scientific discovery and technical research.

AI scienceresearch automationdiscoveryfrontier AI

Related · Research-to-Practice Translation

Video

From arXiv AI Paper: Transformers Explained

YouTube video · Technical Learning

A technical explainer grounding modern AI systems in transformers, attention, tokenization, and sequence modelling.

transformersattentionLLM fundamentalstechnical learning

Related · AI Foundations

Video

10 Papers Every Future AI Engineer Must Read

YouTube video · Technical Learning

A curated technical-learning signal around papers that shape modern AI engineering practice.

papersAI engineeringtechnical learningcurriculum

Related · Library

Video

Josh Payne: How Coframe Is Building Living Interfaces

YouTube video · Interfaces

A signal about adaptive interfaces that change based on user behaviour, context, and AI-driven experimentation.

living interfacesproduct UXagentsexperimentation

Related · AI-Native Interfaces

Video

The Next $100B Market: Selling to AI Agents

YouTube video · Interfaces

A market-structure signal about agents becoming buyers, intermediaries, or decision participants.

agent economyAI-native commercemachine-readable web

Related · Agent-Readable Systems

Video

Inside the Shift Toward Enterprise AI

YouTube video · AI Operating Models

A broad enterprise-AI signal about organisations moving from experimentation toward adoption.

enterprise AIadoptionoperationsAI strategy

Related · Knowledge Systems Audit

Paper

LLM-Oriented Information Retrieval

arXiv · Retrieval

Frames RAG and agentic search around usable evidence density, not raw recall.

information retrievalRAGdenoisingevidence

Related · Vault Indexing

Paper

Trace-Based Assurance Framework

arXiv · Governance

Contracts, tests, traces, and governance for agentic orchestration with side effects.

assurancetracescontractstesting

Related · Run Ledger

Paper

Agentic AI with Orchestrator-Agent Trust

arXiv · Governance

Uses trust calibration and RAG-based re-evaluation across agent roles.

trustorchestrationRAGmulti-agent

Related · HKE Routing

Paper

Drop the Hierarchy and Roles

arXiv · Agentic Workflows

Argues self-organising agents can outperform designed hierarchies under some conditions.

self-organisationagentsroleshierarchy

Related · Holon Role Design

Paper

Building Persona-Based Agents On Demand

arXiv · Agentic Workflows

Dynamic agent personas tailored to user/task context instead of static workflows.

personasmulti-agentworkflowsuser needs

Related · Agent Faces

Vault synthesis

HIPIF

HKE vault synthesis · Planning and Work Loops

Hierarchical planning and information folding for long-horizon LLM agent learning.

planningfoldingwork loopslong-horizon agents

Related · Work Loops Runtime

Paper

Recursive Language Models

arXiv · Planning and Work Loops

Lets models decompose and recursively inspect long prompts beyond native context windows.

recursive modelslong contexttool useplanning

Related · Query Orchestration

Paper

Cutscene Agent

arXiv · Planning and Work Loops

Multi-agent framework for end-to-end 3D cutscene generation with engine integration.

creative agents3Dtool useorchestration

Related · Applied Agentic Workflows

Paper

DMoE

arXiv · Memory Alternatives

Modular experts inject knowledge into model behaviour.

parametric memoryMoEknowledge injectionrisk

Related · Governed Memory