BestHub
Sign in
Tagged articles

long-running agents

7 articles · Page 1 of 1
DataFunTalk
DataFunTalk
Oct 25, 2026 · Artificial Intelligence

EvoMemBench: Why AI Agent Memory Must Evolve, Not Just Remember

The article introduces EvoMemBench, a benchmark evaluating AI agent memory across four dimensions — task-internal knowledge, task-internal execution, cross-task knowledge, and cross-task execution — revealing five key findings: memory isn't always better than raw context, helps more on hard tasks, is critical for long-horizon execution, retrieval does not equal revision, and no single memory type fits all tasks. It concludes with five engineering principles for building effective memory systems.

AI agent memoryAgent EvaluationEvoMemBench
0 likes · 20 min read
EvoMemBench: Why AI Agent Memory Must Evolve, Not Just Remember
DeepHub IMBA
DeepHub IMBA
Oct 13, 2026 · Artificial Intelligence

Harness-of-Harness: Converting Long-Running AI Agent Drift into Checkable Engineering Constraints

Shanghai AI Lab's Harness-of-Harness (HoH) framework organizes AI coding agents into planning, development, and QA roles with persistent state across iterations, achieving 52.25% average improvement over vanilla agents on three benchmarks and demonstrating 70-round autonomous FPS game development with 81 tracked issues.

AI coding agentsBenchmark EvaluationHarness-of-Harness
0 likes · 18 min read
Harness-of-Harness: Converting Long-Running AI Agent Drift into Checkable Engineering Constraints
PaperAgent
PaperAgent
Jun 13, 2026 · Artificial Intelligence

Best Practices for Building Long‑Running Claude Fable 5 Agents

The leaked 1585‑line Claude Fable 5 system prompt reveals a full operating system for long‑running agents, detailing tool definitions, strict copyright limits, three‑layer file semantics, persistent artifact storage, MCP flow, and Claudeception, all enforced by hard‑coded rules.

Artifact PersistenceClaude Fable 5Claudeception
0 likes · 10 min read
Best Practices for Building Long‑Running Claude Fable 5 Agents
AndroidPub
AndroidPub
May 11, 2026 · Artificial Intelligence

Is Harness Engineering Just Hype? A Deep Dive into Agent Harnesses

The article traces the evolution of the "Harness" concept from traditional test harnesses to modern AI agent engineering, explains the Planner‑Generator‑Evaluator architecture, evaluates its trade‑offs, and argues that Harness Engineering is a transitional technique rather than mere hype.

AI agentsPlanner Generator Evaluatorharness engineering
0 likes · 16 min read
Is Harness Engineering Just Hype? A Deep Dive into Agent Harnesses
Architect
Architect
May 10, 2026 · Artificial Intelligence

Long‑Running Agents: From Ralph Loop to Hand‑over‑Ready Harness

The article analyzes the challenges of long‑running AI agents, showing that persistence alone is insufficient and that reliable hand‑over requires explicit specifications, external state files, drift mitigation, sub‑agents, and a verifiable evidence chain to keep the work understandable for the next model or human.

AI agentsHarnessRalph Loop
0 likes · 25 min read
Long‑Running Agents: From Ralph Loop to Hand‑over‑Ready Harness
DeepHub IMBA
DeepHub IMBA
Apr 29, 2026 · Artificial Intelligence

From Stateless to Stateful: 5 Architecture Patterns for Long‑Running Agents

The article outlines five concrete design patterns—Checkpoint‑and‑Resume, Delegated Approval, Memory‑Layered Context, Ambient Processing, and Fleet Orchestration—that enable production‑grade, multi‑day AI agents to persist state, handle failures, and scale safely.

AI agentsHuman-in-the-Loopcheckpointing
0 likes · 12 min read
From Stateless to Stateful: 5 Architecture Patterns for Long‑Running Agents
Machine Learning Algorithms & Natural Language Processing
Machine Learning Algorithms & Natural Language Processing
Apr 11, 2026 · Artificial Intelligence

From Claude Code to Codex: Migrating Anthropic’s Harness Design

The author reproduces Anthropic’s long‑running harness architecture on a Codex + GPT stack, separates planner, generator, and evaluator roles, persists state to concrete artifacts, adds strict execution constraints, and demonstrates that the approach improves task success despite higher costs, while highlighting practical pitfalls and cost‑control strategies.

Agent HarnessAnthropicClaude Code
0 likes · 12 min read
From Claude Code to Codex: Migrating Anthropic’s Harness Design