agentic workflows

8 articles · 15 co-occurring · 1 contradictions · 2 briefs

Paper advances understanding of agentic long-context challenges by introducing dynamic evaluation where models generate context (queries, reflections) that can degrade performance.

Related concepts

multi agent orchestration 6 tool integration patterns 5 context window management 5 token efficiency 2 prompt engineering 2 model selection strategy 2 token economy 1 task decomposition 1 system prompt optimization 1 state management 1 spec driven development 1 retrieval augmented generation 1 prompt engineering iteration 1 prompt composition 1 production reliability gap 1

Contradictions

@IntuitMachine: Good counterintuitive advice.

Author argues against heavy toolcalls/thinking, positioning as overhead vs. specificity

Signal history

2026-W15

Evidence chain (8 articles, showing 8)

Haystack Engineering: Context Engineering for Heterogeneous and Agentic Long-Context Evaluation | OpenReview extends

Paper advances understanding of agentic long-context challenges by introducing dynamic evaluation where models generate context (queries, reflections) that can degrade performance.

The End of Infinite Context: Engineering Reliability in the Age of Agentic Workflows | by Ali moradi | Medium supports

Article describes specific agentic loop (Research→Plan→Reset→Implement) as the architectural pattern that solves context management—agents reset state to prevent context bloat.

AI Agent Architectures example_of

Multi-agent agentic workflows are explicitly discussed as a context orchestration pattern—separating concerns across agents rather than forcing single agent to hold all context.

@IntuitMachine: Good counterintuitive advice. contradicts

Author argues against heavy toolcalls/thinking, positioning as overhead vs. specificity

@dexhorthy: everything is context engineering extends

Task composition naturally leads to agent-like orchestration; this is a precursor architecture pattern.

How to create a new release using the MCP server with Claude example_of

Article frames the use case as 'agentic workflows' where Claude takes autonomous action (adding releases) via MCP

Context Engineering: The Future of AI Development extends

Post positions context engineering as the next evolution after agentic workflows, suggesting these are complementary rather than competing concerns.

State of AI Engineering | Datadog extends

Article finds that agent workloads (multi-step control flow, tool execution) rely heavily on context for reliability; context engineering becomes more critical in agentic systems.

query this concept

$ db.articles("agentic-workflows")

$ db.cooccurrence("agentic-workflows")

$ db.contradictions("agentic-workflows")