context engineering

Evidence (14)

• 7InfoQ·10/11/2026research
  Context Language Models Aim to Self-Manage Context

  Researchers from Meta, MIT, and the University of Washington introduced Context Language Models (CLMs), which let models manage and edit their own context instead of depending on fixed summarization, compression, or retrieval pipelines. The approach reportedly improves performance while reducing compute costs.

• 5The New Stack·8/31/2026framework_update
  Why coding agents need a context development lifecycle

  The article argues that skills, prompt instructions, agent configurations, and rules files should be treated as software artifacts with their own development lifecycle. It proposes a Context Development Lifecycle (CDLC) with four phases: generate, evaluate, distribute, and observe, to improve reliability and reduce regressions in coding agents.

• 3Simon Willison·8/23/2026research
  Drew Breunig on Fable and the end of the free lunch

  Simon Willison quotes Drew Breunig arguing that after Fable arrived, it no longer made sense to ignore coding harnesses or context strategies and simply wait for better models. The note suggests that model quality and pricing no longer automatically cover shortcomings in workflow design, so teams must think more carefully about where work is done.

• 5InfoQ·8/14/2026framework_update
  Context Engineering for Reliable Coding Agents

  Patrick Debois and Baruch Sadogursky explain why coding agents often fail when prompts are overloaded with noisy context. They outline practical context engineering patterns such as lazy-loaded skills, versioned context artifacts, externalized memory banks, and LLM-as-a-judge evaluations to build more reliable agent workflows.

• 6Hacker News·7/25/2026framework_update
  Claude 5 context engineering guidelines

  Anthropic published guidance on how to structure context for Claude 5 generation models to improve reliability and output quality. The article frames context engineering as a practical discipline for building better prompts, inputs, and agent workflows around the model.

• 5InfoQ·7/25/2026research
  Root Cause Analysis Shifts Toward Context Engineering

  Engineers are increasingly arguing that modern LLMs can perform root cause analysis effectively when supplied with well-prepared context, moving the main challenge to telemetry correlation pipelines. A Coroot experiment across eleven models provides early evidence supporting this view.

• 5Hacker News·7/23/2026research
  Why Software Factories Fail

  This Hacker News item highlights an essay arguing that software factories do not succeed by relying on engineering harnesses alone. The piece appears to focus on advanced context engineering for coding agents and the limits of trying to industrialize software creation without addressing broader workflow and system constraints.

• 5InfoQ·7/17/2026framework_update
  QCon AI Boston Highlights Production AI Infrastructure Needs

  QCon AI Boston 2026 centered on the operational demands of deploying AI agents in production. Speakers emphasized stronger context management, security controls around agents, and a more complete engineering approach that includes evaluation and platform support.

• 5Martin Fowler·7/13/2026research
  Fowler on AI Harnesses and Self-Hosted Models

  Martin Fowler reports that teams are focusing more on AI harness engineering, especially context management and stronger validation techniques such as property-based testing and formal methods. He also notes growing interest in self-hosted open-weight models, driven by cost, data security, model sovereignty, and the possibility of fine-tuning domain-specific models to reduce token usage.

• 5Hacker News·7/9/2026research
  AI changes the economics of software rewrites

  The author argues that AI makes software rewrites more attractive when the target codebase has clear, consistent patterns that models can readily learn from. In contrast, proprietary languages, legacy systems, and inconsistent frameworks increase context requirements, token usage, and output variance, making AI-assisted rewrites more costly and less effective.

• 5The New Stack·7/4/2026research
  Cheaper Models Alone Won’t Fix AI Token Costs

  The article argues that the main cost problem in agentic AI is not just model price, but excessive token consumption across agent workflows. It highlights three approaches to lower spend: compressing context, routing simpler tasks to cheaper models, and using semantic caching to avoid repeating similar reasoning.

• 6Martin Fowler·6/16/2026research
  Fowler on LLMs, DDD, and AI’s impact on programming

  Martin Fowler highlights Prag Dave Thomas’s view that programming with LLMs can be more enjoyable by reducing drudgery, speeding feedback, reviving old projects, and helping exploration. He also reflects on DDD Europe talks about managing LLM context windows and conversation modes, and shares Charity Majors’s warning that AI adoption creates both real productivity gains and real risks to code quality and organizational knowledge.

• 4Martin Fowler·4/8/2026framework_update
  Feedback Flywheel for AI-Assisted Development

  The article proposes a structured feedback practice for AI-assisted development, where teams capture useful prompts, missing context, recurring failures, and successful patterns from AI sessions. Those learnings are then fed back into shared artifacts such as priming documents, runbooks, and review checklists so the team improves over time instead of repeating the same mistakes.

• 5Martin Fowler·4/2/2026research
  Harness engineering for coding agent users

  The article proposes a harness engineering mental model for coding agents to improve trust in AI-generated code. It argues for combining feedforward guides and feedback sensors, using both deterministic tooling and semantic evaluation to reduce errors before human review.