[{"data":1,"prerenderedAt":1603},["ShallowReactive",2],{"navigation":3,"\u002Fnews\u002Finsights\u002Fcontext-is-new-code":341,"\u002Fnews\u002Finsights\u002Fcontext-is-new-code-surround":881},[4,8,17,21,25,29,33,329,333,337],{"title":5,"path":6,"stem":7},"About","\u002Fabout","about",{"title":9,"path":10,"stem":11,"children":12},"Authentication","\u002Fauth","auth",[13],{"title":14,"path":15,"stem":16},"Email Confirmation","\u002Fauth\u002Fconfirmation","auth\u002Fconfirmation",{"title":18,"path":19,"stem":20},"Industry Briefs","\u002Fcase-studies","case-studies",{"title":22,"path":23,"stem":24},"Contact","\u002Fcontact","contact",{"title":26,"path":27,"stem":28},"Thinkata | Production AI Engineering","\u002F","index",{"title":30,"path":31,"stem":32},"Insights","\u002Finsights","insights",{"title":34,"path":35,"stem":36,"children":37},"News","\u002Fnews","news",[38,41,65],{"title":39,"path":35,"stem":40},"News & Insights","news\u002Findex",{"title":18,"path":42,"stem":43,"children":44},"\u002Fnews\u002Fcase-studies","news\u002Fcase-studies",[45,49,53,57,61],{"title":46,"path":47,"stem":48},"Building Secure and Scalable AI Infrastructure: Integrating with Existing Systems through Modern Cloud Frameworks","\u002Fnews\u002Fcase-studies\u002Fcloud-infrastructure-ai","news\u002Fcase-studies\u002Fcloud-infrastructure-ai",{"title":50,"path":51,"stem":52},"Making Sense of Financial Regulations: How AI Teams Can Tackle Complex Documents","\u002Fnews\u002Fcase-studies\u002Ffinancial-regulations","news\u002Fcase-studies\u002Ffinancial-regulations",{"title":54,"path":55,"stem":56},"AI-Powered Transformations in Healthcare","\u002Fnews\u002Fcase-studies\u002Fhealth-care","news\u002Fcase-studies\u002Fhealth-care",{"title":58,"path":59,"stem":60},"Generative AI in Upstream Natural Gas: Shell's Exploration Initiative","\u002Fnews\u002Fcase-studies\u002Foil-gas","news\u002Fcase-studies\u002Foil-gas",{"title":62,"path":63,"stem":64},"Optimizing Manufacturing with AI-Driven Multi-Agent Systems","\u002Fnews\u002Fcase-studies\u002Fsupply-chain-optimization","news\u002Fcase-studies\u002Fsupply-chain-optimization",{"title":30,"path":66,"stem":67,"children":68},"\u002Fnews\u002Finsights","news\u002Finsights",[69,73,77,81,85,89,93,97,101,105,109,113,117,121,125,129,133,137,141,145,149,153,157,161,165,169,173,177,181,185,189,193,197,201,205,209,213,217,221,225,229,233,237,241,245,249,253,257,261,265,269,273,277,281,285,289,293,297,301,305,309,313,317,321,325],{"title":70,"path":71,"stem":72},"Nothing Gets Deleted, Just Blurred","\u002Fnews\u002Finsights\u002Fadaptive-context-memory","news\u002Finsights\u002Fadaptive-context-memory",{"title":74,"path":75,"stem":76},"The Capability-Reliability Split in Agent Systems","\u002Fnews\u002Finsights\u002Fagent-capability-reliability-split","news\u002Finsights\u002Fagent-capability-reliability-split",{"title":78,"path":79,"stem":80},"The Rise of AI Agents in Cyberattacks: Latest Research and Threats","\u002Fnews\u002Finsights\u002Fai-agent-cyber-threats","news\u002Finsights\u002Fai-agent-cyber-threats",{"title":82,"path":83,"stem":84},"Winning the Bid Is Not the Same as Knowing the Price","\u002Fnews\u002Finsights\u002Fai-agent-economics","news\u002Finsights\u002Fai-agent-economics",{"title":86,"path":87,"stem":88},"The Smart Enterprise AI Stack: Why Teams of AI Agents Beat Solo Models Consistently","\u002Fnews\u002Finsights\u002Fai-architecture","news\u002Finsights\u002Fai-architecture",{"title":90,"path":91,"stem":92},"When Seeing Everything Becomes the Only Option","\u002Fnews\u002Finsights\u002Fai-comprehensive-observability","news\u002Finsights\u002Fai-comprehensive-observability",{"title":94,"path":95,"stem":96},"The Data Infrastructure AI-Native Systems Can't Ignore","\u002Fnews\u002Finsights\u002Fai-data-layer","news\u002Finsights\u002Fai-data-layer",{"title":98,"path":99,"stem":100},"Enterprise AI Triage Systems: Intelligent Automation for Large-Scale Operations","\u002Fnews\u002Finsights\u002Fai-enterprise-triage","news\u002Finsights\u002Fai-enterprise-triage",{"title":102,"path":103,"stem":104},"When Oversight Becomes Infrastructure","\u002Fnews\u002Finsights\u002Fai-governed-autonomy","news\u002Finsights\u002Fai-governed-autonomy",{"title":106,"path":107,"stem":108},"Designing for Graceful Failure in Compound AI Systems","\u002Fnews\u002Finsights\u002Fai-graceful-failure","news\u002Finsights\u002Fai-graceful-failure",{"title":110,"path":111,"stem":112},"Intelligent Composability: Building AI Systems Like Orchestra, Not Soloists","\u002Fnews\u002Finsights\u002Fai-intelligent-composability","news\u002Finsights\u002Fai-intelligent-composability",{"title":114,"path":115,"stem":116},"Building the Plane While Flying It — Migrating from Monolith to AI-Native Without Stopping","\u002Fnews\u002Finsights\u002Fai-migration-path","news\u002Finsights\u002Fai-migration-path",{"title":118,"path":119,"stem":120},"Stability Through Continuous Adaptation","\u002Fnews\u002Finsights\u002Fai-native-overview","news\u002Finsights\u002Fai-native-overview",{"title":122,"path":123,"stem":124},"Provable Stability: Mathematical Guarantees for Adaptive AI Systems","\u002Fnews\u002Finsights\u002Fai-provable-stability","news\u002Finsights\u002Fai-provable-stability",{"title":126,"path":127,"stem":128},"How Temperature Tuning Makes or Breaks Reinforcement Learning","\u002Fnews\u002Finsights\u002Fai-soft-actor-critic-entropy-collapse","news\u002Finsights\u002Fai-soft-actor-critic-entropy-collapse",{"title":130,"path":131,"stem":132},"Testing What Can't Be Predicted","\u002Fnews\u002Finsights\u002Fai-systems-testing","news\u002Finsights\u002Fai-systems-testing",{"title":134,"path":135,"stem":136},"Delegating the Job Is Not the Same as Delegating the Rules","\u002Fnews\u002Finsights\u002Fauthorization-drift-multi-agent","news\u002Finsights\u002Fauthorization-drift-multi-agent",{"title":138,"path":139,"stem":140},"Closing the Loop: How Human Corrections Can Make AI Systems Smarter Over Time","\u002Fnews\u002Finsights\u002Fclosing-the-loop","news\u002Finsights\u002Fclosing-the-loop",{"title":142,"path":143,"stem":144},"A Panel Can Be Safer Than Any Member of It","\u002Fnews\u002Finsights\u002Fcoalitional-alignment-panel-review","news\u002Finsights\u002Fcoalitional-alignment-panel-review",{"title":146,"path":147,"stem":148},"Multi-Path Reasoning: Collaborative and Competitive Approaches in AI","\u002Fnews\u002Finsights\u002Fcollaborative-competitive-agents","news\u002Finsights\u002Fcollaborative-competitive-agents",{"title":150,"path":151,"stem":152},"Why Challenges Supercharge Smarts for Humans and AI","\u002Fnews\u002Finsights\u002Fcompetition-improves-ai","news\u002Finsights\u002Fcompetition-improves-ai",{"title":154,"path":155,"stem":156},"Context is Infrastructure, Not Instructions","\u002Fnews\u002Finsights\u002Fcontext-is-infrastructure","news\u002Finsights\u002Fcontext-is-infrastructure",{"title":158,"path":159,"stem":160},"Context is the New Code","\u002Fnews\u002Finsights\u002Fcontext-is-new-code","news\u002Finsights\u002Fcontext-is-new-code",{"title":162,"path":163,"stem":164},"Continuous Thought Machines","\u002Fnews\u002Finsights\u002Fcontinuous-thought-machines","news\u002Finsights\u002Fcontinuous-thought-machines",{"title":166,"path":167,"stem":168},"Don't Vibe, Architect","\u002Fnews\u002Finsights\u002Fdont-vibe-architect","news\u002Finsights\u002Fdont-vibe-architect",{"title":170,"path":171,"stem":172},"The Edge of the Underdefined","\u002Fnews\u002Finsights\u002Fedge-of-the-underdefined","news\u002Finsights\u002Fedge-of-the-underdefined",{"title":174,"path":175,"stem":176},"Experts All the Way Down","\u002Fnews\u002Finsights\u002Fexperts-all-the-way","news\u002Finsights\u002Fexperts-all-the-way",{"title":178,"path":179,"stem":180},"A Multi-Tier Safety Architecture for Critical Applications","\u002Fnews\u002Finsights\u002Ffour-tier-architecture","news\u002Finsights\u002Ffour-tier-architecture",{"title":182,"path":183,"stem":184},"Green Dashboard, Unhappy Users","\u002Fnews\u002Finsights\u002Fgreen-dashboard-unhappy-users","news\u002Finsights\u002Fgreen-dashboard-unhappy-users",{"title":186,"path":187,"stem":188},"Hybrid Autoregressive Residual Tokens","\u002Fnews\u002Finsights\u002Fhart-model","news\u002Finsights\u002Fhart-model",{"title":190,"path":191,"stem":192},"Hierarchical Reasoning in Artificial Intelligence","\u002Fnews\u002Finsights\u002Fhierarchical-approaches","news\u002Finsights\u002Fhierarchical-approaches",{"title":194,"path":195,"stem":196},"Taking Turns Instead of Splitting the Difference","\u002Fnews\u002Finsights\u002Finterleaved-distillation-rl","news\u002Finsights\u002Finterleaved-distillation-rl",{"title":198,"path":199,"stem":200},"Latent Diffusion for Language Generation: A Comprehensive Overview","\u002Fnews\u002Finsights\u002Flatent-diffusion-for-language","news\u002Finsights\u002Flatent-diffusion-for-language",{"title":202,"path":203,"stem":204},"Breaking Language Barriers: How AI Can Translate Without Examples","\u002Fnews\u002Finsights\u002Flearning-languages","news\u002Finsights\u002Flearning-languages",{"title":206,"path":207,"stem":208},"The Emergence of AI Deception: How Large Language Models Have Learned to Strategically Mislead Users","\u002Fnews\u002Finsights\u002Fllm-deception","news\u002Finsights\u002Fllm-deception",{"title":210,"path":211,"stem":212},"Grading on a Shared Curve","\u002Fnews\u002Finsights\u002Fllm-judge-correlated-errors","news\u002Finsights\u002Fllm-judge-correlated-errors",{"title":214,"path":215,"stem":216},"Synergizing Specialized Reasoning and General Capabilities in AI","\u002Fnews\u002Finsights\u002Fllm-reasoning-advances","news\u002Finsights\u002Fllm-reasoning-advances",{"title":218,"path":219,"stem":220},"The Expensive Default","\u002Fnews\u002Finsights\u002Fllm-routing-cost-quality","news\u002Finsights\u002Fllm-routing-cost-quality",{"title":222,"path":223,"stem":224},"The AI That Rewrites Itself: MIT's Breakthrough in Self-Adapting Language Models","\u002Fnews\u002Finsights\u002Fllm-seal","news\u002Finsights\u002Fllm-seal",{"title":226,"path":227,"stem":228},"Forgetting on Request","\u002Fnews\u002Finsights\u002Fmachine-unlearning-removal-suppression","news\u002Finsights\u002Fmachine-unlearning-removal-suppression",{"title":230,"path":231,"stem":232},"The Map and the Walk","\u002Fnews\u002Finsights\u002Fmap-and-walk","news\u002Finsights\u002Fmap-and-walk",{"title":234,"path":235,"stem":236},"Metacognitive Reinforcement Learning for Self-Improving AI Systems","\u002Fnews\u002Finsights\u002Fmetacognitive-reinforcement-learning","news\u002Finsights\u002Fmetacognitive-reinforcement-learning",{"title":238,"path":239,"stem":240},"Revolutionary Advancements in Mixture of Experts (MoE) Architectures","\u002Fnews\u002Finsights\u002Fmixture-of-experts","news\u002Finsights\u002Fmixture-of-experts",{"title":242,"path":243,"stem":244},"One Model, Many Customers, and the Leak Nobody Tests For","\u002Fnews\u002Finsights\u002Fmulti-tenant-ai-isolation","news\u002Finsights\u002Fmulti-tenant-ai-isolation",{"title":246,"path":247,"stem":248},"Balancing Neural Plasticity and Stability","\u002Fnews\u002Finsights\u002Fneural-plasticity","news\u002Finsights\u002Fneural-plasticity",{"title":250,"path":251,"stem":252},"Offline RL and the Data Flywheel","\u002Fnews\u002Finsights\u002Foffline-rl-data-flywheel","news\u002Finsights\u002Foffline-rl-data-flywheel",{"title":254,"path":255,"stem":256},"Orchestration Without a Signal From the Room","\u002Fnews\u002Finsights\u002Fplanner-in-the-loop","news\u002Finsights\u002Fplanner-in-the-loop",{"title":258,"path":259,"stem":260},"Second-Guessing Has a Price","\u002Fnews\u002Finsights\u002Freasoning-budget-allocation","news\u002Finsights\u002Freasoning-budget-allocation",{"title":262,"path":263,"stem":264},"Reasoning You Can Check","\u002Fnews\u002Finsights\u002Freasoning-you-can-check","news\u002Finsights\u002Freasoning-you-can-check",{"title":266,"path":267,"stem":268},"When Optimization Optimizes Itself","\u002Fnews\u002Finsights\u002Frecursive-goodhart","news\u002Finsights\u002Frecursive-goodhart",{"title":270,"path":271,"stem":272},"Reward Design as Architecture","\u002Fnews\u002Finsights\u002Freward-design-as-architecture","news\u002Finsights\u002Freward-design-as-architecture",{"title":274,"path":275,"stem":276},"When Success Has No Author: The Temporal Credit Assignment Problem","\u002Fnews\u002Finsights\u002Frl-credit-assignment-problem","news\u002Finsights\u002Frl-credit-assignment-problem",{"title":278,"path":279,"stem":280},"Beyond Entropy Collapse: When Exploration Succeeds but Learning Fails","\u002Fnews\u002Finsights\u002Frl-optimization-gaps","news\u002Finsights\u002Frl-optimization-gaps",{"title":282,"path":283,"stem":284},"The Body Was Supposed to Be the Easy Part","\u002Fnews\u002Finsights\u002Frobot-embodiment-interface","news\u002Finsights\u002Frobot-embodiment-interface",{"title":286,"path":287,"stem":288},"The Path to Practical Confidential Computing for AI Systems","\u002Fnews\u002Finsights\u002Fsecure-ai-architectures","news\u002Finsights\u002Fsecure-ai-architectures",{"title":290,"path":291,"stem":292},"When an AI Edits Its Own Memory","\u002Fnews\u002Finsights\u002Fself-editing-agent-memory","news\u002Finsights\u002Fself-editing-agent-memory",{"title":294,"path":295,"stem":296},"Guess First, Check Later","\u002Fnews\u002Finsights\u002Fspeculative-execution-pattern","news\u002Finsights\u002Fspeculative-execution-pattern",{"title":298,"path":299,"stem":300},"Spiking Neural Networks for Energy-Efficient AI","\u002Fnews\u002Finsights\u002Fspiking-neural-networks","news\u002Finsights\u002Fspiking-neural-networks",{"title":302,"path":303,"stem":304},"When Replay Is Not an Option: Streaming Q-Learning and SARSA Get a Second Look","\u002Fnews\u002Finsights\u002Fstreaming-q-learning-revival","news\u002Finsights\u002Fstreaming-q-learning-revival",{"title":306,"path":307,"stem":308},"The Turn as the Unit of Quality","\u002Fnews\u002Finsights\u002Fstructured-iteration-quality","news\u002Finsights\u002Fstructured-iteration-quality",{"title":310,"path":311,"stem":312},"AI Speech Translation: Breaking Down Language Barriers","\u002Fnews\u002Finsights\u002Fsts-performance-advances","news\u002Finsights\u002Fsts-performance-advances",{"title":314,"path":315,"stem":316},"Test-Time Training Layers: The Next Evolution in Transformer Architecture","\u002Fnews\u002Finsights\u002Ftest-time-training-layers","news\u002Finsights\u002Ftest-time-training-layers",{"title":318,"path":319,"stem":320},"Breakthrough: Large Language Models Pass the Turing Test","\u002Fnews\u002Finsights\u002Fturing-tests","news\u002Finsights\u002Fturing-tests",{"title":322,"path":323,"stem":324},"Algorithms Used in Autonomous Fighter Jet Flight Research","\u002Fnews\u002Finsights\u002Fvenom-f16-flight-autonomy","news\u002Finsights\u002Fvenom-f16-flight-autonomy",{"title":326,"path":327,"stem":328},"Training in a World That Does Not Exist Yet","\u002Fnews\u002Finsights\u002Fworld-models-as-infrastructure","news\u002Finsights\u002Fworld-models-as-infrastructure",{"title":330,"path":331,"stem":332},"Privacy Policy","\u002Fprivacy","privacy",{"title":334,"path":335,"stem":336},"Research","\u002Fresearch","research",{"title":338,"path":339,"stem":340},"Terms of Service","\u002Fterms","terms",{"id":342,"title":158,"body":343,"date":862,"description":863,"extension":864,"image":865,"meta":867,"navigation":878,"path":159,"seo":879,"stem":160,"__hash__":880},"insights\u002Fnews\u002Finsights\u002Fcontext-is-new-code.md",{"type":344,"value":345,"toc":851},"minimark",[346,365,372,389,397,400,404,424,473,501,518,528,538,542,545,562,593,603,613,617,620,639,642,645,649,652,666,674,690,693,697,700,703,706,709,712,715],[347,348,351,352,351,358],"div",{"className":349},[350],"page-title","\n  ",[353,354,158],"h1",{"className":355,"id":357},[356],"page-title__main","context-is-the-new-code",[359,360,364],"h2",{"className":361,"id":363},[362],"page-title__sub","the-discipline-the-artifacts-and-the-first-signs-of-compression","The discipline, the artifacts, and the first signs of compression",[366,367,368],"p",{},[369,370,371],"em",{},"This is the first article in \"The Meta-Engineer,\" a three-part series examining how AI is reshaping the identity and skill set of software engineers.",[366,373,374,375,379,380,388],{},"Sometime in mid-2025, a shift began among engineers building production AI systems. The previous two years had been dominated by a single idea, that the key to getting good results from a language model was learning to talk to it well. Entire job titles sprang up around the skill. Courses, certifications, and prompt libraries proliferated. And for a while, the idea held. Careful phrasing did produce better outputs. But as AI coding tools evolved from autocomplete assistants into autonomous agents, the engineers working with them found that \"prompt engineering,\" however refined, was no longer sufficient. The tasks they faced, getting an agent to navigate a 100,000-line codebase, maintain architectural consistency across sessions, and avoid repeating past mistakes, had little to do with crafting a clever sentence. They needed something more systematic. The emerging answer is ",[376,377,378],"strong",{},"context engineering",", a discipline that treats the entire informational environment surrounding an AI agent as a designed artifact ",[381,382,383],"sup",{},[384,385,387],"a",{"href":386},"#source-1","[1]",".",[366,390,391,392,396],{},"The distinction is more than semantic. Prompt engineering focuses on the instruction itself, the text sent to a language model. Context engineering encompasses everything the model sees at inference time, from system prompts and retrieved documents to session memory, tool definitions, and the structure organizing all of it ",[381,393,394],{},[384,395,387],{"href":386},". If prompt engineering is writing a memo to a new employee, context engineering is designing the entire onboarding program, complete with reference materials, reporting lines, institutional knowledge, and decision-making protocols. The memo matters, but it cannot compensate for a badly designed information environment.",[366,398,399],{},"The need for systematic context design became especially visible as coding agents moved from autocomplete tools to autonomous systems capable of multi-step reasoning. An agent that only completes the next line of code can function adequately with a short prompt. An agent that independently creates a feature branch, writes an implementation spanning multiple files, runs tests, diagnoses failures, and iterates until the build passes needs far more than an instruction. It needs to understand the project's technology stack, its conventions for error handling and logging, its test infrastructure, which directories contain which types of code, and the architectural rationale behind structural decisions that might otherwise look arbitrary. Providing all of this reliably, economically, and in the right format at the right time is a design problem, and it is the problem that context engineering exists to solve.",[359,401,403],{"id":402},"a-discipline-takes-shape","A Discipline Takes Shape",[366,405,406,407,411,412,415,416,419,420,423],{},"A comprehensive survey covering over 1,400 research papers formalized this field, establishing a taxonomy that decomposes context engineering into three foundational components ",[381,408,409],{},[384,410,387],{"href":386},". The first, ",[376,413,414],{},"context retrieval and generation",", addresses where relevant information comes from, whether through search over documents, tool calls to external APIs, or synthesis from prior interactions. The second, ",[376,417,418],{},"context processing",", covers how that information is filtered, compressed, and structured for relevance. The third, ",[376,421,422],{},"context management",", deals with the ongoing challenge of maintaining context within a model's context window, the maximum amount of text it can consider at once, across multi-step interactions. Each stage introduces its own design decisions and failure modes, and the survey reveals that treating any single stage in isolation produces fragile systems.",[347,425,427,428,427,438],{"style":426},"display: flex; flex-wrap: wrap; gap: 20px; margin: 20px 0;","\n    ",[347,429,431,432,427],{"style":430},"flex: 1; min-width: 300px;","\n        ",[433,434],"img",{"src":435,"alt":436,"style":437},"https:\u002F\u002Fimages.unsplash.com\u002Fphoto-1544396821-4dd40b938ad3?w=800&auto=format&fit=crop","Organized file folders arranged on shelves, analogous to how context engineering structures layered information into retrievable categories for AI agents","width: 100%; aspect-ratio: 1\u002F1; object-fit: cover;",[347,439,431,441,427],{"style":440},"flex: 1; min-width: 300px; display: flex; flex-direction: column; justify-content: center;",[347,442,443,444,443,450,431],{},"\n            ",[445,446,449],"h3",{"style":447,"id":448},"margin: 0 0 1rem 0;","from-craft-to-maturity-model","From Craft to Maturity Model",[366,451,453,454,460,461,464,465,468,469,388],{"style":452},"margin: 0;","Just as a well-organized notebook helps a researcher locate the right reference at the right moment, context engineering structures the informational landscape an AI agent draws from. A separate framework proposes a four-level maturity pyramid for what it calls \"agent engineering\" ",[381,455,456],{},[384,457,459],{"href":458},"#source-2","[2]",". At the base sits prompt engineering, the craft of writing individual queries. Above it sits context engineering, the design and management of the entire informational environment. The third level, ",[376,462,463],{},"intent engineering",", encodes organizational goals and trade-off hierarchies into agent infrastructure, moving beyond operational instructions to strategic alignment. At the top, ",[376,466,467],{},"specification engineering"," creates machine-readable corpora of corporate policies enabling multi-agent systems to operate autonomously at scale. Each level subsumes the one below it as a necessary foundation ",[381,470,471],{},[384,472,459],{"href":458},[366,474,475,476,480,481,484,485,488,489,492,493,496,497,500],{},"The same framework proposes five quality criteria for evaluating engineered context ",[381,477,478],{},[384,479,459],{"href":458},". ",[376,482,483],{},"Relevance"," means the agent receives only what pertains to the current task. ",[376,486,487],{},"Sufficiency"," means nothing critical is left out. ",[376,490,491],{},"Isolation",", especially important in multi-agent architectures where multiple AI sub-agents collaborate on different parts of a task, ensures each sub-agent's context does not leak into another's. ",[376,494,495],{},"Economy"," demands minimum token expenditure for maximum informational value. ",[376,498,499],{},"Provenance"," requires that every element of context be traceable to a verified source. Most teams operating at the prompt engineering level address one or two of these criteria at best, and typically only by instinct rather than by design.",[502,503,505,506,509,511,512],"blockquote",{"style":504},"color: #0066CC; font-size: 1em; border-left: 4px solid #0066CC; padding-left: 1em;","\n  \"Whoever controls the agent's context controls its behavior; whoever controls its intent controls its strategy; whoever controls its specifications controls its scale.\"",[507,508],"br",{},[507,510],{},"\n  — ",[384,513,517],{"href":514,"style":515,"target":516},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2603.09619","color: #0066CC; text-decoration: none;","_blank","Vishnyakova, 2026",[366,519,520,521,527],{},"The gap between this vision and current practice is wide. An exploratory survey of 74 software professionals across six countries found that prompt usage in software engineering remains \"largely ad hoc,\" with prompts refined through trial-and-error, rarely reused, and shaped more by individual heuristics than standardized practices ",[381,522,523],{},[384,524,526],{"href":525},"#source-3","[3]",". Most organizations are still at level one of the maturity pyramid. The knowledge to do better exists, but the institutional habits have not caught up.",[366,529,530,531,537],{},"A related line of work pushes further by arguing that prompts should be treated not as informal text but as first-class software artifacts, subject to the same lifecycle of requirements engineering, design, testing, and versioning as traditional code ",[381,532,533],{},[384,534,536],{"href":535},"#source-4","[4]",". That paper describes the present state as a \"promptware crisis,\" an echo of the original \"software crisis\" of the 1960s that gave rise to software engineering as a discipline. The parallel is illuminating. Early software development was also trial-and-error, driven by individual skill rather than systematic method. It took decades of accumulated failures, ballooning complexity, and hard-won professional norms to establish the field. Context engineering may be at a similar inflection point, the moment before a craft becomes a discipline.",[359,539,541],{"id":540},"the-artifacts-practitioners-actually-build","The Artifacts Practitioners Actually Build",[366,543,544],{},"While the academic literature establishes frameworks and taxonomies, a parallel development is happening in practice. Developers working with agentic coding tools like Claude Code, Codex, and Cursor have begun creating a new category of software artifact, configuration files that serve as persistent, structured instructions for AI agents. Files named CLAUDE.md, AGENTS.md, and .cursorrules are essentially \"READMEs for AI,\" machine-readable documents that encode project-specific knowledge an agent needs to operate effectively within a particular codebase.",[366,546,547,548,554,555,561],{},"Several empirical studies have examined what developers actually put in these files. An analysis of 328 CLAUDE.md files from popular GitHub projects found that 72.6% specify application architecture, making it the most common concern, followed by testing instructions, development guidelines, and project overviews ",[381,549,550],{},[384,551,553],{"href":552},"#source-5","[5]",". A separate study of 253 Claude Code manifests confirmed consistent structural patterns, typically one main heading with several subsections, dominated by operational commands, technical implementation notes, and high-level architectural descriptions ",[381,556,557],{},[384,558,560],{"href":559},"#source-6","[6]",". The shallow structure is not a sign of immaturity. It appears to reflect what agents actually need, a flat, scannable set of instructions rather than deeply nested documentation.",[347,563,427,564,427,587],{"style":426},[347,565,431,566,427],{"style":440},[347,567,443,568,443,572,431],{},[445,569,571],{"style":447,"id":570},"scaling-across-tools","Scaling Across Tools",[366,573,574,575,581,582,586],{"style":452},"Just as a growing organization eventually needs written policies that work across departments rather than relying on informal tribal knowledge, the expanding ecosystem of AI coding tools needs configuration standards that work across platforms. The broadest study to date examined 2,923 GitHub repositories and identified eight distinct configuration mechanisms spanning a spectrum from static context files to executable integrations ",[381,576,577],{},[384,578,580],{"href":579},"#source-7","[7]",". Context Files, simple Markdown documents like CLAUDE.md and AGENTS.md, dominate the landscape. More advanced mechanisms such as Skills (structured packages with executable resources) and Subagents remain only shallowly adopted, with most repositories defining just one or two configuration artifacts. AGENTS.md has emerged as a de facto interoperable standard, recognized across multiple tools ",[381,583,584],{},[384,585,580],{"href":579},". The picture is of an ecosystem in its early days, where the simplest approach, a well-written Markdown file, is doing the heavy lifting.",[347,588,431,589,427],{"style":430},[433,590],{"src":591,"alt":592,"style":437},"https:\u002F\u002Fimages.unsplash.com\u002Fphoto-1635859890085-ec8cb5466806?w=800&auto=format&fit=crop","A professional reviewing and signing layered documents, analogous to the layered configuration files that developers now maintain as formal agreements between human intent and AI agent behavior",[366,594,595,596,602],{},"These files are not just documentation. A controlled study of 10 repositories and 124 pull requests found that the presence of an AGENTS.md file was associated with a 29% reduction in median agent runtime and a 17% reduction in output token consumption, while maintaining comparable task completion behavior ",[381,597,598],{},[384,599,601],{"href":600},"#source-8","[8]",". The researchers hypothesize that agents spend less time on exploratory navigation when they have explicit project context, needing fewer planning iterations and fewer repeated calls to the model. In practical terms, a well-crafted context file can cut both the time and cost of an agent session by roughly a quarter.",[366,604,605,606,612],{},"Yet adoption remains strikingly low. A study of open-source software projects found that only about 5% of surveyed repositories have adopted any context file format ",[381,607,608],{},[384,609,611],{"href":610},"#source-9","[9]",". This is a field where the early adopters are seeing real gains, but the vast majority of projects have not yet begun to invest in structured agent context. The parallel to early version control adoption, or early unit testing adoption, is hard to miss. A practice that starts as optional among a skilled minority tends to become standard once enough teams experience the cost of not doing it.",[359,614,616],{"id":615},"what-goes-in-and-why-it-matters","What Goes In, and Why It Matters",[366,618,619],{},"The content of these files reveals something important about what developers have learned through experience with agents. Architecture specifications dominate because agents without architectural context tend to generate code that works in isolation but violates the system's structural assumptions. A microservices project with strict domain boundaries, for example, will see an unconstrained agent casually import across those boundaries, creating coupling that takes hours to untangle. An agent working without knowledge of a project's event-driven architecture might implement a synchronous function call where an asynchronous message was expected, producing code that compiles but behaves incorrectly under load. The agent has no way to infer architectural intent from the code alone. Architectural decisions are often conventions enforced by humans rather than patterns enforced by compilers.",[366,621,622,623,629,630,634,635,388],{},"Testing instructions appear frequently, and a recent empirical study reveals exactly why. An analysis of over 1.2 million commits across 2,168 repositories found that coding agents are significantly more likely to add mock objects to tests than human developers ",[381,624,625],{},[384,626,628],{"href":627},"#source-10","[10]",". Specifically, 36% of agent commits that modify test files introduce mocks, compared with 26% for human-authored commits. The study also found that 23% of commits made by coding agents add or change test files, compared with only 13% by non-agents, and that 68% of repositories with agent test activity also contain agent mock activity ",[381,631,632],{},[384,633,628],{"href":627},". Repositories created more recently showed even higher proportions of agent-generated test and mock commits, suggesting the trend is accelerating as agent adoption grows. Mock objects, which substitute simplified stand-ins for real system components during testing, are easier for agents to generate automatically but less effective at validating how components actually interact. Tests that mock everything pass reliably but verify very little about the real system's behavior. The researchers explicitly recommend including guidance on mocking practices in agent configuration files ",[381,636,637],{},[384,638,628],{"href":627},[366,640,641],{},"Developers have independently arrived at the same conclusion. Anti-mock instructions appear in CLAUDE.md files across many projects, a concrete example of the feedback loop between agent output and human judgment. The chain of reasoning behind such an instruction is worth unpacking. Someone had to encounter the problematic tests, recognize the pattern of excessive mocking, diagnose that the agent was reaching for mocks as the path of least resistance, and then encode a corrective instruction that prevents recurrence. That entire chain, from recognizing a quality problem to articulating a rule that addresses its root cause, is precisely the kind of reasoning that context engineering formalizes.",[366,643,644],{},"Project overviews also appear frequently, and their function is subtler than it first appears. An agent that knows it is working on a distributed event-processing system written in Rust makes different choices than one operating under the assumption that it is working on a standard web application. The overview is not there for the agent's curiosity. It establishes the interpretive frame within which every subsequent instruction and code change should be understood. Without that frame, the agent optimizes locally, generating code that satisfies the immediate request. With it, the agent's local decisions become more likely to cohere with the system's global design intent. Software projects accumulate unstated assumptions over time, assumptions about performance targets, deployment environments, backward compatibility requirements, and acceptable trade-offs between code clarity and runtime efficiency. A human developer absorbs these assumptions gradually through code review, team conversations, and debugging sessions. An agent has none of that ambient context. The project overview and its associated configuration files are the only mechanism for transmitting what would otherwise require months of socialization.",[359,646,648],{"id":647},"the-first-signs-of-compression","The First Signs of Compression",[366,650,651],{},"The configuration files described above are brand new, barely a year old as a widespread practice. They represent a distinctly human contribution, the product of engineering judgment, project-specific knowledge, and hard-won experience. And yet, there are already early signs that the same systems these files were designed to guide are learning to generate and refine similar artifacts autonomously.",[366,653,654,655,661,662,388],{},"The ACE (Agentic Context Engineering) framework treats context not as a static human-authored artifact but as an \"evolving playbook\" ",[381,656,657],{},[384,658,660],{"href":659},"#source-11","[11]",". Through a modular cycle of generation, reflection, and curation, ACE accumulates, refines, and organizes strategies without any labeled training data, relying instead on natural execution feedback. In practice, the generation phase creates new strategy elements from recent task experiences. The reflection phase evaluates which strategies contributed to successes or failures. And the curation phase integrates promising strategies into the evolving playbook while pruning elements that have proven unhelpful. What distinguishes ACE from simple prompt optimization is the cumulative, structured nature of the updates. Rather than rewriting the entire context on each iteration, the framework makes targeted additions and modifications, preserving the accumulated knowledge that prior iterations have validated ",[381,663,664],{},[384,665,660],{"href":659},[366,667,668,669,673],{},"ACE demonstrated a 10.6% improvement over strong baselines on agent benchmarks and 8.6% on domain-specific financial reasoning tasks ",[381,670,671],{},[384,672,660],{"href":659},". On the AppWorld leaderboard, ACE matched the top-ranked production-level agent on the overall average and surpassed it on the harder test-challenge split, despite using a smaller open-source model.",[366,675,676,677,680,681,684,685,689],{},"The ACE researchers identified two failure modes that plague simpler, static approaches. ",[376,678,679],{},"Brevity bias"," is the tendency for iterative optimization to collapse rich context into short, generic summaries that strip away domain-specific heuristics. ",[376,682,683],{},"Context collapse"," occurs when iterative rewriting gradually erodes important details over time ",[381,686,687],{},[384,688,660],{"href":659},". ACE addresses both with structured, incremental updates guided by a \"grow-and-refine\" principle that preserves detailed knowledge rather than compressing it. The framework argues, counterintuitively, that large language models are actually more effective with long, detailed contexts than with tight summaries. Unlike humans, LLMs can autonomously distill relevance from comprehensive inputs, so stripping context down may sacrifice the edge-case knowledge that separates correct output from output that merely compiles.",[366,691,692],{},"This is proto-self-context-engineering. The artifacts that feel novel and distinctly human today, the carefully authored CLAUDE.md files and AGENTS.md specifications that encode project architecture and testing conventions, are already beginning to be optimized by the very systems they were written to guide.",[359,694,696],{"id":695},"the-automation-ladder","The Automation Ladder",[366,698,699],{},"There is a pattern worth noticing, and it recurs so reliably across the history of software that it probably qualifies as structural rather than coincidental. Every major abstraction layer eventually got formalized, stabilized, and then partially or fully automated.",[366,701,702],{},"In the 1950s, programmers encoded instructions in raw machine language, addressing memory registers by number. Compilers eliminated that work. In the decades that followed, programmers managed memory by hand, tracking every allocation and deallocation. Garbage collectors eliminated that work. By the 1990s, developers wrote boilerplate business logic from scratch for every project, implementing authentication, database access, and request routing by hand. Frameworks and libraries eliminated most of that work. Entire product categories, e-commerce, content management, analytics, became platforms. And in the last three years, code generation itself has undergone a dramatic shift. What began as autocomplete suggestions in IDEs evolved into autonomous agents capable of creating features, writing tests, and issuing pull requests with minimal human direction.",[366,704,705],{},"Context engineering sits at the latest step on this ladder. It feels like the domain of uniquely human judgment, and for now, in most practical settings, it is. Designing the right information environment for an AI agent requires understanding the project, its architecture, its failure modes, and its quality standards in ways that demand genuine expertise. The decision to include anti-mock instructions in a CLAUDE.md file, for instance, reflects not just a knowledge of testing patterns but a judgment about what \"good\" means for that particular codebase. That judgment currently lives in human heads.",[366,707,708],{},"But the ACE framework demonstrates that at least the refinement of context, the iterative improvement of playbooks based on execution feedback, can be automated today. The generate-reflect-curate loop does not need labeled data. It does not need a human reviewing each iteration. It learns from the natural consequences of its own decisions, and it demonstrably outperforms static, human-authored baselines on agent benchmarks.",[366,710,711],{},"A question the remaining articles in this series will explore, is where will the ladder lead? If agents can learn to refine their own context, and the orchestration patterns that coordinate multi-agent work are themselves being learned by self-improving systems, what remains durably human? Professional developers are already shifting from writing code to designing context. If context design itself begins to compress, as the evidence tentatively suggests, the next shift may not be upward to a higher rung on the same ladder. It may be toward a different kind of work entirely.",[366,713,714],{},"As the evidence from practitioner studies, scaled infrastructure projects, and self-improving agent systems will suggest across this series, has less to do with any particular abstraction layer and more to do with the nature of the work itself. Humans persist wherever meaning is still underdefined. That edge moves, and it moves fast. But it does not disappear, because the world keeps generating new ambiguity faster than systems can resolve it.",[347,716,351,720,351,723],{"className":717},[718,719],"references","mt-8",[359,721,722],{"id":718},"References",[724,725,427,731,427,747,427,757,427,769,427,779,427,789,427,800,427,810,427,820,427,830,427,840,351],"ol",{"className":726},[727,728,729,730],"list-decimal","list-inside","space-y-2","mt-4",[732,733,735,736,739,740],"li",{"id":734},"source-1","L. Mei et al., \"A Survey of Context Engineering for Large Language Models,\" ",[369,737,738],{},"arXiv",", 2025, ",[384,741,746],{"href":742,"target":516,"className":743},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2507.13334",[744,745],"text-blue-600","underline","[Online]",[732,748,750,751,753,754],{"id":749},"source-2","V. V. Vishnyakova, \"Context Engineering: From Prompts to Corporate Multi-Agent Architecture,\" ",[369,752,738],{},", 2026, ",[384,755,746],{"href":514,"target":516,"className":756},[744,745],[732,758,760,761,764,765],{"id":759},"source-3","H. Villamizar et al., \"Prompts as Software Engineering Artifacts: A Research Agenda and Preliminary Findings,\" in ",[369,762,763],{},"Proc. PROFES 2025, Lecture Notes in Computer Science",", vol. 16361, Springer, 2025, ",[384,766,746],{"href":767,"target":516,"className":768},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.17548",[744,745],[732,770,772,773,739,775],{"id":771},"source-4","Z. Chen et al., \"Promptware Engineering: Software Engineering for Prompt-Enabled Systems,\" ",[369,774,738],{},[384,776,746],{"href":777,"target":516,"className":778},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2503.02400",[744,745],[732,780,782,783,739,785],{"id":781},"source-5","H. V. F. Santos et al., \"Decoding the Configuration of AI Coding Agents: Insights from Claude Code Projects,\" ",[369,784,738],{},[384,786,746],{"href":787,"target":516,"className":788},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.09268",[744,745],[732,790,792,793,795,796],{"id":791},"source-6","W. Chatlatanagulchai et al., \"On the Use of Agentic Coding Manifests: An Empirical Study of Claude Code,\" in ",[369,794,763],{},", Springer, 2025, ",[384,797,746],{"href":798,"target":516,"className":799},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.14744",[744,745],[732,801,803,804,753,806],{"id":802},"source-7","M. Galster et al., \"Configuring Agentic AI Coding Tools: An Exploratory Study,\" ",[369,805,738],{},[384,807,746],{"href":808,"target":516,"className":809},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2602.14690",[744,745],[732,811,813,814,753,816],{"id":812},"source-8","J. L. Lulla et al., \"On the Impact of AGENTS.md Files on the Efficiency of AI Coding Agents,\" ",[369,815,738],{},[384,817,746],{"href":818,"target":516,"className":819},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2601.20404",[744,745],[732,821,823,824,739,826],{"id":822},"source-9","S. Mohsenimofidi et al., \"Context Engineering for AI Agents in Open-Source Software,\" ",[369,825,738],{},[384,827,746],{"href":828,"target":516,"className":829},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.21413",[744,745],[732,831,833,834,753,836],{"id":832},"source-10","A. Hora and R. Robbes, \"Are Coding Agents Generating Over-Mocked Tests? An Empirical Study,\" ",[369,835,738],{},[384,837,746],{"href":838,"target":516,"className":839},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2602.00409",[744,745],[732,841,843,844,753,847],{"id":842},"source-11","Q. Zhang et al., \"Agentic Context Engineering: Evolving Contexts for Self-Improving Language Models,\" in ",[369,845,846],{},"Proc. International Conference on Learning Representations (ICLR)",[384,848,746],{"href":849,"target":516,"className":850},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.04618",[744,745],{"title":852,"searchDepth":853,"depth":853,"links":854},"",2,[855,856,857,858,859,860,861],{"id":363,"depth":853,"text":364},{"id":402,"depth":853,"text":403},{"id":540,"depth":853,"text":541},{"id":615,"depth":853,"text":616},{"id":647,"depth":853,"text":648},{"id":695,"depth":853,"text":696},{"id":718,"depth":853,"text":722},"2026-04-05","The rise of context engineering as a formal discipline, the configuration files practitioners actually build, and the first signs that these artifacts are already being automated.","md",{"src":866},"https:\u002F\u002Fimages.unsplash.com\u002Fphoto-1607705703571-c5a8695f18f6?w=800&auto=format&fit=crop",{"authors":868,"badge":874,"source":876},[869],{"avatar":870,"name":872,"to":873},{"src":871},"\u002Fimg\u002Fmark_avatar.png","Mark Williams","https:\u002F\u002Fthinkata.com",{"label":875},"AI Engineering",{"name":877,"url":873},"Thinkata Research",true,{"title":158,"description":863},"-zAqujYny5wnGiBY18JTNib3V8QsFYn9upMGNtqMpY4",[882,1185],{"id":883,"title":166,"body":884,"date":1173,"description":1174,"extension":864,"image":1175,"meta":1177,"navigation":878,"path":167,"seo":1183,"stem":168,"__hash__":1184,"_path":167},"insights\u002Fnews\u002Finsights\u002Fdont-vibe-architect.md",{"type":344,"value":885,"toc":1166},[886,898,906,913,916,920,928,939,947,960,963,967,973,997,1000,1008,1016,1020,1023,1031,1055,1063,1066,1069],[347,887,351,889,351,893],{"className":888},[350],[353,890,166],{"className":891,"id":892},[356],"dont-vibe-architect",[359,894,897],{"className":895,"id":896},[362],"how-professionals-work-with-agents-how-context-scales-and-why-orchestration-is-a-transitional-skill","How professionals work with agents, how context scales, and why orchestration is a transitional skill",[366,899,900],{},[369,901,902,903],{},"This is the second article in \"The Meta-Engineer,\" a three-part series examining how AI is reshaping the identity and skill set of software engineers. The first article is ",[384,904,905],{"href":159},"\"Context is the New Code.\"",[366,907,908,909,912],{},"The ",[384,910,911],{"href":159},"first article"," in this series described a new category of software artifact, configuration files that tell AI coding agents how to behave within a particular codebase. Those files have measurable impact on agent efficiency and output quality. But they immediately raise a deeper question. If structured context is the foundation of effective agent use, who creates it, and what does the rest of the work actually look like?",[366,914,915],{},"The popular narrative about coding agents splits into two contradictory claims. One holds that agents are replacing developers, writing code at a pace no human can match. The other insists they are merely fancier autocomplete, useful for boilerplate but incapable of real engineering. A growing body of field research, large-scale repository analysis, and detailed practitioner case studies supports neither version. Professional developers are using agents extensively, but in a mode that looks nothing like \"vibe coding,\" the practice of trusting AI output without careful review. They plan, supervise, validate, and increasingly build elaborate infrastructure to keep agents effective across complex, long-running projects. The work has not disappeared. It has changed shape.",[359,917,919],{"id":918},"what-professionals-actually-do","What Professionals Actually Do",[366,921,922,923,927],{},"A field study combining 13 in-depth observations with a qualitative survey of 99 experienced developers found a consistent pattern ",[381,924,925],{},[384,926,387],{"href":386},". Professional developers value agents as a productivity boost, but they retain authority over software design and implementation. They plan before implementing, validate all agent outputs, and insist on fundamental quality attributes like maintainability, test coverage, and architectural coherence. Developers found agents well-suited to straightforward, well-described tasks but not to complex ones involving architectural judgment or unfamiliar domains. The relationship resembles less a pair programming partnership and more a delegation arrangement where the human sets the specification and reviews the results.",[502,929,930,931,933,511,935],{"style":504},"\n  \"The role is more… if you think of it like a conductor of sorts as opposed to the actual instrument player.\"",[507,932],{},[507,934],{},[384,936,938],{"href":937,"style":515,"target":516},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2512.23982","Practitioner quoted in Chang et al., 2025",[366,940,941,942,946],{},"A separate qualitative analysis of 57 practitioner videos published between late 2024 and October 2025 confirmed a complementary picture ",[381,943,944],{},[384,945,459],{"href":458},". Developers consistently describe their evolving role using the metaphor of a conductor, someone who directs rather than plays. The cognitive load has not decreased so much as shifted. Instead of grappling with syntax, APIs, and repetitive implementation details, developers devote greater attention to domain modeling, architectural decisions, and system integration. Natural language has become the primary medium of software composition, but the reasoning behind that language, the judgment about what to build and why, remains firmly human. The study also raised a specific warning about junior engineers who accept AI output without understanding it, creating what practitioners described as \"house of cards\" solutions that compile and pass tests but rest on foundations no one in the room actually understands.",[366,948,949,950,954,955,959],{},"The scale of adoption is already substantial and growing fast. A study of over 129,000 GitHub projects found that between 15.8% and 22.6% show traces of coding agent use, a remarkably high figure for tools that have existed in their current form for less than a year ",[381,951,952],{},[384,953,526],{"href":525},". Agent-assisted commits tend to be larger than purely human commits and focus disproportionately on features and bug fixes, suggesting developers use agents for substantive production work rather than experimentation. A complementary dataset of over 456,000 agent-generated pull requests (proposed code changes submitted to a repository for review) across 61,000 repositories reinforced the trend ",[381,956,957],{},[384,958,536],{"href":535},". OpenAI Codex alone produced more than 400,000 pull requests within two months of its release. Developers appear to work in two distinct modes, using agents for \"acceleration\" on familiar tasks where the goal is speed, and for \"exploration\" of unfamiliar design spaces where the goal is learning. The relevant productivity question, one that frameworks like SPACE address by measuring satisfaction, collaboration, and efficiency alongside raw throughput, is not how fast agents generate code but how effectively the combined human-agent system produces correct, maintainable software.",[366,961,962],{},"What these studies collectively describe is neither replacement nor mere assistance. The developer's contribution has shifted from producing code to producing specifications, constraints, and quality judgments, a transition that turns out to demand more expertise rather than less.",[359,964,966],{"id":965},"when-a-config-file-isnt-enough","When a Config File Isn't Enough",[366,968,969,970,972],{},"The configuration files described in the ",[384,971,911],{"href":159},", CLAUDE.md and AGENTS.md, work well for modest-sized projects. A few hundred lines of instructions can orient an agent to a codebase's conventions, testing expectations, and architectural patterns. But what happens when a project reaches 108,000 lines of code, spans 45 subsystems, and defines 35 network message types? A single file no longer suffices.",[347,974,427,975,427,981],{"style":426},[347,976,431,977,427],{"style":430},[433,978],{"src":979,"alt":980,"style":437},"https:\u002F\u002Fimages.unsplash.com\u002Fphoto-1481627834876-b7833e8f5570?w=800&auto=format&fit=crop","Library shelves organized in layered tiers of books and reference materials, analogous to how complex software projects require layered knowledge architectures for AI agents rather than a single instruction file",[347,982,431,983,427],{"style":440},[347,984,443,985,443,989,431],{},[445,986,988],{"style":447,"id":987},"three-tiers-of-machine-memory","Three Tiers of Machine Memory",[366,990,991,992,996],{"style":452},"Just as a large library organizes its holdings into different levels of accessibility, with reference materials on open shelves, specialized texts in reserve, and archival documents retrieved on request, a sufficiently complex software project needs layered knowledge infrastructure for its AI agents. A detailed case study documented exactly what this looks like ",[381,993,994],{},[384,995,553],{"href":552},". A researcher built a 108,000-line C# distributed system using Claude Code as the sole code-generation tool, developing a three-tier context architecture across 283 development sessions. The first tier, a \"hot memory\" constitution of roughly 660 lines, loaded into every agent session automatically. It encoded naming conventions, build commands, and orchestration protocols. The second tier comprised 19 specialized domain-expert agents, each responsible for a specific subsystem like networking, physics, or UI, totaling around 9,300 lines. The third tier was a cold-memory knowledge base of 34 on-demand specification documents served through a retrieval tool only when relevant. The total context infrastructure amounted to about 26,000 lines, roughly 24% of the codebase it supported.",[366,998,999],{},"The detail that the researcher's primary background is in chemistry, not software engineering, inverts a common assumption about who can do this kind of work. Building complex software with agents may depend less on traditional coding skill and more on the ability to design knowledge architectures, to decompose a problem domain into structured components and write clear specifications for each. That is an architectural competency, but not necessarily a programming one. The context infrastructure itself was AI-generated under human architectural direction, with the human's role being to decide what knowledge to capture and how to organize it.",[366,1001,1002,1003,1007],{},"Similar infrastructure patterns appear in other systems. A technical report on the OpenDev terminal agent described five-stage progressive context compaction that activates at increasing token pressure thresholds, from 70% to 99% of the model's context window capacity (the maximum amount of text it can consider at once) ",[381,1004,1005],{},[384,1006,560],{"href":559},". To counteract \"instruction fade-out,\" the phenomenon where agents gradually stop following their original instructions as a conversation grows longer, the system injects event-driven reminders at key decision points rather than relying solely on the initial prompt. A three-tier Skills hierarchy, spanning built-in, project-level, and user-defined instructions, manages reusable templates through lazy loading, injecting only what each specific task requires. These are infrastructure-level solutions to a problem that anyone running a long agent session has encountered.",[366,1009,1010,1011,1015],{},"Multi-agent approaches add another dimension of complexity and capability. A study of context engineering for coordinated coding systems found that retrieving both external knowledge (research papers and documentation) and internal codebase context (project files and conventions) substantially improved task resolution on the SWE-Bench Lite benchmark, a widely used test suite for evaluating whether agents can resolve real GitHub issues ",[381,1012,1013],{},[384,1014,580],{"href":579},". The multi-agent approach yielded higher single-shot success rates than single-agent baselines, at the cost of roughly 3 to 5 times more tokens per task. Dividing work among specialized sub-agents, each operating within a focused context window, reduced hallucinations (plausible but incorrect AI-generated content) and improved adherence to project conventions. But orchestrating multiple agents introduced its own complexity. Someone had to design the task decomposition, define agent roles, and ensure shared state remained consistent. For now, that someone is a human.",[359,1017,1019],{"id":1018},"the-orchestration-paradox","The Orchestration Paradox",[366,1021,1022],{},"The orchestration patterns that professionals develop, decomposing tasks, routing work to specialized agents, maintaining shared memory across sessions, represent genuine engineering skill. They also represent the next thing likely to be automated.",[366,1024,1025,1026,1030],{},"The Darwin Gödel Machine demonstrated this directly ",[381,1027,1028],{},[384,1029,601],{"href":600},". Rather than relying on a fixed, human-designed coordinator to direct improvements, the system iteratively modified its own codebase, including its own orchestration logic, and empirically validated each change against coding benchmarks. On SWE-bench, it improved performance from 20% to 50%. On the Polyglot benchmark, which tests across six programming languages, it improved from 14.2% to 30.7%. The key architectural insight is that this is a single system that both solves coding problems and refines its own implementation, removing the need for a separate, hand-crafted meta-agent. The better tools and workflows it discovered were not anticipated by its designers.",[347,1032,427,1033,427,1049],{"style":426},[347,1034,431,1035,427],{"style":440},[347,1036,443,1037,443,1041,431],{},[445,1038,1040],{"style":447,"id":1039},"when-the-wrapper-becomes-redundant","When the Wrapper Becomes Redundant",[366,1042,1043,1044,1048],{"style":452},"Just as a machine tool capable of manufacturing other machine tools represents a fundamentally different category than one that merely stamps out parts, a coding agent that can edit its own source code occupies a different position than one that simply follows instructions. The SICA system (Self-Improving Coding Agent) demonstrated this by autonomously modifying its own Python codebase, improving from 17% to 53% on a subset of SWE-Bench Verified ",[381,1045,1046],{},[384,1047,611],{"href":610},". One finding proved particularly telling. When a reasoning model was provided as a sub-component, crude reasoning scaffolds that SICA had built for itself actually hurt performance, because the model's native reasoning was better than the agent's self-designed wrapper. This is a concrete instance of a recurring compression pattern, where a layer that was necessary at one capability level becomes counterproductive when the underlying system matures.",[347,1050,431,1051,427],{"style":430},[433,1052],{"src":1053,"alt":1054,"style":437},"https:\u002F\u002Fimages.unsplash.com\u002Fphoto-1518770660439-4636190af475?w=800&auto=format&fit=crop","Close-up of a circuit board with intricate interconnected pathways, analogous to self-modifying agent systems that rewire their own logic to improve performance",[366,1056,1057,1058,1062],{},"Meanwhile, trajectory-informed memory generation already automates the extraction of structured lessons from agent execution histories ",[381,1059,1060],{},[384,1061,628],{"href":627},". Rather than relying on humans to document what worked and what failed after each session, the system analyzes completed task trajectories, identifies which decisions led to successes or failures through causal attribution, and generates categorized guidance for future runs, including strategy tips from successful patterns, recovery tips from failure handling, and optimization tips from inefficient successes. On the AppWorld benchmark, this approach improved task completion by up to 14.3 percentage points, with the strongest gains on the most complex tasks. This is essentially automating the \"lessons learned\" process that the codified context researcher performed manually across 283 development sessions.",[366,1064,1065],{},"The pattern is consistent across these studies. The conductor role that practitioners are developing right now is structurally similar to what self-improving systems are learning to do autonomously. Decomposing tasks, routing to specialists, and refining strategies based on execution feedback are exactly the capabilities that agent systems are acquiring through their own operation. Code completion automated the first layer of developer effort. Context configuration is being formalized now. Orchestration appears to be next.",[366,1067,1068],{},"The final article in this series will take up the question this observation raises. If the orchestration layer compresses too, what remains durably human? The answer, the evidence across these studies suggests, has less to do with any particular abstraction level and more to do with wherever meaning is still underdefined.",[347,1070,351,1072,351,1074],{"className":1071},[718,719],[359,1073,722],{"id":718},[724,1075,427,1077,427,1086,427,1094,427,1103,427,1112,427,1121,427,1130,427,1139,427,1148,427,1157,351],{"className":1076},[727,728,729,730],[732,1078,1079,1080,739,1082],{"id":734},"R. Huang et al., \"Professional Software Developers Don't Vibe, They Control: AI Agent Use for Coding in 2025,\" ",[369,1081,738],{},[384,1083,746],{"href":1084,"target":516,"className":1085},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2512.14012",[744,745],[732,1087,1088,1089,739,1091],{"id":749},"H.-F. Chang et al., \"Coding With AI: From a Reflection on Industrial Practices to Future Computer Science and Software Engineering Education,\" ",[369,1090,738],{},[384,1092,746],{"href":937,"target":516,"className":1093},[744,745],[732,1095,1096,1097,753,1099],{"id":759},"R. Robbes et al., \"Agentic Much? Adoption of Coding Agents on GitHub,\" ",[369,1098,738],{},[384,1100,746],{"href":1101,"target":516,"className":1102},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2601.18341",[744,745],[732,1104,1105,1106,739,1108],{"id":771},"H. Li et al., \"The Rise of AI Teammates in Software Engineering (SE) 3.0: How Autonomous Coding Agents Are Reshaping Software Engineering,\" ",[369,1107,738],{},[384,1109,746],{"href":1110,"target":516,"className":1111},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2507.15003",[744,745],[732,1113,1114,1115,753,1117],{"id":781},"A. Vasilopoulos, \"Codified Context: Infrastructure for AI Agents in a Complex Codebase,\" ",[369,1116,738],{},[384,1118,746],{"href":1119,"target":516,"className":1120},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2602.20478",[744,745],[732,1122,1123,1124,753,1126],{"id":791},"N. D. Q. Bui, \"Building Effective AI Coding Agents for the Terminal: Scaffolding, Harness, Context Engineering, and Lessons Learned,\" ",[369,1125,738],{},[384,1127,746],{"href":1128,"target":516,"className":1129},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2603.05344",[744,745],[732,1131,1132,1133,739,1135],{"id":802},"M. Haseeb, \"Context Engineering for Multi-Agent LLM Code Assistants Using Elicit, NotebookLM, ChatGPT, and Claude Code,\" ",[369,1134,738],{},[384,1136,746],{"href":1137,"target":516,"className":1138},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.08322",[744,745],[732,1140,1141,1142,753,1144],{"id":812},"J. Zhang et al., \"Darwin Gödel Machine: Open-Ended Evolution of Self-Improving Agents,\" in ",[369,1143,846],{},[384,1145,746],{"href":1146,"target":516,"className":1147},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2505.22954",[744,745],[732,1149,1150,1151,739,1153],{"id":822},"M. Robeyns et al., \"A Self-Improving Coding Agent,\" ",[369,1152,738],{},[384,1154,746],{"href":1155,"target":516,"className":1156},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2504.15228",[744,745],[732,1158,1159,1160,753,1162],{"id":832},"G. Fang et al., \"Trajectory-Informed Memory Generation for Self-Improving Agent Systems,\" ",[369,1161,738],{},[384,1163,746],{"href":1164,"target":516,"className":1165},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2603.10600",[744,745],{"title":852,"searchDepth":853,"depth":853,"links":1167},[1168,1169,1170,1171,1172],{"id":896,"depth":853,"text":897},{"id":918,"depth":853,"text":919},{"id":965,"depth":853,"text":966},{"id":1018,"depth":853,"text":1019},{"id":718,"depth":853,"text":722},"2026-04-12","How professional developers actually work with coding agents, what happens when projects outgrow a single config file, and why orchestration itself is a transitional skill.",{"src":1176},"https:\u002F\u002Fimages.unsplash.com\u002Fphoto-1503387762-592deb58ef4e?w=800&auto=format&fit=crop",{"authors":1178,"badge":1181,"source":1182},[1179],{"avatar":1180,"name":872,"to":873},{"src":871},{"label":875},{"name":877,"url":873},{"title":166,"description":1174},"ng9uxPUpaoTajXGUXOsMHYLq5yjuexLiLQxbAxiiILc",{"id":1186,"title":138,"body":1187,"date":1590,"description":1591,"extension":864,"image":1592,"meta":1594,"navigation":878,"path":139,"seo":1601,"stem":140,"__hash__":1602,"_path":139},"insights\u002Fnews\u002Finsights\u002Fclosing-the-loop.md",{"type":344,"value":1188,"toc":1580},[1189,1202,1214,1226,1230,1257,1260,1264,1272,1280,1288,1310,1314,1342,1350,1353,1357,1364,1372,1376,1379,1394,1404,1420,1423,1427,1434,1442,1450,1473],[347,1190,351,1192,351,1197],{"className":1191},[350],[353,1193,1196],{"className":1194,"id":1195},[356],"closing-the-loop","Closing the Loop",[359,1198,1201],{"className":1199,"id":1200},[362],"how-human-corrections-can-make-ai-systems-smarter-over-time","How Human Corrections Can Make AI Systems Smarter Over Time",[366,1203,1204,1205,1209,1210,388],{},"Every day, thousands of domain experts in law firms, hospitals, and financial institutions review the outputs of AI systems and quietly fix the mistakes. A legal automation tool misclassifies a contract clause. A clinical decision support system recommends the wrong risk category. A customer service bot generates an irrelevant response. In each case, a human steps in, corrects the output, and moves on. But what happens to those corrections? In most production systems today, the answer is surprisingly little. The same mistakes keep recurring, reviewers grow frustrated, and the promised value of automation slowly erodes ",[381,1206,1207],{},[384,1208,580],{"href":579},". Even at companies with sophisticated ML infrastructure, model update cycles often stretch to months before corrections feed back into training ",[381,1211,1212],{},[384,1213,628],{"href":627},[366,1215,1216,1217,1221,1222,388],{},"The fundamental challenge is architectural. Converting scattered human corrections into durable improvements requires a carefully designed feedback pipeline ",[381,1218,1219],{},[384,1220,387],{"href":386},". That pipeline must respect privacy constraints, handle noisy annotations, and adapt at the right speed for each use case. Recent advances in reinforcement learning, adaptive routing, and noise-robust supervision are making this feedback loop increasingly practical ",[381,1223,1224],{},[384,1225,459],{"href":458},[359,1227,1229],{"id":1228},"the-core-problem-two-timescales-of-improvement","The Core Problem: Two Timescales of Improvement",[347,1231,427,1232,427,1238],{"style":426},[347,1233,431,1234,427],{"style":430},[433,1235],{"src":1236,"alt":1237,"style":437},"https:\u002F\u002Fimages.unsplash.com\u002Fphoto-1540962351504-03099e0a754b?w=800&auto=format&fit=crop","An aircraft cockpit instrument panel filled with analog gauges and dials, analogous to the real-time monitoring layer in an ML correction system that tracks confidence scores, correction rates, and model performance",[347,1239,431,1240,427],{"style":440},[347,1241,443,1242,443,1246,431],{},[445,1243,1245],{"style":447,"id":1244},"like-a-pilots-instrument-panel","Like a Pilot's Instrument Panel",[366,1247,1248,1249,1252,1253,1256],{"style":452},"A pilot monitors altitude and heading in real time, making constant small corrections. But deeper analysis happens only after landing, from mechanical inspections to route adjustments. An effective correction system works the same way. A ",[376,1250,1251],{},"fast loop"," provides immediate, lightweight adjustments without changing the model's core parameters. A ",[376,1254,1255],{},"slow loop"," periodically retrains the model using accumulated, quality-filtered correction data. Mixing these two timescales creates a system that is either too slow to fix obvious errors or too unstable for high-stakes deployment.",[366,1258,1259],{},"Production correction systems also face constraints that academic benchmarks rarely address. Privacy regulations in healthcare and finance may prohibit storing full model outputs, limiting the system to structured metadata about each correction. Annotation quality varies across reviewers, meaning a single careless override can push the model in the wrong direction. In platforms that serve multiple client organizations, different clients may need distinct model behaviors, making a single shared update inappropriate.",[359,1261,1263],{"id":1262},"learning-from-preferences-rlhf-and-dpo","Learning from Preferences: RLHF and DPO",[366,1265,1266,1267,1271],{},"Reinforcement Learning from Human Feedback (RLHF) is one of the most influential approaches to aligning model behavior with human intent. The technique works in two stages ",[381,1268,1269],{},[384,1270,387],{"href":386},". First, it trains a reward model from human preference data, meaning pairs of outputs where a human has indicated which is better. Then it uses reinforcement learning to fine-tune the target model so it produces outputs the reward model scores highly. A landmark demonstration showed that a relatively small RLHF-aligned model could be preferred by human raters over a much larger unaligned model. Alignment through feedback can be more efficient than simply making models bigger.",[366,1273,1274,1275,1279],{},"A notable trend in 2024-2025 is the growing adoption of online iterative RLHF, where feedback is collected continuously from the current model rather than from a pre-collected dataset ",[381,1276,1277],{},[384,1278,459],{"href":458},". This matters because reward models trained on outputs from a previous version of the model often struggle with outputs from the current version. The data goes stale. Online iterative approaches solve this by keeping feedback current, ensuring the training data matches what the model is actually producing now.",[366,1281,1282,1283,1287],{},"A cost-effective variant called RLTHF (Targeted Human Feedback) achieves comparable alignment in benchmark evaluations using only about 6-7% of the typical human annotation effort ",[381,1284,1285],{},[384,1286,387],{"href":386},". It does this by focusing corrections on the hardest samples, the ones the reward model itself flags as uncertain. Whether these efficiency gains hold in production, where error distributions and reviewer behavior differ from controlled benchmarks, remains an open question. But the direction is promising for settings where human review time is the scarcest resource.",[366,1289,1290,1291,1295,1296,1300,1301,1305,1306,388],{},"Direct Preference Optimization (DPO) takes a different path by eliminating the separate reward model entirely ",[381,1292,1293],{},[384,1294,526],{"href":525},". Instead of the two-stage RLHF process, DPO converts preference pairs directly into a training signal for the model. The math works out so that the model can learn the same alignment objective in a single, simpler step. Because DPO skips the reward-model stage, it is substantially more stable and computationally lighter than traditional RLHF, making it practical for teams that batch corrections on a weekly schedule ",[381,1297,1298],{},[384,1299,553],{"href":552},". A comprehensive 2025 survey organizes the growing DPO research into four dimensions covering data strategy, learning framework, constraint mechanisms, and model properties ",[381,1302,1303],{},[384,1304,560],{"href":559},". One important finding is that including ambiguous or difficult preference pairs in training data can actually harm alignment, underscoring the importance of careful data curation ",[381,1307,1308],{},[384,1309,553],{"href":552},[359,1311,1313],{"id":1312},"smart-routing-contextual-bandits-for-model-selection","Smart Routing: Contextual Bandits for Model Selection",[347,1315,427,1316,427,1336],{"style":426},[347,1317,431,1318,427],{"style":440},[347,1319,443,1320,443,1324,431],{},[445,1321,1323],{"style":447,"id":1322},"choosing-where-to-eat-at-machine-speed","Choosing Where to Eat, at Machine Speed",[366,1325,1326,1327,1331,1332,1335],{"style":452},"Imagine walking down a street lined with restaurants. Should the diner return to a familiar spot or try someplace new? This is the exploration-exploitation dilemma, and it is exactly the trade-off that contextual bandits solve for AI systems. These algorithms provide a principled way to route each incoming query to the best-suited model or configuration ",[381,1328,1329],{},[384,1330,536],{"href":535},". The key insight is that in deployment, only the outcome of the ",[369,1333,1334],{},"chosen"," model is observed. The system never learns what would have happened with a different choice, a constraint that most simpler routing approaches ignore.",[347,1337,431,1338,427],{"style":430},[433,1339],{"src":1340,"alt":1341,"style":437},"https:\u002F\u002Fimages.unsplash.com\u002Fphoto-1760742712910-ba9483fbd691?w=800&auto=format&fit=crop","A narrow street lined with restaurants and colorful signage, representing the exploration-exploitation trade-off where an algorithm must choose between familiar options and new possibilities",[366,1343,1344,1345,1349],{},"The BaRP (Bandit-feedback Routing with Preferences) framework, introduced in 2025, treats routing as a balancing act between performance and cost ",[381,1346,1347],{},[384,1348,536],{"href":535},". Operators can adjust that trade-off on the fly without retraining, simply by specifying how much they value accuracy versus cost savings. In preprint results not yet peer-reviewed, experiments across diverse benchmarks show BaRP outperforming strong alternatives by at least 12% while simultaneously reducing costs, and generalizing well to tasks never seen during training.",[366,1351,1352],{},"In a production correction loop, each time a human corrects a model output, that correction updates the router's estimate of how well that model handles similar queries. Over time, the router learns to steer traffic away from models that consistently underperform on certain query types. Each client organization can maintain its own routing preferences, while new clients benefit from patterns already learned across the broader user base.",[359,1354,1356],{"id":1355},"when-corrections-themselves-are-wrong","When Corrections Themselves Are Wrong",[366,1358,1359,1360,388],{},"Human corrections are imperfect. Reviewers vary in expertise, attention, and consistency. A correction loop that treats every override as ground truth will inevitably amplify errors. Programmatic Weak Supervision (PWS) addresses this by treating each labeling source, including each human reviewer, as an imperfect signal whose reliability can be measured and weighted accordingly ",[381,1361,1362],{},[384,1363,601],{"href":600},[366,1365,1366,1367,1371],{},"Recent work has advanced this idea significantly. A 2025 methodology attaches confidence scores to the labels produced by weak supervision systems, enabling the learning pipeline to quantify uncertainty and reduce the influence of unreliable labels ",[381,1368,1369],{},[384,1370,611],{"href":610},". This connects to a broader principle in production ML. Label noise should be treated as a first-class design concern, with explicit mechanisms for detection and mitigation, rather than as a data-cleaning afterthought.",[359,1373,1375],{"id":1374},"putting-it-together-a-correction-to-improvement-pipeline","Putting It Together: A Correction-to-Improvement Pipeline",[366,1377,1378],{},"One way to organize these techniques into a practical architecture is a three-stage correction pipeline, all under a shared governance layer. The specific design draws on patterns from the literature cited above, though the overall structure is an editorial synthesis rather than any single paper's proposal.",[366,1380,1381,1384,1385,1389,1390,388],{},[376,1382,1383],{},"Ingestion and signal processing."," Every corrected output event produces structured metadata (error type, model version, tenant ID, confidence score) written to a permanent event log ",[381,1386,1387],{},[384,1388,628],{"href":627},". Raw corrections then pass through several quality filters, including noise reduction, confidence scoring, and prioritization of the most informative examples, before reaching any model ",[381,1391,1392],{},[384,1393,611],{"href":610},[366,1395,1396,1399,1400,388],{},[376,1397,1398],{},"Fast loop (real-time)."," Between retraining cycles, the fast loop improves behavior without changing the model itself. It injects prompt hints based on common confusion patterns, adds validated corrections to a reference knowledge base the model can consult at query time, updates the routing system's performance estimates, and monitors correction rates in real time ",[381,1401,1402],{},[384,1403,536],{"href":535},[366,1405,1406,1409,1410,1414,1415,1419],{},[376,1407,1408],{},"Slow loop (periodic)."," On a weekly or event-triggered schedule, accumulated preference pairs feed fine-tuning through either DPO ",[381,1411,1412],{},[384,1413,526],{"href":525}," or online RLHF workflows ",[381,1416,1417],{},[384,1418,459],{"href":458},". Updated models must pass a quality check before deployment, verifying that accuracy has not dropped and that correction rates on held-out test samples remain below baseline. Validated updates then roll out gradually, initially serving only 5-10% of traffic before expanding.",[366,1421,1422],{},"A governance layer spans all three stages, enforcing a permanent audit log and filtering of personally identifiable information at ingestion. It also provides independent rollback capabilities for the model and routing system, along with access controls that prevent one client's correction data from leaking to another.",[359,1424,1426],{"id":1425},"open-questions-and-limits","Open Questions and Limits",[366,1428,1429,1430,1433],{},"Not every correction loop should be closed. When correction volume is too low to be statistically meaningful, feeding sparse overrides into training risks overfitting to noise rather than learning genuine patterns. When the operating environment shifts, older corrections may no longer reflect current conditions. And when the task is inherently subjective, with reasonable experts regularly disagreeing on the right answer, consensus-based retraining can suppress legitimate diversity of judgment. Recognizing when ",[369,1431,1432],{},"not"," to retrain is as important as building the pipeline to do so.",[366,1435,1436,1437,1441],{},"Among the challenges that do apply, reward hacking is perhaps the most concerning. Models optimized repeatedly against imperfect reward signals can learn to game the system, producing outputs that score well on the reward model but miss the mark on true human intent ",[381,1438,1439],{},[384,1440,387],{"href":386},". This can be subtle. A customer service model might learn to generate responses that match evaluator style preferences without actually resolving the underlying issue. Detecting this kind of drift requires monitoring not just the reward signal but also downstream task outcomes, an additional layer of instrumentation that many teams underinvest in.",[366,1443,1444,1445,1449],{},"Annotation cost remains a major bottleneck. Even with active learning and targeted feedback strategies like RLTHF, correction loops demand sustained human effort. One promising approach, demonstrated in production at Airbnb, embeds annotation directly into operational workflows rather than treating it as a separate labeling task, compressing model update cycles from months to weeks ",[381,1446,1447],{},[384,1448,628],{"href":627},". AI-generated feedback offers another path toward partial automation at lower cost per data point, but it introduces its own risks and should complement rather than replace human review in high-stakes domains.",[366,1451,1452,1453,1457,1458,1462,1463,1467,1468,1472],{},"The central thesis emerging from recent research is clear. A robust correction loop requires the separation of timescales. Fast-loop mechanisms like prompt hints, retrieval augmentation, and bandit routing deliver immediate responsiveness ",[381,1454,1455],{},[384,1456,536],{"href":535},". Slow-loop mechanisms deliver principled fine-tuning on accumulated, quality-filtered preference data, whether through DPO's single-step approach ",[381,1459,1460],{},[384,1461,526],{"href":525}," or iterative online RLHF pipelines ",[381,1464,1465],{},[384,1466,459],{"href":458},". The convergence of targeted feedback strategies, smart routing, and confidence-aware weak supervision ",[381,1469,1470],{},[384,1471,611],{"href":610}," means that a production-grade human-correction loop is now within reach, for the right kinds of tasks and with clear-eyed awareness of its limits. The organizations that invest in closing this loop will find their AI systems not just tolerating human oversight but actively benefiting from it, getting measurably better with every correction.",[347,1474,351,1476,351,1478],{"className":1475},[718,719],[359,1477,722],{"id":718},[724,1479,427,1481,427,1491,427,1501,427,1512,427,1521,427,1530,427,1539,427,1549,427,1560,427,1569,351],{"className":1480},[727,728,729,730],[732,1482,1483,1484,739,1487],{"id":734},"T. Kaufmann et al., \"A Survey of Reinforcement Learning from Human Feedback,\" ",[369,1485,1486],{},"Transactions on Machine Learning Research",[384,1488,746],{"href":1489,"target":516,"className":1490},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2312.14925",[744,745],[732,1492,1493,1494,1496,1497],{"id":749},"H. Dong et al., \"RLHF Workflow: From Reward Modeling to Online RLHF,\" ",[369,1495,1486],{},", 2024, ",[384,1498,746],{"href":1499,"target":516,"className":1500},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2405.07863",[744,745],[732,1502,1503,1504,1507,1508],{"id":759},"R. Rafailov et al., \"Direct Preference Optimization: Your Language Model is Secretly a Reward Model,\" ",[369,1505,1506],{},"Advances in Neural Information Processing Systems",", vol. 36, 2023. DOI: ",[384,1509,746],{"href":1510,"target":516,"className":1511},"https:\u002F\u002Fdoi.org\u002F10.48550\u002FarXiv.2305.18290",[744,745],[732,1513,1514,1515,739,1517],{"id":771},"W. Wei et al., \"Learning to Route LLMs from Bandit Feedback: One Policy, Many Trade-offs,\" ",[369,1516,738],{},[384,1518,746],{"href":1519,"target":516,"className":1520},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2510.07429",[744,745],[732,1522,1523,1524,1496,1526],{"id":781},"W. Xiao et al., \"A Comprehensive Survey of Direct Preference Optimization: Datasets, Theories, Variants, and Applications,\" ",[369,1525,738],{},[384,1527,746],{"href":1528,"target":516,"className":1529},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2410.15595",[744,745],[732,1531,1532,1533,739,1535],{"id":791},"S. Liu et al., \"A Survey of Direct Preference Optimization,\" ",[369,1534,738],{},[384,1536,746],{"href":1537,"target":516,"className":1538},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2503.11701",[744,745],[732,1540,1541,1542,739,1545],{"id":802},"A. Challapally, C. Pease, R. Raskar, and P. Chari, \"The GenAI Divide: State of AI in Business 2025,\" ",[369,1543,1544],{},"MIT Project NANDA",[384,1546,746],{"href":1547,"target":516,"className":1548},"https:\u002F\u002Fwww.artificialintelligence-news.com\u002Fwp-content\u002Fuploads\u002F2025\u002F08\u002Fai_report_2025.pdf",[744,745],[732,1550,1551,1552,1555,1556],{"id":812},"A. Ratner et al., \"Snorkel: Rapid Training Data Creation with Weak Supervision,\" ",[369,1553,1554],{},"Proceedings of the VLDB Endowment",", vol. 11, pp. 269–282, 2017. DOI: ",[384,1557,746],{"href":1558,"target":516,"className":1559},"https:\u002F\u002Fdoi.org\u002F10.14778\u002F3157794.3157797",[744,745],[732,1561,1562,1563,739,1565],{"id":822},"V. Álvarez et al., \"Reliable Programmatic Weak Supervision with Confidence Intervals for Label Probabilities,\" ",[369,1564,738],{},[384,1566,746],{"href":1567,"target":516,"className":1568},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.03896",[744,745],[732,1570,1571,1572,1575,1576],{"id":832},"C. Zhao et al., \"Agent-in-the-Loop: A Data Flywheel for Continuous Improvement in LLM-based Customer Support,\" ",[369,1573,1574],{},"Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing: Industry Track",", pp. 1919–1930, 2025, ",[384,1577,746],{"href":1578,"target":516,"className":1579},"https:\u002F\u002Faclanthology.org\u002F2025.emnlp-industry.135\u002F",[744,745],{"title":852,"searchDepth":853,"depth":853,"links":1581},[1582,1583,1584,1585,1586,1587,1588,1589],{"id":1200,"depth":853,"text":1201},{"id":1228,"depth":853,"text":1229},{"id":1262,"depth":853,"text":1263},{"id":1312,"depth":853,"text":1313},{"id":1355,"depth":853,"text":1356},{"id":1374,"depth":853,"text":1375},{"id":1425,"depth":853,"text":1426},{"id":718,"depth":853,"text":722},"2026-03-29","Most AI systems throw away the corrections humans make every day. A well-designed feedback pipeline can turn those fixes into lasting improvements.",{"src":1593},"https:\u002F\u002Fimages.unsplash.com\u002Fphoto-1773162148680-e502cdce0bf6?w=800&auto=format&fit=crop",{"authors":1595,"badge":1598,"source":1600},[1596],{"avatar":1597,"name":872,"to":873},{"src":871},{"label":1599},"Reinforcement Learning",{"name":877,"url":873},{"title":138,"description":1591},"jKvhVxJkDNtzutDylJw_Syw0yKkZqa-Pj_gJJP3N5rE",1792582877305]