{
  "schema_version": "newruntime-topic-hub-v0.1",
  "type": "topic_hub",
  "slug": "coding-agents",
  "title": "Coding agents - New Runtime",
  "description": "A living map of coding agents as they move from autocomplete into supervised software work, review loops, environments, and operating cost.",
  "answer": [
    "Coding agents are becoming operational workers inside software teams, not just better code completion.",
    "The durable bottlenecks are review bandwidth, environment setup, context quality, permissions, and cost per completed task.",
    "The useful question is no longer whether an agent can write code, but whether the surrounding harness can verify and land work repeatedly."
  ],
  "search_intents": [
    "coding agent workflow",
    "AI coding agent verification",
    "software engineering agents"
  ],
  "status": "featured",
  "last_updated": "2026-07-22",
  "counts": {
    "total": 49,
    "signals": 39,
    "patterns": 1,
    "posts": 6,
    "atlas": 3,
    "sources": 81
  },
  "routes": {
    "html": "https://newruntime.com/topics/coding-agents/",
    "markdown": "https://newruntime.com/topics/coding-agents.md",
    "json": "https://newruntime.com/topics/coding-agents.json"
  },
  "top_sources": [
    "https://addyo.substack.com/p/own-the-outer-loop",
    "https://addyosmani.com/blog/new-sdlc-vibe-coding",
    "https://agent-cookbook.com/tutorial/ai-20260512163240-54x4",
    "https://agentos-sdk.dev/",
    "https://anthropic.com/news/claude-sonnet-5",
    "https://anthropic.com/research/AI-assistance-coding-skills",
    "https://anthropic.com/research/claude-code-expertise",
    "https://arxiv.org/abs/2207.06991",
    "https://arxiv.org/abs/2412.09871",
    "https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-6-flash-3-5-flash-lite-3-5-flash-cyber/",
    "https://claude.com/blog/agent-view-in-claude-code",
    "https://claude.com/blog/artifacts-in-claude-code"
  ],
  "patterns": [
    {
      "kind": "Pattern",
      "slug": "verification-bandwidth-is-the-scarce-resource",
      "title": "Verification bandwidth is the scarce engineering resource",
      "description": "The primary bottleneck in agentic software delivery is moving from code production to the human and machine capacity required to verify it.",
      "date": "2026-07-16",
      "topics": [
        "coding-agents",
        "evals",
        "engineering-management"
      ],
      "source_count": 5,
      "metric": "high",
      "url": "https://newruntime.com/patterns/verification-bandwidth-is-the-scarce-resource/"
    }
  ],
  "field_notes": [
    {
      "kind": "Field Note",
      "slug": "devin-outposts-agent-execution-plane",
      "title": "Devin Outposts Splits the Agent Brain from the Execution Plane",
      "description": "Devin Outposts moves command execution, repository access, and sandbox lifecycle into customer-controlled infrastructure while the agent loop remains in Cognition's cloud.",
      "date": "2026-07-22",
      "topics": [
        "coding-agents",
        "agent-runtime",
        "infrastructure"
      ],
      "source_count": 4,
      "url": "https://newruntime.com/posts/devin-outposts-agent-execution-plane/"
    },
    {
      "kind": "Field Note",
      "slug": "gemini-3-6-flash-agent-efficiency",
      "title": "Gemini 3.6 Flash Moves the Agent Race Toward Cost per Task",
      "description": "Google's Gemini 3.6 Flash release frames the model race around token efficiency, built-in computer use, and specialized cyber agents rather than raw chat intelligence alone.",
      "date": "2026-07-22",
      "topics": [
        "gemini",
        "agent-economics",
        "coding-agents"
      ],
      "source_count": 1,
      "url": "https://newruntime.com/posts/gemini-3-6-flash-agent-efficiency/"
    },
    {
      "kind": "Field Note",
      "slug": "ai-coding-workflow-verifiable-work",
      "title": "AI Coding Workflow: From Idea to Verifiable Work",
      "description": "AI coding works better when checkable artifacts stand between the idea and the code: specs, tickets, TDD, fresh-context review, and manual QA.",
      "date": "2026-07-21",
      "topics": [
        "coding-agents",
        "workflow",
        "verification"
      ],
      "source_count": 2,
      "url": "https://newruntime.com/posts/ai-coding-workflow-verifiable-work/"
    },
    {
      "kind": "Field Note",
      "slug": "coding-agent-cost-environment-config",
      "title": "Coding Agent Cost Is Cut in Environment Config, Not Prompts",
      "description": "Claude Code and Codex costs are reduced by environment design, not by asking the agent to read less: output filtering, repo maps, model routing, and stable prompt caching.",
      "date": "2026-07-21",
      "topics": [
        "coding-agents",
        "cost-control",
        "prompt-caching"
      ],
      "source_count": 3,
      "url": "https://newruntime.com/posts/coding-agent-cost-environment-config/"
    },
    {
      "kind": "Field Note",
      "slug": "pillar-sandbox-escapes",
      "title": "Coding Agent Sandboxes Break in Places Teams Do Not Expect",
      "description": "Pillar shows that agent sandboxes must be assessed not only around the agent process, but around files, configs, allowlisted commands, and local daemons the host later trusts.",
      "date": "2026-07-21",
      "topics": [
        "agent-security",
        "coding-agents",
        "sandbox"
      ],
      "source_count": 1,
      "url": "https://newruntime.com/posts/pillar-sandbox-escapes/"
    },
    {
      "kind": "Field Note",
      "slug": "kimi-k3-frontend-model-race",
      "title": "Kimi K3: Frontend Is Becoming the Model Race Arena",
      "description": "Kimi K3 moves open model competition into visual software engineering: frontend benchmarks require not only code, but layout, screenshots, accessibility, and human preference.",
      "date": "2026-07-21",
      "topics": [
        "open-models",
        "frontend",
        "coding-agents"
      ],
      "source_count": 2,
      "url": "https://newruntime.com/posts/kimi-k3-frontend-model-race/"
    }
  ],
  "raw_signals": [
    {
      "kind": "Raw Signal",
      "slug": "opencode-model-swappable-agent-harness",
      "title": "OpenCode makes the model a swappable dependency",
      "description": "OpenCode keeps rules, skills, permissions, sessions, and tools stable while the underlying model changes.",
      "date": "2026-07-16",
      "topics": [
        "coding-agents",
        "harness-engineering",
        "model-routing",
        "developer-tools"
      ],
      "source_count": 4,
      "metric": "structural",
      "url": "https://newruntime.com/signals/opencode-model-swappable-agent-harness/"
    },
    {
      "kind": "Raw Signal",
      "slug": "verification-capacity-coding-agent-bottleneck",
      "title": "Verification capacity becomes the coding-agent bottleneck",
      "description": "Coding agents can parallelize implementation faster than engineering teams can expand review judgment.",
      "date": "2026-07-16",
      "topics": [
        "coding-agents",
        "evals",
        "agent-ops",
        "engineering-management"
      ],
      "source_count": 4,
      "metric": "structural",
      "url": "https://newruntime.com/signals/verification-capacity-coding-agent-bottleneck/"
    },
    {
      "kind": "Raw Signal",
      "slug": "fable-cost-per-completed-task",
      "title": "A pricier model can be cheaper per completed task",
      "description": "Cognition reports that Fable 5 completed coding work with fewer steps and output tokens than its previous lead model.",
      "date": "2026-07-15",
      "topics": [
        "agent-economics",
        "coding-agents",
        "model-routing"
      ],
      "source_count": 1,
      "metric": "structural",
      "url": "https://newruntime.com/signals/fable-cost-per-completed-task/"
    },
    {
      "kind": "Raw Signal",
      "slug": "airbnb-agentic-coding-loop",
      "title": "Airbnb frames agentic coding as a loop inside development",
      "description": "The workflow separates planning, delegated implementation, verification, and integration instead of treating the coding agent as the whole process.",
      "date": "2026-07-14",
      "topics": [
        "coding-agents",
        "developer-workflow",
        "agent-loops"
      ],
      "source_count": 1,
      "metric": "notable",
      "url": "https://newruntime.com/signals/airbnb-agentic-coding-loop/"
    },
    {
      "kind": "Raw Signal",
      "slug": "claude-code-model-and-effort-are-separate-controls",
      "title": "Claude Code separates model choice from effort",
      "description": "Anthropic exposes model selection and effort level as different controls for capability, token use, latency, and persistence.",
      "date": "2026-07-13",
      "topics": [
        "coding-agents",
        "model-routing",
        "agent-economics"
      ],
      "source_count": 2,
      "metric": "notable",
      "url": "https://newruntime.com/signals/claude-code-model-and-effort-are-separate-controls/"
    },
    {
      "kind": "Raw Signal",
      "slug": "cheap-code-moves-bottleneck-to-review",
      "title": "Cheap code moves the engineering bottleneck to review",
      "description": "An agent-heavy development model treats implementation as abundant while specifications, validation, security, and integration remain scarce.",
      "date": "2026-07-12",
      "topics": [
        "engineering-management",
        "coding-agents",
        "verification"
      ],
      "source_count": 1,
      "metric": "structural",
      "url": "https://newruntime.com/signals/cheap-code-moves-bottleneck-to-review/"
    },
    {
      "kind": "Raw Signal",
      "slug": "databricks-benchmarks-real-coding-agent-economics",
      "title": "Databricks benchmarks coding agents on its own codebase",
      "description": "Databricks evaluates agents on fresh internal pull-request tasks and measures success alongside runtime, tokens, and cost.",
      "date": "2026-07-12",
      "topics": [
        "coding-agents",
        "evals",
        "agent-economics"
      ],
      "source_count": 1,
      "metric": "structural",
      "url": "https://newruntime.com/signals/databricks-benchmarks-real-coding-agent-economics/"
    },
    {
      "kind": "Raw Signal",
      "slug": "large-task-planning-becomes-ticket-graph",
      "title": "Large-task planning moves from chat into a ticket graph",
      "description": "Matt Pocock's workflow turns discovery, specification, tickets, implementation, and review into durable linked artifacts.",
      "date": "2026-07-12",
      "topics": [
        "coding-agents",
        "planning",
        "task-graphs",
        "agent-skills"
      ],
      "source_count": 1,
      "metric": "notable",
      "url": "https://newruntime.com/signals/large-task-planning-becomes-ticket-graph/"
    },
    {
      "kind": "Raw Signal",
      "slug": "an-ai-coding-workflow-starts-with-data-and-interfaces",
      "title": "An AI Coding Workflow Starts with Data and Interfaces",
      "description": "ThePrimeagen's workflow separates evidence gathering, interface design, and implementation so the coding agent receives a constrained problem instead of an open-ended prompt.",
      "date": "2026-07-11",
      "topics": [
        "coding-agents",
        "specification",
        "workflow"
      ],
      "source_count": 1,
      "metric": "notable",
      "url": "https://newruntime.com/signals/an-ai-coding-workflow-starts-with-data-and-interfaces/"
    },
    {
      "kind": "Raw Signal",
      "slug": "trycase-disposable-linux-agent-verification",
      "title": "TryCase gives coding agents disposable Linux verification",
      "description": "TryCase creates short-lived Linux environments where agents can run generated code and tests away from the developer's machine.",
      "date": "2026-07-10",
      "topics": [
        "sandboxes",
        "coding-agents",
        "verification"
      ],
      "source_count": 1,
      "metric": "notable",
      "url": "https://newruntime.com/signals/trycase-disposable-linux-agent-verification/"
    },
    {
      "kind": "Raw Signal",
      "slug": "parallel-agents-make-large-migrations-tractable",
      "title": "Parallel coding agents make large migrations tractable",
      "description": "A large Bun pull request shows many independent agents working against explicit file boundaries and a common test contract.",
      "date": "2026-07-08",
      "topics": [
        "coding-agents",
        "parallel-agents",
        "software-migrations"
      ],
      "source_count": 2,
      "metric": "notable",
      "url": "https://newruntime.com/signals/parallel-agents-make-large-migrations-tractable/"
    },
    {
      "kind": "Raw Signal",
      "slug": "pxpipe-trades-text-tokens-for-rendered-visual-context",
      "title": "pxpipe Trades Text Tokens for Rendered Visual Context",
      "description": "The proxy compresses large tool results and old context into images, reducing input cost while exposing precision limits for byte-exact data.",
      "date": "2026-07-08",
      "topics": [
        "context-compression",
        "token-cost",
        "coding-agents"
      ],
      "source_count": 4,
      "metric": "notable",
      "url": "https://newruntime.com/signals/pxpipe-trades-text-tokens-for-rendered-visual-context/"
    },
    {
      "kind": "Raw Signal",
      "slug": "claude-code-recovers-after-failed-command",
      "title": "Claude Code can recover after a failed command",
      "description": "A runtime update lets the coding agent inspect a command failure and continue its loop without waiting for a generic human prompt.",
      "date": "2026-07-07",
      "topics": [
        "coding-agents",
        "error-recovery",
        "agent-loops"
      ],
      "source_count": 1,
      "metric": "incremental",
      "url": "https://newruntime.com/signals/claude-code-recovers-after-failed-command/"
    },
    {
      "kind": "Raw Signal",
      "slug": "openwiki-codebase-docs-as-agent-memory",
      "title": "OpenWiki treats repository documentation as agent memory",
      "description": "OpenWiki generates and refreshes a structured codebase explanation that other agents and humans can inspect before acting.",
      "date": "2026-07-07",
      "topics": [
        "codebase-memory",
        "documentation",
        "coding-agents"
      ],
      "source_count": 2,
      "metric": "notable",
      "url": "https://newruntime.com/signals/openwiki-codebase-docs-as-agent-memory/"
    },
    {
      "kind": "Raw Signal",
      "slug": "herdr-adds-agent-awareness-to-terminal-multiplexing",
      "title": "Herdr Adds Agent Awareness to Terminal Multiplexing",
      "description": "Herdr organizes parallel coding-agent sessions around projects, process state, and operator navigation rather than presenting them as undifferentiated terminal panes.",
      "date": "2026-07-04",
      "topics": [
        "coding-agents",
        "orchestration",
        "terminal"
      ],
      "source_count": 2,
      "metric": "notable",
      "url": "https://newruntime.com/signals/herdr-adds-agent-awareness-to-terminal-multiplexing/"
    },
    {
      "kind": "Raw Signal",
      "slug": "twenty-four-gigabytes-of-vram-becomes-a-practical-local-coding-tier",
      "title": "Twenty-Four Gigabytes of VRAM Becomes a Practical Local Coding Tier",
      "description": "Local coding models are becoming useful on high-end consumer hardware, creating a viable private lane even when frontier cloud agents remain stronger overall.",
      "date": "2026-07-04",
      "topics": [
        "local-models",
        "coding-agents",
        "hardware"
      ],
      "source_count": 1,
      "metric": "notable",
      "url": "https://newruntime.com/signals/twenty-four-gigabytes-of-vram-becomes-a-practical-local-coding-tier/"
    },
    {
      "kind": "Raw Signal",
      "slug": "sonnet-moves-agent-capability-downmarket",
      "title": "Sonnet moves agent capability down the price curve",
      "description": "Anthropic positions Claude Sonnet 5 for planning, terminal work, browser use, and multi-step agent tasks at a lower tier.",
      "date": "2026-07-01",
      "topics": [
        "models",
        "agent-economics",
        "coding-agents"
      ],
      "source_count": 1,
      "metric": "notable",
      "url": "https://newruntime.com/signals/sonnet-moves-agent-capability-downmarket/"
    },
    {
      "kind": "Raw Signal",
      "slug": "warp-reframes-product-engineers-as-factory-engineers",
      "title": "Warp Reframes Product Engineers as Factory Engineers",
      "description": "An AI-native engineering organization invests more effort in task decomposition, agent infrastructure, review systems, and throughput management than in direct implementation.",
      "date": "2026-06-30",
      "topics": [
        "engineering-org",
        "coding-agents",
        "orchestration"
      ],
      "source_count": 1,
      "metric": "notable",
      "url": "https://newruntime.com/signals/warp-reframes-product-engineers-as-factory-engineers/"
    },
    {
      "kind": "Raw Signal",
      "slug": "agentos-lightweight-runtime-for-coding-agents",
      "title": "agentOS packages a lightweight runtime for coding agents",
      "description": "Rivet's agentOS combines isolated execution, durable workflows, resource limits, and host-side secrets in a WebAssembly and Rust runtime.",
      "date": "2026-06-29",
      "topics": [
        "agent-runtime",
        "sandboxes",
        "coding-agents"
      ],
      "source_count": 3,
      "metric": "notable",
      "url": "https://newruntime.com/signals/agentos-lightweight-runtime-for-coding-agents/"
    },
    {
      "kind": "Raw Signal",
      "slug": "loop-library-packages-fifteen-repeatable-coding-agent-cycles",
      "title": "Loop Library Packages Fifteen Repeatable Coding-Agent Cycles",
      "description": "A public repository turns common maintenance and delivery tasks into explicit loops with commands, checks, and completion conditions.",
      "date": "2026-06-28",
      "topics": [
        "loop-engineering",
        "coding-agents",
        "automation"
      ],
      "source_count": 2,
      "metric": "notable",
      "url": "https://newruntime.com/signals/loop-library-packages-fifteen-repeatable-coding-agent-cycles/"
    },
    {
      "kind": "Raw Signal",
      "slug": "shopify-treats-code-as-output-and-guardrails-as-engineering",
      "title": "Shopify Treats Code as Output and Guardrails as Engineering",
      "description": "The engineering role shifts toward defining constraints, interfaces, tests, and product intent while agents perform a larger share of implementation.",
      "date": "2026-06-28",
      "topics": [
        "engineering-org",
        "guardrails",
        "coding-agents"
      ],
      "source_count": 1,
      "metric": "notable",
      "url": "https://newruntime.com/signals/shopify-treats-code-as-output-and-guardrails-as-engineering/"
    },
    {
      "kind": "Raw Signal",
      "slug": "cursor-is-expanding-from-ide-to-agent-factory",
      "title": "Cursor Is Expanding from IDE to Agent Factory",
      "description": "Cursor's workflow increasingly centers on parallel autonomous tasks, remote execution, review, and orchestration rather than one developer editing one file at a time.",
      "date": "2026-06-27",
      "topics": [
        "cursor",
        "coding-agents",
        "orchestration"
      ],
      "source_count": 1,
      "metric": "notable",
      "url": "https://newruntime.com/signals/cursor-is-expanding-from-ide-to-agent-factory/"
    },
    {
      "kind": "Raw Signal",
      "slug": "glm-5-2-changes-the-economics-of-claude-code-workflows",
      "title": "GLM-5.2 Changes the Economics of Claude Code Workflows",
      "description": "A capable open model routed through compatible infrastructure can replace premium inference for parts of a coding-agent workload.",
      "date": "2026-06-27",
      "topics": [
        "open-models",
        "model-routing",
        "coding-agents"
      ],
      "source_count": 4,
      "metric": "notable",
      "url": "https://newruntime.com/signals/glm-5-2-changes-the-economics-of-claude-code-workflows/"
    },
    {
      "kind": "Raw Signal",
      "slug": "parallel-agent-workflow-needs-operator-stack",
      "title": "Parallel coding agents need an operator stack",
      "description": "Kun Chen's workflow combines durable project context, task plans, worktrees, independent review, tests, and cleanup around multiple agents.",
      "date": "2026-06-27",
      "topics": [
        "coding-agents",
        "parallel-agents",
        "developer-workflow"
      ],
      "source_count": 2,
      "metric": "notable",
      "url": "https://newruntime.com/signals/parallel-agent-workflow-needs-operator-stack/"
    }
  ],
  "atlas_records": [
    {
      "kind": "Project",
      "slug": "coding-agent-cost-controls",
      "title": "Coding Agent Cost Controls",
      "description": "A forkable setup for reducing coding-agent cost by shaping environment output, repo maps, model routing, cache stability, and task-level budgets.",
      "date": "2026-07-21",
      "topics": [
        "coding-agents",
        "cost-control"
      ],
      "source_count": 2,
      "metric": "B2",
      "url": "https://newruntime.com/projects/coding-agent-cost-controls/"
    },
    {
      "kind": "Shift",
      "slug": "goal-scoped-agent-loops",
      "title": "Manual continue prompting -> goal-scoped agent loops",
      "description": "Long-running agent work is moving from manual continue prompts toward explicit goals, limits, subagents, schedules, and verifiable completion.",
      "date": "2026-07-21",
      "topics": [
        "coding-agents",
        "agent-ops",
        "automation"
      ],
      "source_count": 3,
      "metric": "emerging",
      "url": "https://newruntime.com/shifts/goal-scoped-agent-loops/"
    },
    {
      "kind": "Project",
      "slug": "verifiable-ai-coding-workflow",
      "title": "Verifiable AI Coding Workflow",
      "description": "A small agent-assisted coding workflow where the useful artifact is not generated code, but a checked path from idea to tests, diff, review, and handoff.",
      "date": "2026-07-21",
      "topics": [
        "coding-agents",
        "developer-workflow"
      ],
      "source_count": 1,
      "metric": "B2",
      "url": "https://newruntime.com/projects/verifiable-ai-coding-workflow/"
    }
  ]
}