feat(content): rewrite blogs for readers and improve editorial covers

.evolve/skill-runs.jsonl

@@ -9,3 +9,4 @@
 {"skill":"/signal-distill","ts":"2026-08-02T00:21:12Z","project":"tangle-website-blog","target":"Tangle blog SEO/AEO/GEO query map and SERP research","operatorPrompt":"","durationMin":null,"verdict":"PASS","dispatchedTo":"/stop","operatorOverride":null,"transcriptPath":null,"traceDir":null}
 {"skill":"/docs-slop-audit","ts":"2026-08-02T00:21:20Z","project":"tangle-website-blog","target":"85-post reader and SEO discovery audit; search research notes","operatorPrompt":"","durationMin":null,"verdict":"PASS","dispatchedTo":"/reflect","operatorOverride":null,"transcriptPath":null,"traceDir":null}
 {"skill":"/docs-slop-audit","ts":"2026-08-02T00:23:30Z","project":"tangle-website-blog","target":"SEO search research notes after seo-engine capability check","operatorPrompt":"","durationMin":null,"verdict":"PASS","dispatchedTo":"/reflect","operatorOverride":null,"transcriptPath":null,"traceDir":null}
+{"skill":"/bad","ts":"2026-08-03T07:07:30Z","project":"tangle-website","target":"tangle blog cover and article page","operatorPrompt":"","durationMin":null,"verdict":"PASS","dispatchedTo":"/reflect","operatorOverride":null,"transcriptPath":null,"traceDir":null}

src/content/blog/agent-profile-materialization-contracts.mdx

@@ -12,9 +12,9 @@ tags:
   - ai-infrastructure
 series: Agent Runtime Infrastructure
 seriesOrder: 1
-coverImage: /images/covers/agent-intent-infrastructure.svg
-heroImage: /images/covers/agent-intent-infrastructure.svg
-imageAlt: 'An agent profile divided into identity, model, prompt, tools, and resource fields'
+coverImage: /images/covers/editorial/agent-runtime-v2.webp
+heroImage: /images/covers/editorial/agent-runtime-v2.webp
+imageAlt: 'An editorial still life about the software that runs an agent'
 ---
 
 Imagine saving an agent profile that says, “Use this model, these tools, this prompt, and these files.”

src/content/blog/agent-readiness-index-methodology.mdx

@@ -10,6 +10,9 @@ tags:
   - agents
   - evals
   - api
+coverImage: /images/covers/editorial/instrument-problem-v2.webp
+heroImage: /images/covers/editorial/instrument-problem-v2.webp
+imageAlt: 'An editorial still life about measuring whether an agent can use a changing API'
 ---
 
 Developers increasingly hand integration work to coding agents.
@@ -27,7 +30,7 @@ This post explains how it works, what a score means, and what it cannot tell you
 A developer evaluating a payments provider today often starts by asking an agent to wire up a checkout flow.
 Here is what happens if the agent writes from memory.
 Stripe renamed the Checkout Session display-mode values on 2026-03-25: `embedded` became `embedded_page`.
-An agent recalling the pre-2026 contract sends `ui_mode=embedded` and gets a 400.
+An agent recalling the older contract sends `ui_mode=embedded` and gets a 400.
 The developer does not see a stale model.
 The developer sees a payments API that returned an error on the vendor's own quickstart flow, and tries the next vendor.
 
@@ -120,7 +123,7 @@ All 9 were discarded and re-run.
 
 Detection was not the fix.
 The mock's source file is now deleted from disk the moment the server boots; the process keeps its loaded copy, so there is nothing left to read.
-The cheat is impossible now, not merely detected.
+The cheat is impossible now, more than detected.
 
 The general lesson: if the grader or a reference implementation is reachable from the agent's execution environment, a capable agent will find it, and nothing in the score will tell you.
 Treat any benchmark without transcript-level auditing and a leak-proof grading boundary as contaminated.
@@ -140,7 +143,7 @@ A candidate task that fails any one of them is not admitted.
 
 Configuration-level results are published on the board itself rather than reproduced here: a leaderboard pasted into a post goes stale the moment the suite is re-run or the metric is corrected, and a post that disagrees with the board is worse than no post. The live table, including per-task detail and the roster behind each row, is at [/benchmarks/stripe](/benchmarks/stripe).
 
-The shape worth understanding is that **rows are configurations, not models** — a model plus the CLI agent driving it — because the tooling moves the score materially. The same model driven by two different agents can differ by tens of points on identical tasks, which is why a "model leaderboard" for integration work is misleading.
+The shape worth understanding is that **rows are configurations, not models**; a model plus the CLI agent driving it; because the tooling moves the score materially. The same model driven by two different agents can differ by tens of points on identical tasks, which is why a "model leaderboard" for integration work is misleading.
 
 A pass means the agent's code executed correctly against the current contract.
 Each configuration attempted each task twice, and each attempt is a single agent session with no retries and no human help.
@@ -152,7 +155,7 @@ The sample is small.
 Every board states the attempt count behind each row; at the tens-of-attempts scale these suites run, treat gaps under roughly 10 points as noise rather than a ranking.
 
 Coverage is not always identical across rows.
-A configuration can carry fewer attempts than its peers when a cell was lost to an infrastructure failure and quarantined rather than scored. The board reports the real denominator per row, so check it before comparing two configurations — an unequal denominator is the single easiest way to misread a benchmark.
+A configuration can carry fewer attempts than its peers when a cell was lost to an infrastructure failure and quarantined rather than scored. The board reports the real denominator per row, so check it before comparing two configurations; an unequal denominator is the single easiest way to misread a benchmark.
 
 This is one API and 12 tasks, all Python HTTP clients.
 It measures readiness to integrate Stripe's current API, not general coding ability, repo-scale migration, or UI work.

src/content/blog/agent-runtime-environments.mdx

@@ -1,5 +1,5 @@
 ---
-title: 'What an AI Agent Needs Beyond a Model'
+title: 'AI Agent Runtime Environment: What an Agent Needs Beyond a Model'
 slug: agent-runtime-environments
 summary: >-
   A plain-language map of the infrastructure an agent needs to route models, run tools, use a browser, remember state, accept payment, and return evidence.
@@ -11,9 +11,9 @@ tags:
   - infrastructure
 series: Agent Intent Infrastructure
 seriesOrder: 7
-coverImage: /images/covers/agent-intent-infrastructure.svg
-heroImage: /images/covers/agent-intent-infrastructure.svg
-imageAlt: 'An AI agent surrounded by model, execution, browser, payment, and evidence boundaries'
+coverImage: /images/covers/editorial/agent-intent-v2.webp
+heroImage: /images/covers/editorial/agent-intent-v2.webp
+imageAlt: 'An editorial still life about describing and running an agent task'
 ---
 
 A model can write a response from text.

src/content/blog/agent-runtime-generic-executor-deletion.mdx

@@ -1,5 +1,5 @@
 ---
-title: 'Removing a Second Agent Path Without Hiding the Losses'
+title: 'AI Agent Runtime Architecture: Removing a Second Execution Path'
 slug: agent-runtime-generic-executor-deletion
 summary: >-
   We removed a second program for starting coding agents only after a shared path could deliver the same saved settings and expose the missing run signals.
@@ -12,9 +12,9 @@ tags:
   - software-architecture
 series: Agent Runtime Infrastructure
 seriesOrder: 2
-coverImage: /images/covers/agent-intent-infrastructure.svg
-heroImage: /images/covers/agent-intent-infrastructure.svg
-imageAlt: 'A shared path replacing a backend-specific path for starting an agent'
+coverImage: /images/covers/editorial/agent-runtime-v2.webp
+heroImage: /images/covers/editorial/agent-runtime-v2.webp
+imageAlt: 'An editorial still life about the software that runs an agent'
 ---
 
 Suppose two pieces of software start the same kind of coding job.

src/content/blog/agent-runtime-worker-observability.mdx

@@ -1,5 +1,5 @@
 ---
-title: 'A Running Agent Is Not Observable Until Its Work Is Visible'
+title: 'AI Agent Observability: Make the Work Visible'
 slug: agent-runtime-worker-observability
 summary: >-
   The shared agent path now exposes live progress and requested tool calls, while leaving completion and duration unknown when the remote service never sends them.
@@ -12,9 +12,9 @@ tags:
   - coding-agents
 series: Agent Runtime Infrastructure
 seriesOrder: 3
-coverImage: /images/covers/agent-intent-infrastructure.svg
-heroImage: /images/covers/agent-intent-infrastructure.svg
-imageAlt: 'A live agent worker stream showing progress and recorded tool requests'
+coverImage: /images/covers/editorial/agent-runtime-v2.webp
+heroImage: /images/covers/editorial/agent-runtime-v2.webp
+imageAlt: 'An editorial still life about the software that runs an agent'
 ---
 
 When an automated program runs for several minutes, silence is not a useful status.

src/content/blog/ai-accountant-complex-tax-returns.mdx

@@ -11,16 +11,16 @@ tags:
   - complex-tax
 series: Building an AI Tax Agent
 seriesOrder: 4
-coverImage: /images/products/tax-agent-app.png
-heroImage: /images/products/tax-agent-app.png
-imageAlt: 'Tangle Tax Agent workspace showing source-backed accounting workpapers and review controls'
+coverImage: /images/covers/editorial/ai-tax-v2.webp
+heroImage: /images/covers/editorial/ai-tax-v2.webp
+imageAlt: 'An editorial still life about preparing a complex tax return with software'
 ---
 
-An AI accountant for complex tax returns is useful when it behaves like a preparer of evidence: it organizes documents, reconciles facts, drafts calculations, identifies missing records, and creates review questions for the taxpayer or advisor. It is not useful when it gives confident tax answers without workpapers. For founders, investors, crypto users, and multi-state households, the value is leverage before professional review, not replacing every professional judgment call.
+An AI accountant for complex tax returns is useful when it behaves like a preparer of evidence: it organizes documents, reconciles facts, drafts calculations, identifies missing records, and creates review questions for the taxpayer or advisor. It is not useful when it gives confident tax answers without workpapers. For founders, investors, crypto users, and multi-state households, the value is use before professional review, not replacing every professional judgment call.
 
 This post is not tax advice. Treat AI accountant output as preparation material until the taxpayer or qualified advisor reviews it.
 
-## What An AI Accountant Should Actually Do
+## What An AI Accountant Should Do
 
 | Job | Good output |
 | --- | --- |

src/content/blog/ai-agent-sandbox.mdx

@@ -11,12 +11,12 @@ tags:
   - infrastructure
 series: Agent Intent Infrastructure
 seriesOrder: 2
-coverImage: /images/covers/agent-intent-infrastructure.svg
-heroImage: /images/covers/agent-intent-infrastructure.svg
-imageAlt: 'Agent runtime diagram showing sandbox files, processes, network policy, snapshots, and evidence'
+coverImage: /images/covers/editorial/agent-intent-v2.webp
+heroImage: /images/covers/editorial/agent-intent-v2.webp
+imageAlt: 'An editorial still life about describing and running an agent task'
 ---
 
-An AI agent sandbox is an isolated runtime where an agent can create files, run processes, call tools, use the network under policy, preserve state, and return evidence. A code interpreter, browser automation session, or serverless job runner is too narrow for that job. Tangle Sandbox gives agents a machine-shaped workspace for real work: install dependencies, execute tests, inspect artifacts, recover from failure, and keep the dangerous parts contained. Start with [Tangle Sandbox](https://github.com/tangle-network/tcloud) when the agent needs an environment, not only an API.
+An AI agent sandbox is an isolated runtime where an agent can create files, run processes, call tools, use the network under policy, preserve state, and return evidence. A code interpreter, browser automation session, or serverless job runner is too narrow for that job. Tangle Sandbox gives agents a machine-shaped workspace for real work: install dependencies, execute tests, inspect artifacts, recover from failure, and keep the dangerous parts contained. Start with [Tangle Sandbox](https://github.com/tangle-network/tcloud) when the agent needs an environment, more than an API.
 
 The hard part of agent infrastructure is not letting a model produce text. It is giving that model a place to act without handing it your laptop, production credentials, or a shared build server.
 
@@ -31,12 +31,7 @@ The hard part of agent infrastructure is not letting a model produce text. It is
 
 Tangle's sandbox contract is intentionally machine-readable. Agents can discover it from the product manifest, package metadata, and health endpoint before trying to run work.
 
-```bash
-npm install @tangle-network/sandbox
-curl -fsS https://sandbox.tangle.tools/health
-curl -fsS https://sandbox.tangle.tools/.well-known/tangle-agent.json
-curl -fsS https://sandbox.tangle.tools/v1/public-templates
-```
+Use the public Sandbox manifest and package documentation to inspect the current service contract. Check the health result and available templates as part of a real workload, not as a substitute for one.
 
 As of this post, the health endpoint returns `{"status":"ok"}` and the public template endpoint is safe to call. Treat the template list as discovery, not as a guarantee that a specific template slug exists.
 

src/content/blog/ai-browser-testing-evidence-traces.mdx

@@ -11,12 +11,12 @@ tags:
   - ai-browser-testing
 series: Browser Agent
 seriesOrder: 4
-coverImage: /images/products/browser-agent-app.png
-heroImage: /images/products/browser-agent-app.png
-imageAlt: 'Tangle Browser Agent workspace showing browser test run, screenshots, actions, and run evidence'
+coverImage: /images/covers/editorial/browser-agent-v2.webp
+heroImage: /images/covers/editorial/browser-agent-v2.webp
+imageAlt: 'An editorial still life about an agent operating a web browser'
 ---
 
-AI browser testing should not mean "ask a model if the page works." It should mean the agent drives the browser, observes the page, takes actions, verifies the goal, and saves the evidence. Without artifacts, a passing run is just a story. With artifacts, product and QA teams can inspect the exact page state that led to the result.
+AI browser testing should not mean "ask a model if the page works." It should mean the agent drives the browser, observes the page, takes actions, verifies the goal, and saves the evidence. Without artifacts, a passing run is only a story. With artifacts, product and QA teams can inspect the exact page state that led to the result.
 
 Tangle Browser Agent is built around that evidence loop. For the broader automation model, read [Browser Automation For AI Agents](/blog/browser-automation-for-ai-agents).
 
@@ -37,14 +37,7 @@ Traditional browser automation has strong standards. [WebDriver](https://www.w3.
 
 The normal run path is direct:
 
-```bash
-npm install -g @tangle-network/browser-agent-driver
-npx playwright install chromium
-
-bad run \
-  --url https://example.com \
-  --goal "Create an account, finish onboarding, and verify the dashboard loads"
-```
+Start with the public Browser Agent documentation and run one representative flow. Save the screenshots, page state, actions, and outcome so another person can inspect the result.
 
 After the run, inspect the evidence:
 

src/content/blog/ai-code-audit-sandboxed-agents.mdx

@@ -11,9 +11,9 @@ tags:
   - ai-audit
 series: Code Auditor
 seriesOrder: 1
-coverImage: /images/covers/agent-intent-infrastructure.svg
-heroImage: /images/covers/agent-intent-infrastructure.svg
-imageAlt: 'Code auditor agent workspace showing repository audit, sandboxed tools, subagents, and finding validation'
+coverImage: /images/covers/editorial/code-auditor-v2.webp
+heroImage: /images/covers/editorial/code-auditor-v2.webp
+imageAlt: 'An editorial still life about reviewing code and producing evidence'
 ---
 
 AI code audit is useful only when the agent can inspect the repository, run tools, test exploit paths, and show why each finding is real. A model that reads code and writes confident prose is not an auditor. Tangle Code Auditor is the upcoming product surface at `audit.tangle.tools`; until that domain is live, public copy should describe the audit runtime without linking to a product URL.
@@ -35,22 +35,7 @@ This is closer to an audit workflow than a scanner workflow. Scanners such as [C
 
 ## Example Audit Command
 
-```bash
-pnpm redteam audit --repo https://github.com/org/protocol
-pnpm redteam audit --repo https://github.com/org/protocol --ref main --focus "reentrancy,flash-loans" --json
-```
-
-The command is only valuable if the report connects each issue to files, exploit path, command output, and a fix direction.
-
-## Sandbox-First Design
-
-Security audits execute unknown code. That should happen in a managed environment, not on a developer laptop with production credentials.
-
-```bash
-pnpm redteam sandbox create --capability evm-foundry
-pnpm redteam sandbox exec <sandbox-id> "forge test"
-pnpm redteam sandbox destroy <sandbox-id>
-```
+Use an isolated audit runner and preserve the report, affected files, reproduction evidence, and proposed fix. The exact command depends on the repository and scanner version.
 
 For the runtime layer behind this pattern, read [LLM Sandbox Environment For Agent Runs](/blog/llm-sandbox-environment-agent-runtime). For scanner triage, read [AI Vulnerability Scanner Vs Agent Audit](/blog/ai-vulnerability-scanner-vs-agent-audit).
 

src/content/blog/ai-coding-assistant-deployment-evidence.mdx

@@ -11,12 +11,12 @@ tags:
   - developer-tools
 series: Blueprint Agent
 seriesOrder: 2
-coverImage: /images/products/blueprint-agent-app.png
-heroImage: /images/products/blueprint-agent-app.png
-imageAlt: 'Blueprint Agent workspace showing AI coding session, project files, quest verification, and deployment controls'
+coverImage: /images/covers/editorial/blueprint-agent-v2.webp
+heroImage: /images/covers/editorial/blueprint-agent-v2.webp
+imageAlt: 'An editorial still life about a runnable service blueprint'
 ---
 
-An AI coding assistant becomes useful to a developer program when it can prove more than "the model wrote code." The bar is running code, verified tasks, deployment evidence, and a session trace that a partner team can review. Tangle Blueprint Agent is aimed at that job: give developers an isolated coding workspace with partner SDKs, indexed docs, AI help, and code-based quest verification.
+A trace is the record of an agent run, including its actions and results. An AI coding assistant becomes useful to a developer program when it can prove more than "the model wrote code." The bar is running code, verified tasks, deployment evidence, and a session trace that a partner team can review. Tangle Blueprint Agent is aimed at that job: give developers an isolated coding workspace with partner SDKs, indexed docs, AI help, and code-based quest verification.
 
 This is different from a generic IDE chat panel. The product surface at [ai.tangle.tools](https://ai.tangle.tools) is designed around partner onboarding and real build outcomes.