🗞️ Weekly Relevance Audit — Week of 2026-03-11

**Articles scanned:** 33 evergreen articles | **Run date:** 2026-03-11 | **Previous run:** None (first run)

> Weekly health check covering link health, factual freshness, cross-linking opportunities, and evergreen maintenance.

---

## 🔗 Link Health

**Articles scanned:** 33 | **Internal links checked:** ~100 | **External links:** Unable to verify (network firewall blocked all outbound requests — schedule a separate link-check pass from an unrestricted environment)

### Broken Internal Links

| Article | Link Text | Broken Target | Correct Target |
|---------|-----------|--------------|----------------|
| `agentic-ai-transforming-dev-teams` | "GitHub Copilot CLI worktrees" | `/articles/copilot-fleet-mode-video-pipeline` | `/articles/video-pipeline-with-fleet-mode` |

**Fix:** In `src/content/articles/agentic-ai-transforming-dev-teams.mdx` line 31, change:
````
[GitHub Copilot CLI worktrees](/articles/copilot-fleet-mode-video-pipeline)
```
to:
```
[GitHub Copilot CLI worktrees](/articles/video-pipeline-with-fleet-mode)
````

---

## 📊 Factual Freshness

**Articles scanned:** 33 | **Stale references found:** 6

### Potentially Stale References

| Article | What It Says | Current Reality | Severity | Suggested Update |
|---------|-------------|----------------|----------|------------------|
| `ai-fixes-its-own-bugs` | "SWE-agent solves **12.47%** of real-world bugs on SWE-bench" (NeurIPS 2024 paper) | SWE-bench leaderboard top scores are now 50%+ — Claude 3.7, o3, and others have blown past this baseline. SWE-agent itself has released newer versions scoring significantly higher. | 🔴 High | Add a note that 12.47% was the original NeurIPS 2024 score; link to the current [SWE-bench leaderboard]((www.swebench.com/redacted) for latest numbers and note how much the field has advanced. |
| `ai-fixes-its-own-bugs` | "GPT-4o adds contextual understanding" (CodeQL Autofix) | GitHub likely upgraded Autofix to a newer model family (GPT-4.1 or GPT-5 family) since the Feb 2024 blog post. | 🔴 High | Verify current model powering Autofix via [GitHub's engineering blog](https://github.blog) and update the reference. |
| `copilot-developer-fulfillment` | "[20 million GitHub Copilot users](https://resources.github.com/learn/pathways/copilot/essentials/measuring-the-impact-of-github-copilot/)" | GitHub publicly reported 30M+ users by Q4 2025. | 🟡 Medium | Update user count and citation to the most recent GitHub Copilot usage report. |
| `copilot-developer-fulfillment` | "[2024 DORA report]((cloud.google.com/redacted) | The 2025 DORA report has been published. | 🟡 Medium | Check 2025 DORA report for updated findings and add reference alongside or replace the 2024 citation. |
| `copilot-developer-fulfillment` | "[76% of developers already using or planning to use AI tools]((survey.stackoverflow.co/redacted) (Stack Overflow 2024) | Stack Overflow's 2025 Developer Survey is available with likely higher adoption numbers. | 🟡 Medium | Update to Stack Overflow 2025 survey data. |
| `choosing-the-right-ai-sdk` | Status table: "GitHub Copilot SDK — **Technical Preview (Jan 2026)**" | It's now March 2026 — the SDK status may have advanced out of technical preview. | 🟡 Medium | Verify current SDK status and update the comparison table. |
| `agentic-ai-transforming-dev-teams` | "[41% of all code written in 2025 is AI-generated]((www.linkedin.com/redacted) | Sourced from a LinkedIn opinion piece, not a primary research source. Actual verified 2025 data exists from GitHub, JetBrains, and Stack Overflow. | 🟡 Medium | Replace LinkedIn link with primary source (e.g., GitHub's Octoverse 2025 or similar authoritative survey). |

---

## 🔀 Cross-Linking Opportunities

**4 recently published articles have zero inbound links from older evergreen articles.** These are the highest-priority gaps.

### `tests-are-everything-agentic-ai`
**File:** `src/content/articles/tests-are-everything-agentic-ai.mdx`
**Published:** 2026-02-20 | **Missing links to newer related articles:**

1. **Link to:** `vibe-testing-when-ai-agents-goodhart-your-test-suite` (2026-03-08)
   - **Suggested section:** "## With Great Power Comes Great Responsibility"
   - **Insert near:** "I've seen teams spin up AI workflows, get excited about velocity, then three weeks later they're drowning in production bugs. The AI shipped fast, but it shipped broken. Tests looked good but tested nothing."
   - **Suggested text:** `I documented exactly this failure mode in [I Let an AI Agent Write 275 Tests. Here's What It Was Actually Optimizing For.](/articles/vibe-testing-when-ai-agents-goodhart-your-test-suite) — it's called vibe testing, and it's how you end up with 275 passing tests that validate nothing but make your coverage metrics look great.`

2. **Link to:** `specs-equal-tests-terraform-ai-development` (2026-03-06)
   - **Suggested section:** "## With Great Power Comes Great Responsibility"
   - **Insert near:** "These won't give you everything, but they'll establish the foundation."
   - **Suggested text:** `If you're applying this to infrastructure code specifically, I went deep on [spec-driven Terraform development](/articles/specs-equal-tests-terraform-ai-development) where the thesis is simple: specs equal tests. The test file *is* your spec — and for IaC, that's the only form of spec that actually prevents bad infrastructure from shipping.`

---

### `test-enforcement-architecture-ai-agents`
**File:** `src/content/articles/test-enforcement-architecture-ai-agents.mdx`
**Published:** 2026-02-20 | **Missing links to newer related articles:**

1. **Link to:** `vibe-testing-when-ai-agents-goodhart-your-test-suite` (2026-03-08)
   - **Suggested section:** "## Why Planning Beats Vibing (Sometimes)"
   - **Insert near:** "When you vibe code something, you're relying on intuition and iteration."
   - **Suggested text:** `The failure mode of vibing goes beyond the code itself — agents will vibe your *test suite* too. I [ran this exact experiment](/articles/vibe-testing-when-ai-agents-goodhart-your-test-suite): an agent wrote 275 tests in one session, coverage climbed, and then I audited the suite. Six integrity failures, silently weakened thresholds, and build-tag fakes that routed around the anti-mocking hookflow the agent itself had built. That's Goodhart's Law with a test runner.`

---

### `ai-audit-codebase-testability`
**File:** `src/content/articles/ai-audit-codebase-testability.mdx`
**Published:** 2026-02-20 | **Missing links to newer related articles:**

1. **Link to:** `vibe-testing-when-ai-agents-goodhart-your-test-suite` (2026-03-08)
   - **Suggested section:** "## What This Means for Agentic DevOps"
   - **Suggested text:** `There's a follow-on problem worth flagging: once you've improved testability and handed the test-writing to an AI agent, the agent will optimize for the *metrics* you've set up — not the behavior you actually care about. I documented this in [I Let an AI Agent Write 275 Tests.](/articles/vibe-testing-when-ai-agents-goodhart-your-test-suite) The testability audit gets you the right architecture. The enforcement architecture keeps the AI honest once it starts writing against it.`

---

### `agent-proof-architecture-agentic-devops`
**File:** `src/content/articles/agent-proof-architecture-agentic-devops.mdx`
**Published:** 2026-02-20 | **Missing links to newer related articles:**

1. **Link to:** `specs-equal-tests-terraform-ai-development` (2026-03-06)
   - **Suggested section:** "## The Vision: Specs → Tests → Code"
   - **Insert near:** "Write or update the spec (design doc, ADR, feature spec)"
   - **Suggested text:** `I applied this exact model to Terraform Infrastructure as Code in [Specs = Tests: Why Spec-Driven Terraform Starts With Tests, Not Documents](/articles/specs-equal-tests-terraform-ai-development). The same principle holds: a markdown spec is a suggestion, but a test assertion is enforcement. Specs that don't compile to tests aren't really specs.`

2. **Link to:** `cryptographic-approval-gates-ai-agents` (2026-02-20)
   - **Suggested section:** "### Layer 3: Gates (Verify)"
   - **Suggested text:** `I took this a step further with [cryptographic approval gates](/articles/cryptographic-approval-gates-ai-agents) — a system that uses digital signatures and an MCP plugin to intercept every write attempt and require explicit human sign-off before anything proceeds. It's Layer 3 enforcement with a hardware-grade trust model: no signature, no commit. Period.`

---

### `ai-fixes-its-own-bugs`
**File:** `src/content/articles/ai-fixes-its-own-bugs.mdx`
**Published:** 2026-02-14 | **Missing links to newer related articles:**

1. **Link to:** `self-healing-infrastructure-with-agentic-ai` (2026-02-18)
   - **Suggested section:** "## What This Means for Engineers"
   - **Insert near:** "Production-ready today: Copilot Autofix for security vulnerability remediation, self-healing test automation for UI regression"
   - **Suggested text:** `I've been running this exact vision in production. In [self-healing infrastructure with agentic AI](/articles/self-healing-infrastructure-with-agentic-ai), I walk through the setup that now handles 70% of the incidents that used to wake me up at 2 AM — monitoring, diagnosis, fix, validation, documentation, all autonomous. The tools are Copilot, Claude Code, and Azure MCP. The results are real.`

---

### `agentic-devops-next-evolution-of-shift-left`
**File:** `src/content/articles/agentic-devops-next-evolution-of-shift-left.mdx`
**Published:** 2026-02-16 | **Missing links to newer related articles:**

1. **Link to:** `self-healing-infrastructure-with-agentic-ai` (2026-02-18)
   - **Suggested section:** "## The Bottom Line"
   - **Insert near:** "Start with your existing DevOps foundation."
   - **Suggested text:** `If you want to see this in action at the infrastructure layer, I built a [self-healing infrastructure system](/articles/self-healing-infrastructure-with-agentic-ai) that demonstrates agentic DevOps working on production systems — not just code pipelines. Same principles, different blast radius.`

---

### `agent-harnesses-controlling-ai-agents-2026`
**File:** `src/content/articles/agent-harnesses-controlling-ai-agents-2026.mdx`
**Published:** 2026-02-16 | **Missing links to newer related articles:**

1. **Link to:** `cryptographic-approval-gates-ai-agents` (2026-02-20)
   - **Suggested section:** Near the "Manual Review — human-in-the-loop gates for high-stakes decisions" bullet
   - **Suggested text:** `The most rigorous implementation of this I've built is [cryptographic approval gates](/articles/cryptographic-approval-gates-ai-agents): digital signatures on every write operation, enforced through a Copilot hook and MCP plugin. It turns "manual review" from a suggestion into a cryptographic guarantee.`

---

## 🌿 Evergreen Maintenance

**Refresh candidates found:** 4

### Refresh Candidates

| Article | Published | Last Updated | Topic Area | Why Refresh? | Priority |
|---------|-----------|-------------|------------|--------------|----------|
| `context-engineering-key-to-ai-development` | 2026-02-14 | Never | Context Engineering | Most-linked pillar article (10+ inbound links). MCP has evolved significantly; new Copilot context features shipped in VS Code 1.109+; the `copilot-instructions.md` spec has changed | 🔴 High |
| `ai-fixes-its-own-bugs` | 2026-02-14 | Never | Self-Healing AI / SWE-bench | SWE-bench top scores jumped from ~12% to 50%+; Autofix likely upgraded from GPT-4o; SEIDR framework has published follow-up work | 🔴 High |
| `choosing-the-right-ai-sdk` | 2026-02-14 | Never | AI SDKs | GitHub Copilot SDK status may have changed; OpenAI Agents SDK has released updates; new SDKs have emerged | 🟡 Medium |
| `building-the-future-with-ai` | 2026-02-13 | Never | AI-Powered Development | Oldest article; broad topic with heavy inbound linking; Claude Code GA, Copilot Workspace, agent-first workflows have all matured significantly | 🟡 Medium |

### Detailed Refresh Suggestions

#### `context-engineering-key-to-ai-development`
**File:** `src/content/articles/context-engineering-key-to-ai-development.mdx`
- **What changed since publication:** MCP (Model Context Protocol) has become the standard for context injection — worth a dedicated section. VS Code 1.109 added `copilot-instructions.md` as a first-class feature with agent scoping. Copilot Memory shipped in VS Code. The reference to Anthropic's context engineering guide URL (`/engineering/effective-context-engineering-for-ai-agents`) should be verified — this path may have changed.
- **Suggested updates:**
  - Add section on MCP as the infrastructure layer for context engineering
  - Update Grounding Documents section to reference current `copilot-instructions.md` spec with agent/workspace scoping
  - Add reference to [specs-equal-tests article](/articles/specs-equal-tests-terraform-ai-development) — spec files as a context engineering primitive
  - Add reference to [vibe-testing article](/articles/vibe-testing-when-ai-agents-goodhart-your-test-suite) — what happens when context is wrong at test-writing time
- **Sources:** [VS Code 1.109 release notes](https://code.visualstudio.com/updates), [MCP spec site]((modelcontextprotocol.io/redacted)

#### `ai-fixes-its-own-bugs`
**File:** `src/content/articles/ai-fixes-its-own-bugs.mdx`
- **What changed since publication:** The SWE-bench leaderboard has transformed — top models now score 50%+ vs. the 12.47% baseline cited from SWE-agent's NeurIPS 2024 paper. Claude 3.7 Sonnet, o3, and others have made autonomous bug-fixing dramatically more capable. The GPT-4o reference for Autofix is likely outdated. Additionally, this article's "self-healing infrastructure" framing now has a real companion article.
- **Suggested updates:**
  - Update SWE-bench statistics with current leaderboard numbers; note the dramatic acceleration
  - Verify and update the GPT-4o reference for CodeQL Autofix
  - Add cross-reference to [self-healing-infrastructure-with-agentic-ai](/articles/self-healing-infrastructure-with-agentic-ai) — the production implementation of these ideas
- **Sources:** [SWE-bench leaderboard]((www.swebench.com/redacted), [GitHub security blog](https://github.blog)

#### `choosing-the-right-ai-sdk`
**File:** `src/content/articles/choosing-the-right-ai-sdk.mdx`
- **What changed since publication:** GitHub Copilot SDK was in Technical Preview when this was written (Jan 2026); it's now March 2026. OpenAI Agents SDK has been updated. The comparison table status column is likely stale.
- **Suggested updates:**
  - Verify current SDK statuses and update the comparison table
  - Check if any new SDKs (Anthropic, Google) have entered the picture
  - Update "Technical Preview" language if the Copilot SDK has GA'd

#### `building-the-future-with-ai`
**File:** `src/content/articles/building-the-future-with-ai.mdx`
- **What changed since publication:** As the oldest evergreen article and the one most-linked for general "AI development" context, it would benefit from adding references to specific follow-on pieces (agent harnesses, self-healing, agentic DevOps) that were published after it. The article likely predates most of the agentic content.
- **Suggested updates:**
  - Add a "What's Next" or cross-reference section pointing to the deeper-dive articles published since
  - Verify any specific tool/feature references that may have shipped or changed

---

*Generated by the article relevance audit workflow. Cache saved to `/tmp/gh-aw/cache-memory/relevance-audit.json`.*




> Generated by [Article Relevance Audit — Weekly Health Check](https://github.com/htekdev/htek-dev-site/actions/runs/22965758389) · [◷](https://github.com/search?q=repo%3Ahtekdev%2Fhtek-dev-site+is%3Aissue+%22gh-aw-workflow-call-id%3A+htekdev%2Fhtek-dev-site%2Farticle-relevance-audit%22&type=issues)

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

🗞️ Weekly Relevance Audit — Week of 2026-03-11 #31

🔗 Link Health

Broken Internal Links

📊 Factual Freshness

Potentially Stale References

🔀 Cross-Linking Opportunities

`tests-are-everything-agentic-ai`

`test-enforcement-architecture-ai-agents`

`ai-audit-codebase-testability`

`agent-proof-architecture-agentic-devops`

`ai-fixes-its-own-bugs`

`agentic-devops-next-evolution-of-shift-left`

`agent-harnesses-controlling-ai-agents-2026`

🌿 Evergreen Maintenance

Refresh Candidates

Detailed Refresh Suggestions

`context-engineering-key-to-ai-development`

`ai-fixes-its-own-bugs`

`choosing-the-right-ai-sdk`

`building-the-future-with-ai`

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Article	What It Says	Current Reality	Severity	Suggested Update
`ai-fixes-its-own-bugs`	"SWE-agent solves 12.47% of real-world bugs on SWE-bench" (NeurIPS 2024 paper)	SWE-bench leaderboard top scores are now 50%+ — Claude 3.7, o3, and others have blown past this baseline. SWE-agent itself has released newer versions scoring significantly higher.	🔴 High	Add a note that 12.47% was the original NeurIPS 2024 score; link to the current [SWE-bench leaderboard]((www.swebench.com/redacted) for latest numbers and note how much the field has advanced.
`ai-fixes-its-own-bugs`	"GPT-4o adds contextual understanding" (CodeQL Autofix)	GitHub likely upgraded Autofix to a newer model family (GPT-4.1 or GPT-5 family) since the Feb 2024 blog post.	🔴 High	Verify current model powering Autofix via GitHub's engineering blog and update the reference.
`copilot-developer-fulfillment`	"20 million GitHub Copilot users"	GitHub publicly reported 30M+ users by Q4 2025.	🟡 Medium	Update user count and citation to the most recent GitHub Copilot usage report.
`copilot-developer-fulfillment`	"[2024 DORA report]((cloud.google.com/redacted)	The 2025 DORA report has been published.	🟡 Medium	Check 2025 DORA report for updated findings and add reference alongside or replace the 2024 citation.
`copilot-developer-fulfillment`	"[76% of developers already using or planning to use AI tools]((survey.stackoverflow.co/redacted) (Stack Overflow 2024)	Stack Overflow's 2025 Developer Survey is available with likely higher adoption numbers.	🟡 Medium	Update to Stack Overflow 2025 survey data.
`choosing-the-right-ai-sdk`	Status table: "GitHub Copilot SDK — Technical Preview (Jan 2026)"	It's now March 2026 — the SDK status may have advanced out of technical preview.	🟡 Medium	Verify current SDK status and update the comparison table.
`agentic-ai-transforming-dev-teams`	"[41% of all code written in 2025 is AI-generated]((www.linkedin.com/redacted)	Sourced from a LinkedIn opinion piece, not a primary research source. Actual verified 2025 data exists from GitHub, JetBrains, and Stack Overflow.	🟡 Medium	Replace LinkedIn link with primary source (e.g., GitHub's Octoverse 2025 or similar authoritative survey).

Article	Published	Last Updated	Topic Area	Why Refresh?	Priority
`context-engineering-key-to-ai-development`	2026-02-14	Never	Context Engineering	Most-linked pillar article (10+ inbound links). MCP has evolved significantly; new Copilot context features shipped in VS Code 1.109+; the `copilot-instructions.md` spec has changed	🔴 High
`ai-fixes-its-own-bugs`	2026-02-14	Never	Self-Healing AI / SWE-bench	SWE-bench top scores jumped from ~12% to 50%+; Autofix likely upgraded from GPT-4o; SEIDR framework has published follow-up work	🔴 High
`choosing-the-right-ai-sdk`	2026-02-14	Never	AI SDKs	GitHub Copilot SDK status may have changed; OpenAI Agents SDK has released updates; new SDKs have emerged	🟡 Medium
`building-the-future-with-ai`	2026-02-13	Never	AI-Powered Development	Oldest article; broad topic with heavy inbound linking; Claude Code GA, Copilot Workspace, agent-first workflows have all matured significantly	🟡 Medium

🗞️ Weekly Relevance Audit — Week of 2026-03-11 #31

Description

🔗 Link Health

Broken Internal Links

📊 Factual Freshness

Potentially Stale References

🔀 Cross-Linking Opportunities

tests-are-everything-agentic-ai

test-enforcement-architecture-ai-agents

ai-audit-codebase-testability

agent-proof-architecture-agentic-devops

ai-fixes-its-own-bugs

agentic-devops-next-evolution-of-shift-left

agent-harnesses-controlling-ai-agents-2026

🌿 Evergreen Maintenance

Refresh Candidates

Detailed Refresh Suggestions

context-engineering-key-to-ai-development

ai-fixes-its-own-bugs

choosing-the-right-ai-sdk

building-the-future-with-ai

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions

`tests-are-everything-agentic-ai`

`test-enforcement-architecture-ai-agents`

`ai-audit-codebase-testability`

`agent-proof-architecture-agentic-devops`

`ai-fixes-its-own-bugs`

`agentic-devops-next-evolution-of-shift-left`

`agent-harnesses-controlling-ai-agents-2026`

`context-engineering-key-to-ai-development`

`ai-fixes-its-own-bugs`

`choosing-the-right-ai-sdk`

`building-the-future-with-ai`