On this page
The New Category Has a Clear Thesis
A new category of developer tools appeared this year. The pitch is coherent: you already have a Claude Code subscription. Point it at your website.
The agent reads your config, your templates, and your content inventory. You give it a plain English command. Output lands in the right directory with the right frontmatter. You review a diff and ship it. No new subscription. No new UI. No new mental model beyond the session you already have open.
That is a real improvement over the previous generation of content tools. Writing in a CMS, managing plugins, exporting to markdown, fighting deployment pipelines. The new approach flattens all of it.
It is also solving the wrong problem.
The problem is, people are messy. Even if data is structured, it sort of is not truly available in one index.
Satya Nadella
Making messy people faster at producing content does not make the content less messy. It makes it messier, faster.
The Bet Wire Makes
Wire bets differently. Not AI as interface. AI as worker inside a validation fence.
The content team never sees HTML. They write markdown, run a command, and the pipeline handles the rest. But Wire does not hand them a diff and ask if the output looks right. Wire runs 90 lint rules before the diff exists. Frontmatter validation. Orphan detection. Keyword cannibalization checks. Broken internal link resolution. If the structure is wrong, the build fails. Not with a warning. With a refusal.
The new AI-native tools treat the human as the final quality gate. Review the diff. Ship it if it looks right. Wire treats the build system as the final quality gate. The human sets the strategy. The system enforces it.
Both approaches involve AI. The philosophy is opposite.
What the New Category Gets Right
Honest accounting matters here. There are three things the tools arriving this year built that Wire has not.
LLM discovery as first-class output. Every build generates llms.txt (summary) and llms-full.txt (complete markdown), plus a .md copy of every page alongside the HTML. Parallel output tracks: traditional SEO for crawlers, structured markdown for models.
Nadella made this point about organizing layers. Google organized the web. App stores organized mobile. Models are organizing the next layer now. A site that publishes only HTML speaks only to the previous organizing layer. Wire builds beautiful, validated, schema-correct HTML. It does not yet publish in the format the next layer speaks.
A structured interface for tooling. Running a Model Context Protocol server alongside the static build gives AI tools something richer than a file system. Typed resources: your config, your content inventory, your available commands. The agent does not hunt for your config. The config exposes itself.
Wire's audit output already has this structure. HEALTH, ACTION, SEO, INFO. Structured triage for a human to steer. The same knowledge should be callable by any tool that wants to act on it. Not locked inside a terminal session that requires a human to read it first.
Compliance as content architecture. A trust center (SOC 2, ISO 27001, GDPR documentation structured as a first-class collection with nested navigation) is not a design problem. It is a content architecture problem. B2B sites get audited. Their compliance documentation lives in PDFs, Google Docs, and wiki pages that nobody maintains and nobody can build a structured site from.
Wire's validation system exists to enforce that content does not rot unnoticed. A compliance archive that Wire validates and refuses to ship when outdated is not a nice-to-have. For some customers, it is the product.
Organizing the data layer turns out to be probably the most complicated thing, which spans the enterprise such that it can meet the intelligence.
Satya Nadella
The Shape of the Gap
The new tools solved the interface problem. How does a non-developer tell an AI to update a homepage? Answer: plain English, reviewed diff, shipped.
Wire solved the quality problem. How does a content team ship 400 pages without the keyword graph collapsing into cannibalization and the internal links dissolving into orphans? Answer: 30 build gates, 90 lint rules, refuse to ship until the structure holds.
Both are real problems. The interface problem is visible and immediate. The quality problem is invisible and compounds for twelve months until it costs thousands to fix manually.
Wire's bet is that the quality problem is harder and therefore more valuable to own.
The llms.txt convention formalizes what the new tools already ship by default: a structured entry point that tells models what your site is and where to start. The Model Context Protocol specification defines the interface those same tools expose. Wire ignores both today. That is not a position. It is a gap.
What Changes Next
The LLM discovery gap is the most urgently real. llms.txt is not merely an emerging convention. It is the emerging convention. A site-level file that tells models what the site is, where content lives, and what to read first. Wire builds for the current organizing layer. It should build for the next one at the same time, in the same pipeline.
The MCP gap is a small step from the audit system that already exists. Wire knows which pages need merging, which keywords are cannibalizing, which articles are thin. That knowledge should be callable, not just printable. The structured audit report that runs in two seconds is already the right data. The question is who else can consume it.
The trust center gap answers itself. Wire serves B2B customers. B2B customers get audited. An enterprise that cannot publish a structured, up-to-date compliance archive from version-controlled markdown is an enterprise that writes compliance documentation in Word and prays nobody asks follow-up questions. Compliance content should be a first-class build output, validated the same way product pages are validated, refused when orphaned the same way articles are refused.
What Wire Does Not Trade Away
The new tools are fast. Compiled languages. Sub-second builds. Single binaries. No runtime dependencies. The pitch is clean and the benchmarks are real.
Wire is slower to build. It runs more checks. It enforces more structure. It refuses more often.
That is not a deficiency. That is the design constraint.
Wire exists because somebody has to impose the schema that people will not impose on themselves.
Speed without structure ships content faster. Wire ships content that holds together. The difference shows up at page 100, not page 1. At month twelve, not month one. When the site has 400 articles and the keyword cannibalization is systemic and the internal link graph has orphaned entire topic clusters and nobody noticed because nothing ever refused to build.
Gates wanted SQL against everything. Nadella says neural networks solved the pattern problem that schemas could not. Wire sits somewhere between those two positions: structured enough to enforce quality, flexible enough that messy humans can still write content without learning to code.
The new tools make messy faster. Wire makes messy impossible. Both will exist. The interesting question is not which one wins. It is which problem you actually have.
For a content site at page ten, the interface problem is the real problem. Getting content out of someone's head and into the right directory is the bottleneck.
For a content site at page two hundred, the quality problem is the real problem. The interface was never the bottleneck. The structure was.
Wire is not the interface. It is the organizing layer that runs under the interface: the thing that enforces the schema after the human has moved on to the next page.
That is still the bet.