Blog / ArchGenerator
Architecture in practice.
Decisions, specifications and software that evolves. Perspectives from the ArchGenerator team for the people building it.
Articles appear in their original languages. Browse all languages
Latest articles
It worked yesterday. What about today?
Why use ArchGenerator if I already have Claude Code, Codex, or Cursor. The agent is a bricklayer; ArchGenerator is the blueprint, the foreman, the inspector and the report — and this is how I prove it.
AI Makes Code Cheaper. Building the Wrong Thing Still Costs.
Lessons from Austin Tech Week: how better context, clearer decisions, and avoided rework can turn faster AI coding into useful software delivery.
PHP2Shell: why your composer.lock won't save you from CVE-2026-109618
The PHP2Shell flaws are patched in the runtime, not in your application code. Here is why the ticket that says "upgrade the dependency" closes nothing, and what a remediation path that survives an audit actually looks like.
Most read
All articles
It worked yesterday. What about today?
Why use ArchGenerator if I already have Claude Code, Codex, or Cursor. The agent is a bricklayer; ArchGenerator is the blueprint, the foreman, the inspector and the report — and this is how I prove it.
AI Makes Code Cheaper. Building the Wrong Thing Still Costs.
Lessons from Austin Tech Week: how better context, clearer decisions, and avoided rework can turn faster AI coding into useful software delivery.
PHP2Shell: why your composer.lock won't save you from CVE-2026-109618
The PHP2Shell flaws are patched in the runtime, not in your application code. Here is why the ticket that says "upgrade the dependency" closes nothing, and what a remediation path that survives an audit actually looks like.
AI Risk Gap: architecture prepares software to evolve with AI
From Readiness Gap to software evolution: how ArchGenerator supports diagnosis, remediation and specifications, with a perspective on Gartner’s research.
Superintelligence: Potential Becomes Real When It Transforms How We Work
Discover how ArchGenerator.com connects technical knowledge, architecture, and specifications to help teams build software with greater engineering maturity.
Your AI Writes Code in Minutes. Where Does Your Team Lose Weeks?
Discover how ArchGenerator.com connects architecture, specifications, and AI coding to help teams avoid rework and reclaim delivery capacity.
Your AI Added a Library. Who Approved the Risk?
An AI-generated feature can introduce more than code. Turn new dependencies into informed architecture decisions while keeping your preferred coding AI.
Your Scanner Found 200 Alerts. Which One Should Delay the Release?
Turn security alerts into engineering decisions with reproducible evidence, system context, business impact, and verifiable remediation.
Your AI Can Change Production. Who Decided It Should?
Give coding agents room to work with clear permission levels, approval gates, and recovery plans grounded in your architecture.
Still Explaining Your Project to AI? Give Your Repository a Memory.
Make AGENTS.md useful: concise instructions, approved architecture, and verifiable checks. Keep your coding AI and give it better project context.
Five AI Agents. One Wrong Assumption. Five Times the Rework?
More agents need clearer responsibilities. Learn when specialization helps and how to keep one wrong assumption from spreading.
Your Coding Agent Worked for Six Hours. What Can You Actually Ship?
An active agent is not necessarily a productive one. Evaluate long coding tasks by accepted outcomes, rework, and architectural fit.
Your AI Can Read the Code. Who Owns the Decisions?
Stop paying to rediscover your product. Turn scattered decisions into shared context that survives every AI coding session.
An SBOM Is Not a Compliance Strategy. Make It a Living Process.
A software bill of materials is only useful when teams can connect components to owners, vulnerabilities, decisions, and remediation evidence.
Modular Monolith or Microservices? Let Your Product Decide.
Choose architecture from evidence: team boundaries, release cadence, scaling needs, data consistency, and operational capacity—not fashion.
Your AI Coding Model Is Not the System. The Harness Is.
Two teams can use the same AI model and get radically different results. The difference is the harness governing context, permissions, memory, and proof.
MCP in production: connecting tools is easy; controlling authority is the real problem
Connecting an agent to tools takes minutes. Defining what it may see, do, and prove is the production work. Use this practical model to limit authority without losing autonomy.
10x with AI—or only 2x? How to measure productivity without counting lines and PRs
AI can multiply code and still delay delivery. A practical scorecard measures what matters: accepted outcomes, rework, risk, durability, and total cost.
Is AI making architects more important—or simply changing their job?
When implementation becomes cheap, decisions about boundaries, contracts, data, and evolution become more valuable. The architect's job is shifting from drawing the future to making it verifiable.
Spend Intelligence Where the Work Is Uncertain
Enterprise agents become more reliable when models focus on judgment instead of repeating mechanical orchestration. The practical pattern is to let code, files, and explicit state handle the routine work while the model concentrates on ambiguity and decisions.
How to Start an AI Initiative Without Losing Control of Your Product
Many companies know they need to introduce AI into their products, but few know where to begin. The right starting point is not a model: it is a clear understanding of the system, the business problem, the available data, and the decisions that AI is allowed to influence.
From Spec: The Three Modes of ArchGenerator
From a greenfield spec to an evidence-backed report, and from the report to an executable repair kit. See ArchGenerator's three modes in one flow.
Follow and republish with RSS
Follow new articles in your feed reader. The RSS feed includes publications in every language.
https://archgenerator.com/blog/rss.xml ↗
To republish, use the RSS integration on dev.to (Settings → Extensions → Publishing from RSS), or Medium's “Import a story” option with the article URL. Keep the link to the original publication.