The Scalably blog

AI agents in business. Agent engineering in practice.

Written from systems we build and run: business implementation and agent engineering first, with SEO operations as one measured vertical underneath.

5 posts Implement agents

AI agents inside real business operations

Choose the work, connect current data, keep authority narrow, verify the result, and measure what actually finished.

Explore business implementation →
22 posts Build agents

Agent engineering across MCP, Claude, OpenAI, and Codex

Working code, SDKs, tool boundaries, coding-agent workflows, and architecture choices from systems we use and operate.

Explore agent engineering →
Field report Run models locally

Qwen3.8-27B NVFP4 on two RTX 5090s in production

The exact vLLM configuration, fourteen days of production counters, the failure modes, and the official recipe compared. Every number traced.

Read the field report →
25 posts Run SEO operations

SEO execution, measured against the live site

White-label delivery, internal linking, audits, reporting, and rank tracking as one practical business vertical.

Explore SEO operations →
Current field notes

Start with the work, then choose the stack

The production field report, the business method, the Codex migration, and the technical guides already earning search traffic.

Qwen3.8-27B NVFP4 on two RTX 5090s: production field report

What Qwen, Unsloth, vLLM and Blackwell each contribute, the exact serving configuration, fourteen days of production counters labelled as observations, the failure modes, and the official recipe compared.

Read the field report →

AI agents for business: what works in practice

Start with a recurring job, separate current facts from model judgment, set the permission boundary, and prove the result where the business receives it.

Read the business implementation guide →

AI agents in agriculture: from field update to decision

A privacy-safe operating architecture for work orders, voice and photo intake, authoritative records, human approval, and recovery.

Read the agriculture field guide →

Codex vs Claude Code: why I moved my workflow

Personal coding and orchestration moved to Codex. The production Claude Agent SDK runtime did not. The distinction matters.

Read the comparison →

Codex skills: build reusable workflows that trigger

A complete SKILL.md, trigger tests, deterministic checks, and the point where a local skill should become a plugin.

Build a Codex skill →

Claude Code settings.json: the configuration guide

The site's proven search entry point: file precedence, permission rules, environment settings, and multi-client boundaries.

Read the configuration guide →

AI agents for SEO: 98% completion over four months

A measured business implementation: what the agents did, what stayed human, and where the failures came from.

Read the production field note →
Browse by subject

One publication, four working layers

Business outcomes sit above agent runtimes, tool infrastructure, and specialist operating systems.

Business implementation

Put agents into real operations

Workflow selection, privacy, permissions, rollout, cost, and measured completion.

Agent foundations

Build the loop and the tools

Start with the control flow, then add MCP and production-safe integrations.

Coding agents and SDKs

Operate Codex and Claude

Configuration, reusable workflows, subagents, hooks, and honest platform choices.

SEO operations

Automate work that can be verified

Reporting, audits, rank tracking, internal links, and white-label delivery.