AI and Automation

AI agents, MCP, AI-native products, Claude and Anthropic stack, automation pipelines.

Page 2 of 3

A short document card on the left, and a straight white path running between two railed guides to a lit tile, with quiet paths drifting away above and below

How to write a CLAUDE.md file that steers Claude Code reliably

CLAUDE.md loads into every Claude Code session, so every line competes for attention. Keep it under 120 lines, write positive rules, and move fixed checks to hooks.

A dim interface screen on the left wired by white lines to a lit, indented tree of nodes on the right

Designing for AI agents: building a UI the non-human user can read

AI agents read your site through the accessibility tree, the structure screen readers use. In a 2026 study, agent task success fell from 78% to 42% on a degraded one.

Left, a pyramid of dark tiles above a stack of report sheets and a long timeline; right, three lit tiles above a lit product window and a short timeline

AI consulting firm vs AI studio: which fits your stage in 2026

Big Four AI engagements run $300K to $2M+ over 6 to 18 months. A boutique AI studio ships in 8 to 12 weeks for $30K to $80K. Which one fits your stage.

Three white tiles at the top wired down into a glowing horizontal band, from which many thin lines fan out to a wide row of lit tiles below

What is a Claude Code studio in 2026 (and how to pick one)

A Claude Code studio builds production software using Claude as the execution layer. We define the label and show how to verify a real one before you sign.

A lit server tile inside a countdown ring with a quarter remaining, beside eight staggered short bars fading from old to new, only the newest lit

Short-lived MCP server URLs: the TTL pattern for AI agents in 2026

1,467 MCP servers are exposed on the public internet with no auth, per Trend Micro. Short-lived URLs with a TTL of minutes, not days, close the hole.

Four rows of small dark tiles, two lit, each wired down to a context bar: forty-six thin quiet slices and two wide lit blocks

Claude Code skills: when to use them, when they cost tokens

Claude Code skills load on demand and cost 75-150 tokens each just to list. Use them for procedures over five lines; keep project-wide rules in CLAUDE.md.

A slide split into three panels showing a tile set into a stack, a rising line against a target, and a tile passed from one frame to another, above ten lit checks and five grey flags

AI integration agency in 2026: what their pitch should actually contain

An AI integration agency pitch in 2026 is judged on three things: what gets deployed, how it gets measured, and how it gets handed back. Ten checks to listen for, five red flags to walk out on.

On a white ground, a central tile with twelve spokes out to twelve black tiles in a ring, each spoke broken by a gap marked with a white ring

12 mistakes we see teams make building their first multi-agent ops system

Multi-agent LLM systems fail in production at 41 to 86 percent. Most of the failures trace back to twelve specific decisions teams make in the first month. Here is what they look like and how to undo them.

A lit tile at the centre of a ring of eight checked white tiles joined to it by spokes, above a row of fourteen segments with the first six solid

Claude integration partner: what to look for before you sign

A Claude integration partner ships production AI on Anthropic's stack. Here are the 8 checks, the 6 to 14 week timeline, and when you should not hire one.

Five dark cards in a row, each holding a different network of lit nodes (a chain, a hub, a tree, a loop, a fan), all joined below to one white rail

Multi-agent system design patterns that survive production

Multi-agent systems cost 15 times more than chat. 40% of pilots fail in 6 months. Five design patterns decide if your AI agent team ships or stalls.

A line of lit calls passing a gate, one grey call turned down to a stop, a lit tool tile, ragged pills combed into straight ones, and a ringing signal at the end

Claude Code hooks: PreToolUse guards, PostToolUse formatters

Three hooks in .claude/settings.json: a PreToolUse guard that exits 2 and cancels the command, a PostToolUse formatter, and a Stop event notifier.

Left, a block of dark tiles with a lit tile bolted to its side and a line rising to flatten under a heavy ceiling; right, a block of lit tiles wired to a glowing centre, its line rising through a broken ceiling

AI bolted-on vs AI-native: why the retrofit hits a ceiling

Bolt-on AI hits a ceiling because the data flow is the constraint, not the model. How to tell which one you have, and when a retrofit still makes sense.

Back to Field notes