The Blog
Claude
Anthropic’s Claude: agents, Claude Code, and the architecture behind them. Plans, limits, and what it is genuinely better at.
27 entries
September 24, 2026
The Day One of My AI Agents Briefed Another One
Orchestration stops being an abstraction the first time you see it happen. One question in chat, and underneath it an agent writing another agent's instructions without anyone typing them.
September 23, 2026
Most Things Sold as AI Agents Are Not Agents
Most of what is being sold as an agent is a prompt with a good interface. That is not a failure and it is not an insult, it is a different product with a different price, and being accurate about it saves you money.
September 15, 2026
Copilot or Claude for Business? Which One, and When
Not a winner. Copilot wins on where the work already lives, Claude wins on long documents and reasoning, and since Microsoft added Anthropic models to Copilot the question changed shape entirely.
September 14, 2026
Can You Use Claude Inside Microsoft Copilot? Yes
Microsoft added Anthropic's models to Copilot. You can pick Claude in Word, Excel, PowerPoint, Chat and Copilot Studio. Here is what that changes, and the one setting that has to be switched on first.
September 12, 2026
How We Build a Website With Claude, GitHub, and Netlify
The real workflow behind this platform: Claude drafts it, GitHub tracks it, Netlify previews it, and I review every single change before it ever goes live.
August 27, 2026
What to Learn About AI Agents
You drive every day and have never rebuilt an engine. Same with agents. Three things are worth learning, none about the model.
August 27, 2026
Where Do I Start With AI Agents? Pick the Job, Not the App
The most common blocker isn't access to AI, it's not knowing where to begin. Three things make a job worth handing to an agent first, and none of them are about which app you choose.
August 18, 2026
What Is AI Agent Context? Why Your AI Sounds Like Everyone Else
An agent with no context does not write like you. It writes like the average of everyone, because that is the only thing it was given. What context actually is, what belongs in it, and why generic output is a supply problem rather than a model problem.
August 18, 2026
What Do AI Agents Actually Cost? Reading the Bill
Cost isn't per agent, it's per run and per token. Which means the expensive line is rarely the thing you meant to automate. It's the small job you forgot is running eleven hundred times a day.
August 18, 2026
What Is an AI Agent Handoff? A Folder Isn't a Handoff
Work between agents doesn't flow downhill on its own. A handoff needs a named receiver and a signal, or the finished thing sits in a place nobody is watching. What a real handoff contains and how work goes missing without one.
August 18, 2026
What Are AI Agent Skills? The Third Time You Ask
A skill is a procedure an agent writes down once and reuses, instead of working the same job out from scratch every time. The third identical request is the signal. What a skill is, when to make one, and why it is not the same as a prompt you saved.
August 18, 2026
What Is a Knowledge Base for AI? Two Agents, Two Answers
When two agents answer the same question differently and both sound certain, the problem is that they read different things, not the agents. What a knowledge base is, why a folder isn't one, and what it costs to skip.
August 18, 2026
What Is a Multi-Agent System? When One Agent Isn't Enough
One agent doing five jobs isn't efficient, it's understaffed. What a multi-agent system actually is, why it's a staffing decision rather than a technical one, and the signal that tells you it's time.
August 18, 2026
What Is MCP? How AI Agents Actually Reach Your Apps
A login is built for a person with hands, eyes and a screen. An agent has none of those. MCP is the standard socket that lets an agent plug into an app directly, and it's the difference between a tool and a screenshot.
August 5, 2026
Agent Instructions: What They Are, and Why an Agent Is Useless Without Them
Instructions are where you say what you actually want done. Without them an agent has access to everything and no reason to touch any of it. Where instructions live, what they need to contain, and why having them does not by itself make something an agent.
August 5, 2026
Giving an Agent Hands: What Tools Actually Change
An agent with no tools can only talk. Tools are access to the calendar, the files, the inbox, whatever the job needs. It is the line between advice and action, and it is where the real risk lives.
August 5, 2026
Triggers: What Starts an Agent When You Don't
A trigger is what starts the work without you typing anything. It is the actual line between using a convenient tool and having something work for you, and it is the piece most people skip.
August 4, 2026
ITT: Instructions, Tools, Triggers
Nina's test for whether something is actually an agent. Three parts, about a minute to apply, and it tells you more than the product demo will.
August 4, 2026
Orchestration, Explained Without the Buzzword
Orchestration means someone or something decides which agent or tool handles which part of a job. Without it you have a pile of tools that don't talk to each other. The real shift is when the routing decision stops being yours.
August 4, 2026
Real AI Agent or Just a Wrapper?
Three questions settle it before you pay agent prices for prompt value. Calling something a wrapper isn't an insult, it's accurate.
August 4, 2026
What Is an AI Agent? Plainly, and Without the Hype
Not a chatbot, not an automation, not a clever prompt. What the word actually means, how it differs from the things it gets confused with, and the three-part test for checking a claim.
August 4, 2026
Two Kinds of AI Orchestration
A person orchestrating tools, and an agent orchestrating agents. The line is whether the routing decision is still yours, not sophistication.
July 16, 2026
Netlify Rebuilt Its Platform Around AI Agents
Netlify shipped a major update built around AI agents: Functions redesigned for Agent Experience, an open benchmark called AXIS, and Claude Opus 4.8 live in its AI Gateway. Here's what it means if you're building with Claude.
July 10, 2026
Why Micromanaging Claude Holds You Back
Most people fail with Claude because they treat it like a vending machine for tiny tasks instead of giving it a real goal and letting it work.
July 10, 2026
The Most Common Mistakes People Make With Claude
The same mistake keeps showing up in room after room: something pretty that solves today's problem and nothing built to hold up past it.
July 10, 2026
How to Manage AI Agents Like an Executive
The real unlock in agent orchestration isn't directing every agent's every move, it's delegating the orchestration itself and reviewing outcomes like an executive.
July 3, 2026
Claude Team vs Enterprise, for a 15-person team
Most teams under 20 to 30 people don't need Enterprise. The real difference isn't the price, it's what happens to your data and who controls it. Here's how to actually decide.