Open Textbook

Automation

5 Layers of an AI Agent: What Every Manager Needs to Know

19 min read

“Most people still think an AI agent is just ChatGPT with a good prompt.” That is the opening line of Sunil Ramlochan’s article “The AI Agent Stack Is Not a Prompt. It’s a Production System”, and the author calls this belief “a comforting myth.” The useful truth, he argues, is different: a real agent is closer to a small operating system for getting work done. It has a brain, hands, memory, rules, logs, recovery plans, and someone accountable when the agent does the wrong thing.

The article’s thesis fits in a single line: an agent is an entire stack. Reliability comes from the architecture around it, while the model itself – or a clever prompt – is just one ingredient. The picture is an engineering one, so let us approach it from the other side: what in this stack actually concerns the manager who does not write code but decides whether to put an agent to work.

Read more
5 Layers of an AI Agent: What Every Manager Needs to Know
P5.express and Agentic AI: Where It Helps, Where It Breaks Things
24 min

P5.express and Agentic AI: Where It Helps, Where It Breaks Things

In the PMI world, portfolio management is typically imagined as something monumental: steering committees, Tableau dashboards, hundreds of Jira fields, weekly status meetings with a deck of slides. P5.express offers a different approach. Three cycles, five documents, two roles. The entire system fits on a single page.

This is exactly the kind of system where agentic AI makes sense: minimalist architecture that’s easy to understand, clearly defined roles, structured data. But “makes sense” doesn’t mean “everywhere.” Some parts of P5.express stop working when automated – not because the AI is bad, but because those parts derive their value from the human process itself.

Below is a cycle-by-cycle breakdown. What’s worth delegating to an agent, what’s better left to people, and which model fits these tasks best.

The Agent Instead of Chat: Data Analysis Without Copy-Paste
11 min

The Agent Instead of Chat: Data Analysis Without Copy-Paste

You have three data files: an activation funnel, A/B test results, and support tickets. The task – figure out why onboarding is underperforming. You open ChatGPT, upload the first file, ask your question. You get an answer. You upload the second file. ChatGPT asks: “Can you remind me of the context?” You upload the third. The context of the first file has already been pushed out.

Forty minutes later you have three separate conversations, none of which answer the original question. Because the question was one, and the data was in three places.

This isn’t a ChatGPT problem. It’s a problem of approach.

Claude Code Costs $100/Month – OpenCode Does the Same Thing for Free
13 min

Claude Code Costs $100/Month – OpenCode Does the Same Thing for Free

In March 2026, Lenny Rachitsky published an article with a telling headline: “Everyone should be using Claude Code”. It went viral on LinkedIn and tech newsletters, picked up hundreds of thousands of views, and now every week a manager somewhere asks: how do I try this?

The answer is uncomfortable. The Anthropic Max subscription that unlocks Claude Code costs $100 per month. That’s $1,200 a year for a single tool – before you’ve even figured out whether it fits your workflow. And it locks you into a single vendor, a single model, and a pricing tier designed for power users, not for someone exploring whether AI agents are worth the investment.

There’s a direct alternative. OpenCode is an open-source project that does exactly the same thing, works with any model (including free ones), and takes 15 minutes to set up.

ClawCon, China, and $100B: OpenClaw Between Ecosystem and Bubble
12 min

ClawCon, China, and $100B: OpenClaw Between Ecosystem and Bubble

320,000 stars on GitHub. A physical conference in Manhattan. Government subsidies in Shenzhen. In six weeks, OpenClaw has gone from a repository with an interesting README to a geopolitical phenomenon discussed by Bernstein analysts, officials at China’s Ministry of Industry, and users on Xianyu posting consulting ads with the headline “will help you uninstall.”

This is the sixth part of the series. In the previous part, we recorded 265,000 stars, the creator joining OpenAI, the first enterprise deployments via Sundae Bar, and three new CVEs. That was 12 days ago.

OpenClaw in Practice: Real Use Cases and the Missing Enterprise Layer
15 min

OpenClaw in Practice: Real Use Cases and the Missing Enterprise Layer

After three articles covering critical security issues, workflow lessons, and 72 hours of patches, the obvious question is: what are people actually doing with OpenClaw?

In the two weeks since its explosive growth (January 22 – February 5, 2026), a substantial body of confirmed use cases has emerged from Reddit, X/Twitter, YouTube tutorials, and developer blogs. Interestingly, the usage pattern reveals not so much revolutionary scenarios as a dramatic drop in the barrier to entry for automation that already existed.

Surprisingly, most of these use cases have been technically achievable through n8n, Make, or Zapier for the past 3–5 years. The difference isn’t in capability – it’s in who can now build it. Which raises the question: is OpenClaw truly a new category of tool, or just a more accessible wrapper around old concepts?

86% of Students Use AI, But Are Getting Worse. One Experiment Changed Everything
16 min

86% of Students Use AI, But Are Getting Worse. One Experiment Changed Everything

Traditional approaches to education are breaking down. AI writes essays and papers in minutes – and that has permanently changed the purpose of creative assignments in schools. Banning neural networks doesn’t work, and isolating students from technology is a dead end. The question is not whether to use AI. The question is how to use it so the technology develops students’ skills rather than replacing their thinking.