Help & quick start

Quick start

1. Search the whole library. The box on the home page searches the full text of every article, not just titles. It copes with several words at once, with partial words, and with typing it wrong — hippa finds HIPAA. Try agentic memory or hipaa.
2. Ask a question instead. The Ask a question helper sits in the corner of every page, on a phone as well as a desktop. It answers by quoting real sentences from the articles and linking the source — so it can point you at the passage rather than the piece. Ask it what is context engineering? and see.
3. Or browse by subject. Topics is six doors into the library — security & compliance, career & leadership, agentic AI, tooling, cloud, healthcare — each with the flagships first and the full archive behind it.

That is the whole of it. Everything below is detail you only need when you want it.

Tutorials

Find everything on one subject

1. Open Research. Every article in the library is there as a card.
2. Use the filter buttons along the top. They are subjects, not folders — an article about agentic security is filed under Agentic AI and Security & Compliance, so it appears under both.
3. The filtered view is a real address you can keep or send to someone: research.html#seccomp opens the archive already filtered to security and compliance. #car is career and leadership, #agt agentic AI.

The small pills at the end of an article are the same subjects. Clicking one copies its hashtag, which is there for sharing rather than for navigating — use the filters to move around.

Ask the assistant, and make it better

1. Open Ask a question and type a real question. Whole sentences work better than keywords, because the assistant is matching against written prose.
2. Ask for it in plain words if you want it plainer. Adding in plain english, in simple terms, explain simply or eli5 changes which sentences it picks: it leans to the shortest, least jargon-heavy ones, and for many topics it can use a plain-language summary written specifically for a reader who does not have the vocabulary.
3. Use More detail under an answer for the supporting quotes, and Read more for the articles themselves.
4. Rate it. 👍 or 👎 on an answer is the single most useful thing you can do here. A 👍 teaches the site that a particular wording was answered well by a particular passage, and it leads with that passage next time somebody asks the same thing. A 👎 flags the answer for rewriting. It is anonymous — see how the assistant works for exactly what is stored.
5. If it finds nothing good it says so, and offers Suggest this topic. That goes onto the writing queue. It is the most direct way to ask for something that has not been written yet.

Read a series or a set end to end

Some writing here is meant to be read in order, and some is a group of pieces that each stand alone. The site tells you which:

• A series assumes the parts before it. The Agentic AI Builder’s Playbook is the long one; each part ends with a link to the next, so you can start at part one and keep going.
• A set does not assume anything — each piece is complete on its own, and read together they make one argument. The Money Language Series is one. Every member names its own number and links the others, under the opening paragraph and again at the foot.

Get around a long article

• A long, sectioned piece opens with a contents block. It is a real list of links; on a phone you can fold it away.
• Technical pieces often end with a glossary, set as a term-and-meaning table.
• Some headline articles hang a dense technical detail page off the bottom — the decision record or the appendix behind the argument, for readers who want the workings.
• Citations in the source are live links, including the ones whose address was hidden behind the words.

Keep up with new writing

1. By email. The box at the foot of any page. You will get one message asking you to confirm — nothing is sent until you click it, and if you never do, the record is deleted after two weeks. Issues go out on Tuesdays when there is something new; a quiet week sends nothing.
2. By feed reader. feed.xml is a normal RSS feed of every listed article, newest first. Every page advertises it, so pasting any address from this site into a reader will find it.
3. On LinkedIn. The links at the foot of the page go to the company page and to Jason’s profile. That is a separate thing from the email list — unsubscribing from one does not touch the other.

Unsubscribing deletes your address rather than flagging it, and the link is in the footer of every issue. You never have to type your address to get out.

Answers to the usual questions

Why does the assistant quote instead of writing an answer?

Because a quote can be checked and an invention cannot. It selects and arranges sentences that are genuinely in the articles, so there is always a source to click through to, and there is nothing for it to make up. When it has no good match it says so rather than filling the space.

Is anything about me being collected?

No name, no email, no account, no IP address, no cookie, no advertising identifier, and nothing shared with anyone. What is kept is what the site is being used for — the questions asked and whether they were answered well — because that is what shows where the library is thin. The assistant page lists every field stored, and that list is checked against the code that writes it.

Does the assistant send my question to an AI service?

No. The matching happens in your browser, against files this site already serves. No third party is involved at any point.

Search found nothing. Now what?

Try fewer words, or the subject rather than the phrasing — the index covers the body of every article, so a distinctive word from the middle of a piece works well. Failing that, ask the assistant, which understands paraphrases, and use Suggest this topic if the subject genuinely is not here.

Can I share or quote this writing?

Yes — link to it freely. Each article carries a ready-made post and its hashtags for exactly that. Please link to the article rather than copying it wholesale.

Something is broken, or I want to get in touch.

Contact has the address. Telling the assistant it got something wrong with a 👎 is also genuinely read.

Where do I start if I am new to agentic AI?

Learning is the curated route in — videos, repositories, guides and books from outside this site — and the Playbook is the route in from inside it.