TechGuild
1Level 1, Curious Newcomer

About

What this material is, and how far to trust it

A study guide that never says where its confidence runs out is asking to be trusted uniformly, and no guide has earned that. This page is the general version of an argument each course makes in full.

3 courses · written here, from published blueprints and public documentation

lessons
114
open without an account
11
exam questions reproduced
0

What TechGuild is

TechGuild publishes independent study material on how AI systems are built, evaluated, secured and governed, together with preparation guides for the certifications that cover the same ground. Everything under the library was written here. It is not courseware bought in, relabelled or paraphrased from somebody else's guide, and it reproduces no exam questions.

It is not affiliated with, endorsed by or sponsored by Anthropic, Amazon Web Services or any certification body, and none of them has reviewed a word of it. Certification names appear in plain text because they are the name of the exam you are sitting; they are trademarks of their owners and no badge or logo is used here to imply anything more.

There is nothing to pay. There is no paid plan, no tier and no upgrade, nothing here is sold, and nothing on this site asks you for payment details. What it does ask for is an account: the orientation module of each course reads with no account at all, which is 11 of the 114 lessons here, and a free account opens the other 103 along with the module quizzes and the practice papers. Creating one costs nothing and needs an email address you can receive mail at.

Where the facts come from

Two kinds of claim run through every certification guide here, and they have different provenance, so they are sourced differently.

The exam facts come from the official exam guide
Published by the certification programme itself: the exam code, the number of items, the question formats, the time limit, the pass mark and the scale it is reported on, the fee, the validity period and the domain weights. None of that is inferred, and all of it can change in a later revision, which is why each course tells you to read the revision history yourself before you book.
The technical material was checked against the vendor's own documentation
Product names, tool names, configuration fields, flags, file locations, component names and status notices were verified there rather than written from memory or carried across from older courseware. A prep site, a video course, a forum thread or another study guide is a secondary source however confident its tone, and the failure mode is specific: material that is coherent, well written and eighteen months out of date.

The check that was run, and what it was not

Each course was put through an adversarial fact-check against those primary sources. Claims were pulled out and checked one at a time, on the working assumption that each was wrong until a source said otherwise. That pass produced findings, and the corrections are in the lessons as published.

One episode is worth reporting because it is the sort of thing a marketing page normally hides. The Claude Architect course was written before the appendix listing the blueprint's out-of-scope topics was recovered, so the whole build proceeded without it. Once it surfaced, every lesson, practice item and card was audited against that list. The finding was small: the course had crossed the boundary in a handful of places, all of them prompt caching mechanics mentioned in passing, and that detail was trimmed. No answer key changed, and nothing in scope was removed to make room.

What these guides refuse to claim

Several numbers a reader reasonably wants are missing on purpose. Each absence is a decision rather than an oversight, and each course states its own list.

No raw percentage needed to pass

A scaled pass mark exists so that a score means the same thing on every form of the exam, which requires equating: a harder form needs slightly fewer correct answers to reach the mark. There is no single raw percentage to publish. Where an exam also mixes in unscored trial items you cannot identify, even your own count of correct answers is not the number the scale is built from. Anyone stating that figure is guessing at something neither they nor you can check.

No per-domain item counts

Blueprints publish weights, not counts. Where a course does the arithmetic anyway, for study planning or for reading a score report, it says out loud that it is arithmetic on a published weight.

No real exam questions

Every one of the 1,476 practice exam questions and 448 quiz questions here was written for this site from the published objectives. That is partly integrity, since reconstructing live items from candidate recall breaches the agreement you accept before you sit, and partly practical: a bank assembled from other people's memories of a rotating pool trains you to recognise phrasings that will not be in front of you.

No numbers that could not be read from a source

Where a vendor's training catalogue could not be checked reliably, the course describes it without stating prices, question counts or which parts are free, and tells you to open the vendor's page on the day you plan.

What is version-sensitive

The products these guides describe ship faster than any study guide is revised. Rather than leave that as a general disclaimer, each course carries a lesson that is nothing but a ledger: a table of every topic in that course known to move, why it moves, and exactly what to verify on your own installed version. Tool renames, permission modes, hook output fields, command flags, frontmatter fields, product names and service statuses are all on it.

Those lessons sit in the orientation module of each course, so they need no account, and they are linked below. They are the least interesting pages in the library and the ones most worth reading before you lean on the rest of it.

Where the authority actually lives

The official exam guide is authoritative for anything about an exam: the fee, the format, the time limit, the pass mark, the weights. The vendor's own documentation is authoritative for anything about the products. These guides organise both into something you can reason with at the end of a working day, and they are worth what a good study aid is worth, which is a lot until the moment you need the final word on a fee, a flag or a field name.

When your installed build and the documentation disagree, learn the documented behaviour, because that is what an exam item is written against, and note your build's behaviour separately so your own systems keep working.

Nothing runs here

There is no hosted sandbox, no browser IDE and no pre-warmed container. The lessons describe the tools as you will meet them on your own machine, with your own installed version and your own configuration, and the version ledger exists precisely because that is where you will be working. It costs you a first evening of setup that a sandbox would have skipped, which is the evening the learning is in.

Corrections

The ledger lessons are an invitation as much as a disclosure. If a claim here is wrong, out of date, or right in a way that reads as wrong, the correction is welcome and the source you checked it against is the useful part. Get in touch.