Editorial standards

Who writes it

Forge is a small team building a platform that teaches professionals to work with AI. The blog is written by that team and published under the byline Forge Team rather than individual names — posts are drafted, reviewed, and edited collectively, so a single author credit would misrepresent how they're made.

We have a commercial interest in you finding this useful: the posts link to drills on our platform, and the platform is what we sell. That shapes what we write about. It does not change what we say about it, and where the honest answer is “don't use AI for this” we write that instead.

How the examples work

Many posts include short accounts of named professionals — Priya triaging support tickets, Ana auditing her week, Ondrej checking record counts.

These are illustrative composites, not reported interviews. The names are invented and the specific figures are chosen to make a pattern concrete. The situations are drawn from documented, recurring problems in how people adopt AI at work — the failure modes are real and common, the individuals are not.

We use them because an abstract failure mode is forgettable and a specific one isn't. We say so here because presenting an invented person as a source would be a lie, and because you should know which claims in a post are evidence and which are illustration.

Where a post cites research, a product change, an incident, or a ruling, that is a factual claim and we stand behind it as one.

How AI is used to produce it

We use AI in drafting, the same way we teach people to. It helps with structure, first drafts, and finding the weak paragraph. Every post is read end to end by a person before it publishes, and the factual claims are checked against sources rather than against the model.

It would be strange to teach AI supervision while pretending not to use the tools, and stranger to publish unverified output while writing about verifying output.

What we publish, and in what mix

There are two kinds of post here. Signals react to something that has just happened and are written to be useful in the week they appear. Explainers, guides and role guides are written to still be right in a year.

The archive currently runs to 152 posts, 64 of them durable rather than news-shaped. We hold the durable share above 40% in every topic deliberately: a blog that is only news reaction becomes worthless the moment the news moves.

• Directing AI — Briefing, framing, and context — getting what you actually asked for.
• Judging AI — Verification, quality, and the gap between confident and correct.
• Supervising Agents — Permissions, guardrails, and deciding what a tool may do without asking.
• Systems & Adoption — Workflows, cost, governance, and getting a team past the pilot.
• Work & Judgement — What AI does to careers, and what it quietly does to your own judgement.

Corrections

If something here is wrong, we want to know and we will fix it. Email hello@forge-learn.app with the post and what's wrong.

Substantive corrections — anything that changes what a post is telling you to do — are made in the post itself and the updated date changes with them. Typos and broken links get fixed silently, because nobody needs a changelog for a missing comma.

What we don't do

• Sponsored posts. Nothing here is paid for by a vendor, and no post has been written or approved by one.
• Affiliate links. When we name a tool, we make nothing from you using it.
• Invented statistics. If a post gives you a number without a source, treat that as a bug and tell us.
• Advice that needs a professional. Posts touching legal, financial, or regulatory ground say so and tell you to get someone qualified. They are not a substitute for that person.

Getting the posts

New posts go out by email and by RSS. The email list is used for new posts and product announcements, nothing else, and unsubscribing takes one click.