AI in Practice

Method

The method behind AI in Practice

Every week thousands of AI announcements compete for attention. We keep only what changes practical work for people who build their own AI business brain, check it against the source, let three AIs score it, and say plainly where it leads. This page explains how, so you can judge the results.

Where we are heading

For a thousand years the wealthy managed resources: land, capital and people working for them. Soon every business owner will manage an AI organization. Agents do most of the work, the AI takes over the routine, and the owner keeps the thinking and the decisions.

It takes means, but far less than a property portfolio. What it takes most is structure: rules, memory, skills, tools and focus that the business owns. That is what Arete AIOS builds: the business brain lives on the company's own computer and server, a local AI handles private data, external models do the deep work, and control stays with the owner.

One week, five steps

  1. 1

    Scout

    Thursday 07:30

    Our growth agent, running on our own server, reads official changelogs, Hacker News, X and practitioner blogs, and proposes 8 to 15 candidates with their primary source.

  2. 2

    Score

    Thursday, before 10:00

    Three AIs, Claude, ChatGPT and Jev, score every candidate from the same text with the same rubric. Where they disagree by more than 15 points, the item is marked for review.

  3. 3

    Pick

    Thursday 10:00

    Egils picks the top three and the rest, adds first-hand finds and decides every review case.

  4. 4

    Check and write

    Thursday

    Every claim is checked against its primary source and dated. Where a launch claim and reality differ, we say so. Prompts are written for your business brain, not for a demo.

  5. 5

    Publish

    Friday

    The issue, a screen-share video, a Markdown copy for your own AI, and an updated AI Monitor.

Two questions for every launch

Practical score, 50 to 100

How much does it change practical work for someone building an AI business brain this year? 90 and up changes how almost everyone works with AI, 75 and up is a major new capability, 55 and up is notable. Below 50 stays out.

Direction

Does it move AI toward business owners or away from them? Does it help a business keep and move its own data, memory and skills, or keep them inside one platform? Does it move work toward the owner's own computers or deeper into a vendor's cloud? The AI Monitor tracks this over time.

Why three AIs, and why Jev

Claude and ChatGPT read the launch and give a score with one sentence of reasoning. Jev, a decision model from TypeSafe AI, does not write at all: it picks from fixed options and says how sure it is. The median of the three becomes the score; Egils can replace it.

On our first test, ten claims from Week 42 checked against their source excerpts, Jev and ChatGPT were right ten times out of ten and Claude eight. Jev was least sure exactly on the three hardest claims, which is where a person should read. Jev knows only the text we give it, so it checks claims against a source; it does not replace finding the source.

See the scores in the AI Monitor

What we do not do

  • No paid placements. Vendors cannot buy a score.
  • No private data goes to the raters: they see the same public text you do.
  • A score is a judgement, not a benchmark. When a test is ours, we say how small it was.
  • A launch claim is not a result. Plans, prices and availability are dated, and we say when something is not available in the EU.

We publish the method because you should be able to judge it. You can build the same for your own business; if you want help building it, that is what Arete AIOS does.

Questions or a launch we missed: egils@areteaios.com.