Stop using AI like Google
You ask questions and expect answers. That's why you never get what you actually want.
6 min read
May 2026
Fundamentals
Read the article →
Latest issue · Week 19 · May MMXXVI
Five model releases this week.
Five model releases this week.
Only one mattered.
This week I tested all four major ones: Claude, ChatGPT, Gemini and Grok — with the same prompt, in English, on a concrete business problem. The results differed more than the price tags suggest.
The interesting thing is not which API computes fastest. It's which one actually understands the context. Idioms, industry terms, decision logic. That model wins your hours — regardless of what it costs per month.
In this issue: a comparison that actually helps you make decisions, three concrete test prompts you can copy directly, and a clear answer to whether you should pay 170 kr/month or 349 kr/month.
Read the full analysis →Knowledge without paywalls.
Every guide is written for individuals, business owners and anyone who wants to learn — concrete, direct and without academic overviews.
-
Beginner · PDF · 28 pagesAI for everyone — get started in 3 days
The four tools that actually make a difference in your daily life, plus a framework for getting started without unnecessary complexity.
-
Intermediate · PDF · 42 pagesFrom tools to systems — your first AI pipeline
Once you've moved past the prompt stage, you need chains that run without you. Here's the framework — without framework-bingo, without hidden costs.
-
Advanced · PDF · 64 pagesBuild your first AI agent locally
Your own hardware, open models, full control. Step by step from an empty machine to an agent that actually does work for you overnight.
Coming soon
Three analyses. One method.
Long-form when needed, short when it's enough. Everything passes through the same filter: is it actually relevant for an entrepreneur who needs to make a decision next Monday?
Claude vs ChatGPT vs Gemini vs Grok — which one should you pay for?
Four models, four pricing plans, four different philosophies. The difference between the right and wrong choice is often hundreds per month — and hours lost when the model simply doesn't understand what you're trying to do.
12 min read
May 2026
All levels
Read the analysis →
Local models — when are they worth the trouble?
A 32 GB GPU costs about the same as three years of Claude subscription. Is it worth it? It depends on what you need to do, how much you run, and how much of your data can't leave the building.
8 min read
May 2026
Advanced
Read the deep dive →
Method
A method that holds up to scrutiny.
Writing about AI is easy. Writing about AI without sounding like a press-release generator is harder. Polaris follows a strict five-step method — every time.
-
01
Source
TLDR AI, Stratechery, papers from arXiv, official blogs. Primary sources always — no summaries of summaries.
-
02
Verify
Test myself with real data. No hearsay, no demo videos. If I can't reproduce it, I don't publish it.
-
03
Filter
Only what matters for Nordic SMEs. The rest is cut — no matter how impressive it is in Silicon Valley.
-
04
Translate
In plain English, with practical examples, for decision-makers. Technical accuracy without technical jargon.
-
05
Publish
Free. No paywall. Without asking anything of you except the time it takes to read it.
Live · updated every hour
The AI field right now.
A readable terminal view of what's actually moving in the AI world this week. No stock prices, no influencer quotes — just hard data.
Latest Claude
Opus 4.7
Latest GPT
GPT-5.2
arXiv today
114 papers
Anthropic API
142 ms
OpenAI API
218 ms
Token average
$2.40
Summarized for you · Week 19
3
Major model releases
−18 %
Token price vs Q1
2
Worth testing yourself
Established information sources.
Polaris doesn't work in a vacuum. These are the five feeds followed every week that form the foundation for everything published here.
-
01
TLDR AI
Daily summary of what's actually being released in the AI world — no clickbait headlines.
-
02
Stratechery
Ben Thompson's strategic analysis of the tech companies driving the AI field forward.
-
03
Latent Space
Practical AI engineering, podcast and newsletter. What works in production.
-
04
Anthropic Blog
Official releases, model cards and safety research straight from the source.
-
05
OpenAI Research
Technical release notes and research papers relating to the next GPT iteration.
DM
Founder & editor
Daniel Merthen
Twenty years at the intersection of technology and business. Not in an office — out in the field. Founder of Hrafninn Industries AB. Building AI infrastructure by day, gaming computers by night. Polaris is my open knowledge base — what I wish someone had handed me when the AI shift began.
Next issue · Sunday
Polaris in your inbox —
Polaris in your inbox —
every Sunday, always free.
The most important AI, for everyone who wants to master it. No noise. Just what actually matters.
Already 312 subscribers · 0 issues missed