Every day there are dozens of AI headlines, and almost all of them promise that this time everything changes. If you try to follow them all, you end up exhausted and on Monday you still don’t know what to do. A filter is all you need.
Six questions before you take a headline seriously
- Who says so? The company always says its thing is the best. Wait for independent tests or for people who actually use it.
- Can I use it today, in my country? Announced doesn’t mean you can use it yet. And available in the US doesn’t mean available in Europe.
- Is it a ranking? They help you get your bearings, but the top spot changes every few weeks and companies train to score well on them.
- Is it a demo? Launch videos are staged. The test that matters is the one you run with your own work.
- Does it announce the end of something? «The end of programmers», «the end of Google». Almost always exaggerated, and real change is slower.
- Does it change anything for me? The most important one. If the answer is no, you can safely ignore it.
What matters and what you can let go
Worth your time:
- A tool you use gets cheaper or changes its limits.
- Something new you can apply to your work this week.
- A tool you depend on is shutting down.
- A legal change that affects how you use AI at work.
You can let it go without guilt:
- Funding rounds and valuations.
- Who is first in this week’s ranking.
- Social media fights between executives.
- Predictions about when superintelligence arrives.
Where to follow AI news
When a piece of news really matters to you, go to the source. To compare models, Artificial Analysis: it is independent and measures them all with the same tests, plus price and speed. It is the one I use for the ranking in the group, and every month I sum it up in the best AI right now.
For launches, the news pages of OpenAI, Anthropic and Google DeepMind. There you see what has really come out, for whom and since when, without the headline gloss.
If you read Spanish, these are the ones I would follow. You don’t need all of them: pick one to read and one to listen to. All of them published something new in the week of 24 September 2026:
- IA en Español, a newsletter on Mondays, Wednesdays and Fridays, with the important news and a tool to try that day.
- Digital Brain, a short email every weekday morning with the news and a tutorial or prompt for work.
- monos estocásticos, a weekly podcast by Antonio Ortiz and Matías S. Zavia, to understand what is behind the news.
- La Tertul-IA, a weekly talk show about AI in business.
- Inteligencia Artificial, by Pocho Costa, a podcast for small business owners with real cases. He’s in Argentina, so some examples are from there.
- Dot CSV Lab, Carlos Santana’s YouTube channel, to know whether the model that just came out changes anything.
- Xavier Mitjana, YouTube tutorials on building agents and automations without coding.
- Xataka, its artificial intelligence section, with several pieces a day, including what affects Spain.
A real example
On 22 September 2026, Claude Opus 5.5 and GPT-6 Sol and Luna came out on the same day. There were headlines everywhere. What actually mattered to someone using AI at work fitted in three lines: Opus 5.5 became the best model and was cheaper per token than the previous one, Luna made automated paperwork almost free, and both companies gave away a reset to refill your usage limit. The rest was coffee-break talk.
How I do it
Every day I go through the AI news and keep what’s useful for you. With that I put together the daily message for the La IA que sirve WhatsApp group: something you can apply at work that day and the news that affects you. Always in the same order, with a practical line for your work in each point and a five-line ranking at the end that only changes when there is a reason. If nothing important happens, I say so and that’s it. The message is in Spanish.
On the home page you can see a real message, exactly as I sent it to the group.
