You should own your content, and you should be free to move it to whatever platform you want. Sadly, that is not always the case. This film from @codeforthepeople.com is a reminder of why people stand for open source and their values. Worth the 20' watch, trust me. www.youtube.com/watch?v=8lQi...

3+1 books worth reading to think better about numbers and biases: 1️⃣ Factfulness, by Hans Rosling, Ola Rosling, and Anna Rosling Rönnlund: Instincts and biases tricks us, this will help you reading the world (with some optimism).

Your agent can use MCPs to control apps and websites, yet sometimes you don't even know what they're doing. It shouldn't be like this. Security should always come first. @wpvip.com just released the Secure MCP, and I hope others will follow this lead, honestly, because it's just that useful.

I didn't expect to write this, but here we are: Slop is actually good for you.

(effective) Agentic Loops are the Deep Work state for LLM. What is your human loop, then?

I always trusted numbers, and while I was often skeptical of stats, only recently I realized how easy it is to game them. The Will Rogers phenomenon changed the way I see the world.

There's a new discussion paper based on 26.000 Chinese students comparing those using AI vs those not using it. The result: AI kids were faster in the short term, got worse after 2 years. No surprise. We're still adapting. We can't just replace our skills with AI and be done with it.

Either you're locked into a system, or you're free to change your mind, with no chains. Having something that stands for that freedom is truly priceless in today's world, where most of the products try to lock you in. Be brave (and read this beautiful post from @zeldman.bsky.social ).

Imagine a self evolving SAAS software where you can pay the tokens to build a feature you needed, in addition to a monthly fee. If others will use it then I the fee you pay will be detracted by your monthly costs. If you’ll be the only one using it, that’s it. Would you use it?

Multiple Dimensions of AI Cerebras just announced Gemma 4 31B running at over 1,800 tokens per second.

This resonates hard with me. It's not happening every day, but it is happening. It's hard to say if it's caused by willpower depletion or just exhaustion. Not so unexpectedly, @elijahpotter.dev added a very smart idea at the end of the blog post. Worth a read. elijahpotter.dev/articles/slo...

OpenAI just announced its first chip, and it’s no surprise. AI is not cheap. It takes quite a lot of money and electricity to run, and there’s a reason companies are starting to build chips.

To understand we need to let go of our ego and stop trying to ask smart questions. Simple, basic questions are incredibly effective to get deeper into any problem and get a superior understanding. Letting the ego go is the hard part.

ChatGPT is offsetting Pulse in favor of Scheduled actions and this is a strong reminder of what AI truly unlocked for product development: The acceptance of experimental features.

AI will happily spend tokens all day long. It won’t question assumptions. It won’t ask whether the idea is worth pursuing. Which brings us back to the same product question we’ve always had: What’s the value of what you want to build?

The Midjourney pivot to medical stuff is mindblowing. I'm really curious to see how that will turn out.

“60% say AI in a brand’s messaging is a turnoff, not a feature.” That’s one of the findings from the @wpvip.com article "Future of the Web 2026: AI Brand Visibility Research." And… I kind of agree. I mean, who really wants to talk to an AI unintentionally? (continues)

This WhatsApp scam had my real booking reference and hotel dates. That’s exactly why it was dangerous and worth sharing so you know how phishing is evolving. Enjoy!

DataBricks released a new meta-harness that's model agnostic and I feel many players in this space are all trying to build similar solutions to similar problems only to see which one will stick.

I've been thinking about what it takes to keep a project opinionated as it grows. It's easy to take a stance early on, but passing that vision down with the same intensity over time is harder than it seems. My take below.

How many times have you thought, “I’d really love to do X (vacation, new hobby, etc) but I have do Y (insert random excuse) so it's not the right time”? Guess what? The time was actually right. I get more into why below

We're treating AI like something that "knows it all", but a recent paper showed how easy it is to bias it, even with one single element in the training data. From a human perspective this is a reminder that you should always question the output (as you would with an actual human).

Anthropic’s ban on AI tools during job interviews is a reminder of what has always been special about humans.

Should you speak or stay silent? There's a new open-source audio model that every 4 seconds,decides whether to keep waiting for more input or interact. Be silent, or speak.

This prompt from one of Anthropic employee is becoming more popular and I can tell you why. It focuses on the one thing we need to structure from scratch: Understanding. AI Is fantastic at producing output but that's often a long form output we're not that used to anymore.

This is a WordPress Admin. Look again if you don't believe me. I can't stop thinking about this project that brings a new look to what the WordPress Admin "means", with personalization and multi-tasking at its core. How is this admin look possible? It's the WordPress desktop-mode plugin.

Two Bets on the Future of Software Engineering I feel that in the future all code will be written by agents. That might not be the time now for every company, but I think that's where we're heading. The interesting part is not whether that happens or not. The interesting part is how we get there.

How can you protect by prompt injections when you're also prompting? There's this recent ChatGPT for Google Sheets attack that, thanks to a hidden prompt, it exfiltrates your data and executes a custom AppScript. It got me thinking "We need an anti-virus but for prompt injections".

Willpower is finite, decision fatigue is real and it's a security risk. Today I stumbled into a neat website simulating decision fatigue with AI (multiple Yes/No questions).

It's kind of surprising to me that the only big competitor to China open weight models is, currently, Google. Gemma4 is competing strongly with the latest Qwen open models.

Recently the CEO of ClickUp announced a layoff of 22%, and that was also bundled with a lot of ideas for the future and what it means from now on. Below my 5 takeaways 👇

Here are two books I just finished reading and my take on it (spoiler, I didn't like one of the two).

Here's what AI allows today that wasn't possible "just" with static code analyzers when thinking about code quality and security. It lowers the entry barrier. Getting the code analysis right can be hard on some projects, it can take quite a while and still you might not be able to catch them all.

We're at a point where a generic model from OpenAI disproves a popular (for lack of a better word) conjecture in discrete geometry.

Want to watch a video in a noisy environment but you miss some subtitles? Accessibility. Can't read a site under strong sunlight because of low contrast? Accessibility. Today is Global Accessibility Awareness Day and I took the opportunity to share why Accessibility benefits everyone (even you).

This is a story about one company experiment that reminded me of how important it is to challenge the complexity. Not to remove it, but to question it and use that as a driver for learning.

I was just checking out Antigravity CLI and it's interesting that, to authenticate, you need to "Authenticate with Antigravity or Antigravity IDE before using the CLI." Additional friction to try a product is always risky. We'll see how this plays out.

GitHub was compromised via a poisoned VS Code extension. Now more than ever is important to build awareness about of supply chain attacks. Here are my thoughts and 5 questions that I always use.

Night reader? We just finished bringing dark mode in the WordPress.com Reader _and_ across WordPress.com yet-to-be announced dashboard 🌙 Now you can manage your sites or catch up on your RSS/blog subscriptions long after the sun goes down.

Now more than ever, it's time to experiment. To try stuff that feels nonsense. Not to build TheNextBigThing but to understand the new raw materials we've been given. Go, try, fail and enjoy the process.

Here’s a new type of AI attack you might be susceptible to: sleeper memory poisoning. We know that context is one of the most important parts of AI, but context is limited. Because of that, we rely on external systems to store information that should only be retrieved when needed.

Bun just rewrote its codebase from Zig to … Bun just rewrote its codebase from Zig to Rust. One PR. 1 Million code lines changed. Besides the obvious worry from people using it, this is a reminder of what's possible now, in this age. Such a gigantic migration would've taken months, at least, with…

Race to the bottom or race to the top? OpenAI is giving companies 2 months of free Codex usage. Anthropic is extending limits until July 13th. This is clearly a race to get as many users as possible to switch ecosystems, and I think it’ll continue until the market stabilizes a bit more.

Google just reported the first AI-assisted zero-day exploitation. And honestly, I think this is one of those moments where the security paradigm starts shifting in real time.

Sharing something very personal about the way illness, loss and uncertainty reshape what we think really matters in life. If you’ve ever felt life only starts to make sense in hindsight, this is for you. givemethechills.com/2026/05/12/w...

If you've ever wished WordPress.com had dark mode, I have good news. We've been building one into a new yet-to-be-announced dashboard at Automattic. If you spend a lot of time in WordPress, especially late at night, I'd love your feedback. Steps to try it 🧵

DON'T TRUST THE DEMOS I recently realized is that some things feel like they help you understand a product but they actually don't help reinforcing the mental map, and one of them can be demos. I wrote about it and how to improve product understanding in my new article.