From his corner at The Constant, the wizard scrolls headlines while sipping his first coffee of the day. Trump's vibe presidenting, Sam Altman has insomnia over ChatGPT memory upgrades, MCP is having a viral moment (minus revolutionary use cases), and everyone is turning AI into their therapist.
Howdy wizards,
It’s been another eventful week in the world of math that’s so good that it feels human. Sit back in your favourite piece of furniture and pour yourself a cup of enchanted bean juice.
Here’s what’s brewing in AI.
PS a quick sponsored reminder about Wordware's launch earlier this week—their new Triggers & Actions feature connects AI agents to 2,000+ apps with free monthly credits (courtesy of their $30M seed round). Try Wordware →
DARIO’S PICKS
“Like my wife, ChatGPT now remembers something I said 1,000 days ago” — @piet_dev
ChatGPT can now seamlessly reference all of your past conversations, not just the limited memory it had before.
Previously, you had to get real specific like “remember that I’m gluten intolerant” and then pray it would actually recall that info when you asked "what should I eat?"
Enter ChatGPT’s new and improved memory:
It now listens all the time and remembers over a long period of time and across conversations (aka all your chats).
The responses you’ll get are likely to get more relevant and useful as new conversations build upon what you’ve already talked about.
You can, to some extent, control what ChatGPT remembers by simply asking it.
For the privacy-conscious people out there concerned about the personal dossier ChatGPT would be building on you (I get you!), it’s also possible opt out of this whole memory-thing entirely in your settings.
It’s also possible to use the mode “temporary chat” which is ChatGPT’s equivalent to your browser’s incognito mode.
Why it matters AI is quickly evolving from a useful tool to a thought partner and companion. How useful this long-term memory actually is still depends on several things, though, like how fast and seamless and how context-aware it is.
I'll personally be testing where this adds value versus where it gets weird (will it start mixing work and personal chats, etc.). I’m thinking that I can always reset or make a new account if things get creepy.
One thing Sam Altman is probably acutely aware of though is how ChatGPT’s new persistent memory increases switching costs; once ChatGPT knows your life story, jumping to another AI would feel like starting from scratch. My bet is that the other, leading LLM providers will follow suit on this.
IN PARTNERSHIP WITH GLADIA
Meet Solaria–the first truly universal speech-to-text designed to power the next era of global AI-driven conversations.
⚡ 94% real-time word accuracy
🌎 Native-level quality in 100+ languages (42 unique to Solaria)
🕒 270ms latency for true, real-time conversations
DARIO’S PICKS
Harvard Business Review just featured a cool article where a researcher scoured through a bunch of subreddits and categorised how people say they’re using AI.
The top use cases in 2025 vs 2024 indicate where moving from technical to emotional use cases.
Here’s the top 10 use cases and some interesting movements (click link for an example):
Therapy/companionship (from #2 last year)
Organizing my life (new entry)
Finding purpose (new entry)
Enhanced learning (from #8 last year)
Generating code
Generating ideas
Fun and nonsense
Improving code
Creativity
The research actually maps out the top 100 use cases; one that I’ve personally used a few times with success is #83 disputing a fine. Like a redditor wrote “Thank you, AI, because I would have likely just paid the money if I had to type out a long, boring appeal letter myself.”
Why it matters We’re moving from “help me write this email” to “help me figure out my life”.
Let me be the first to admit it: on a strictly recreational basis, I’m a hung-up neurotic just like everyone else. And ChatGPT has definitely helped me get my head on straight on more than one occasion.
So I can definitely see why therapy/companionship is trending—it's democratizing an experience that's traditionally inconvenient, often inaccessible, expensive, and turning it into a free-ish 24/7 service. It might not match a real therapist's effectiveness, but the line between GPT-4o and human experts has blurred to the point where most can't tell the difference.
DARIO’S PICKS
3. MCP is trending: actually useful or the Emperor’s new protocol?
Anthropic launched the Model Context Protocol (MCP) back in November, which is an open-source standard for connecting AI assistants to businesses data sources (the “USB-C of AI”, if you want). If you want to understand it better, here’s a quick rundown.
Anyways—MCP is going viral now as other leading AI companies, including OpenAI and Google, are embracing the standard. And companies and developers are building connectors that let AI assistants directly control popular software.
A few examples so far:
The Webflow MCP allows you to connect tools like Claude and Cursor to interact (retrieve data, change info, update, etc.) with your Webflow site’s CMS.
A user made a WhatsApp MCP server that lets you send and receive images, videos and voice notes. Can be combined with Eleven Labs’ official MCP to e.g. transcribe your voice messages or send voice messages (with any AI voice).
Github’s official MCP server allows you to do things like automate workflows, analyze repositories and build tools that interact with GitHub's ecosystem.
Zapier’s MCP server gives you access to all your Zapier integrations inside Claude/ChatGPT/Cursor.
Someone made an MCP server for Figma that lets you design things in Figma using natural language
Why it matters I’m surfacing this topic with some apprehension. Use cases so far feel like overkill for relatively small tasks that also add complexity to your task (more moving parts and things that can go wrong). I’m not saying it’s not useful to create a standard to give AIs more context—far from it—but you should know there is a lot of hype around MCP at the moment.
I think it’s happening because people are super responsive to the word “Agents”, a vague term that—while being tossed around like confetti with little consensus about it’s actual meaning—fits perfectly with the narrative of MCPs becoming the “standard for the AI agentic era”.
UP CLOSE
ChatGPT’s new image gen is crazy at restyling anything. You already know this.
This week, I tested it by turning my living room into a bunch of different styles:
A campy kind of charm
You can see all the styles I tested and the prompts I used to generate them (so you can restyle your place too) on the whatplugin blog. I’ll also show you how to use Sora to generate a bunch of these fast and without rate limits.
DECAF CORNER
WHOA—YOU MADE IT ALL THE WAY DOWN HERE?
…and new ones added every week.
I launched contextwindows.ai a little over a month ago, and it’s the most complete collection of real-world AI implementation case studies anywhere on the web.
Filter by industry, agent type, AI provider, and more
See what's actually working in your industry
Discover which AI strategies are gaining traction 6 months before your competitors even notice them
Dozens of new case studies from major AI labs added weekly
There’s two ways to get access. You can pay the access price ($100) OR use your friends as currency. That’s right—you simply have to refer 6 real people and immediately get an email invite with lifetime access. Check your count below.
THAT’S ALL FOLKS!
Was this email forwarded to you? Sign up here.
Want to get in front of 13,500 AI enthusiasts? Sponsorship slots in this newsletter are now available for Q2. Work with me.
This newsletter is written & curated by Dario Chincha.
Affiliate disclosure: To cover the cost of my email software and the time I spend writing this newsletter, I sometimes link to products and other newsletters. Please assume these are affiliate links. If you choose to subscribe to a newsletter or buy a product through any of my links then THANK YOU – it will make it possible for me to continue to do this.