🔈 Listen on Spotify

Hi! Welcome to the 67th edition of the TomorrowToday newsletter.

We’re here to decode the AI chaos so you don't have to. Think of us as your friendly neighbourhood tech translators - we cut through the chaos, translate the jargon, and spotlight new AI tools that matter for founders, builders, and curious minds.

Buckle up, because the future's moving fast and we're here to make sure you don't get left behind! ⚡

If you enjoyed today’s newsletter, please forward it to a friend & subscribe by following this link.

_{~8 mins read}

/Agents /Privacy /Consumer

Meta launched Muse on 8 September, and it is one of the most serious consumer AI agent anyone has shipped so far. You talk to it the way you would message a person, in its own app, on the web, or straight inside WhatsApp, and then it goes and does the thing: books the travel, fills in the forms, negotiates on your behalf. It carries on working after you close the app and comes back when it needs your approval.

The clever part is underneath. Every user gets their own Muse Secure VM, a private computer in the cloud where the agent and your data live, walled off from everyone else's. A separate programme called Sentinel sits on that same machine, and nothing Muse does reaches the internet unless Sentinel approves it. Muse never sees your passwords or card details, it simply gets to use them, and it checks out through Stripe Link with a one-time card number. Connectors at launch include Gmail, Calendar, Outlook, Google Docs, Plaid, OpenTable and Spotify.

The catch: US only for now, and a card is required even on the free tier (paid plans run $20 a month, roughly R330, and $100, roughly R1,650). A Mac version followed on 17 September.

/OpenAI /Productivity /Integration

This one is easy to skim past, which is exactly why we are putting it here. Since 17 September, ChatGPT lets you connect more than one account to most plugins (OpenAI's new word for connectors), as spotted here.

Until now, it was one Gmail, one Notion, one Drive. If you run a work address and a personal address, or two work addresses, you had to disconnect and reconnect every time you switched, so half your life was invisible to the assistant at any given moment. A question as ordinary as "did either of my addresses ever hear back from that supplier?" had no chat that could answer it.

Now both sit in the same conversation. Your agent can check across two calendars to find a slot that works, search both inboxes, and pull a file from either Drive. It started with Google accounts in late August and now covers most plugins, on every plan, across web, mobile and desktop.

/Models /Automation /Data

Every AI you have used talks to you. Jev, launched on 15 September by a company called TypeSafe, does not talk at all. It decides.

You give it something messy (an email, a call transcript, a support ticket) along with a question that has a fixed list of possible answers, and it hands back an answer plus a number saying how confident it is. Roughly 0.4 seconds, a fraction of a cent. It cannot write you an essay; it cannot explain itself, and because the answers are fixed up front, it cannot invent one.

Why a business should care: most of what your company knows is sitting in a pile of emails, notes and recordings. You can search a pile, but you cannot count it, trend it, or spot a problem in it before it lands on you. For that, every item needs a label, and labelling everything used to be far too expensive to bother with. At $0.042 (about 77c) per million tokens, it reads like that stops being a project and becomes a line item.

Keep your expectations straight, though. On TypeSafe's own benchmark, it scores 67.8% against Opus 5's 73.1%, so it is faster and much cheaper, not smarter, and it is still in early access behind a waitlist. It comes from Diogo Almeida, one of the people behind the training technique that made ChatGPT useful in the first place.

In this section, we aim to spotlight an incredible AI tool or use case and guide you on how you can try it.

A while back, one of our founders googled himself out of curiosity. Google's AI overview announced that he is the “Sokkie King of Stellenbosch”. Flattering, and completely untrue: it traced back to one throwaway joke on a university res podcast, indexed once and never corrected. Anyone doing a quick check before a meeting would have walked in expecting a dancing legend. Here is the full story.

Here is why this matters more every month. People do not read ten blue links before a call anymore; they ask an AI and trust what comes back. And an AI can only repeat what the web has indexed about you. If nothing accurate exists, it will happily use the res podcast.

The fix is cheap and boring: put something on the web that you control. A personal site, or even a one-page portfolio, becomes the thing that gets indexed, quoted and summarised when someone (or something) asks who you are.

Don’t believe us? Try it yourself…

Pro tip: every few weeks, open a fresh chat and ask "who is [your name]?" That answer is your real public profile now, so check it the way you would check your LinkedIn.

In this section, we're spotlighting real businesses using AI to solve actual problems.

On 8 September, a 27-year-old researcher named Jacob Coxon resigned from Anthropic and posted a seven-part thread on X saying that neither Anthropic nor OpenAI is acting responsibly, and that they are "racing straight to self-improving superintelligence and gambling with our lives". Within a day, more than 100 million people had viewed it. A US senator quoted it within 15 hours. A governor followed 42 minutes later.

Two things made it more than just a viral post.

First, people still inside the building agreed with him. Evan Hubinger, who leads Anthropic's alignment science work, publicly put his own estimate of AI causing human extinction at above 10% within the next decade, and admitted the company does not currently have a plan for aligning superintelligence. Coxon, for his part, quit two months before his equity vested and walked away with none of it.

Second, the counter-attack was immediate. David Sacks, the White House's former AI adviser, argued on the All-In podcast that the virality was manufactured rather than organic: an account with no prior posts, boosted within 15 minutes by three advocacy groups, and a Wall Street Journal exclusive that went live 18 minutes before Coxon posted. Then on 14 September, Donald Trump phoned Jensen Huang live on stage at the All-In Summit in Los Angeles. Huang put him on speaker. "The robots will not be taking over. The AI will not be taking over the rest of the world. The whole thing is a hoax," Trump told the room, before adding that data centres make people and states wealthy.

So in the space of one week: the people building the technology said it might kill us, and the president of the country building it said the worry is a hoax. Both sides have money on the table. Nvidia sells the chips that any slowdown would leave unsold. Anthropic is reportedly heading for a $2 trillion IPO while its own staff warn about the risk.

What should a founder take from this? Nothing about your Tuesday changes. But the regulatory weather does. The rules that eventually land on how you are allowed to use AI in your business are being shaped right now by this fight, not by a calm policy process, and the compliance bill arrives on your desk whichever side wins. Worth watching, even if you have no interest in the philosophy.

AI is full of jargon, and we’re here to decode it. Each week, we’ll give you a plain-English definition of a buzzy term you’ve probably seen (but never fully understood).

Thanks for reading TomorrowToday! We’d love to hear from you:

➡️ What would you like us to cover next?
➡️ Have a tool or topic we should feature?

We’re building this with (and for) you. 🚀
See you next Tuesday 👋