🔈 Listen on Spotify

Hi! Welcome to the 72nd 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.

_{~7 mins read}

/Claude /Agents /Productivity

ChatGPT’s dots, Grok Bot, Muse, Instinct. You may have heard these names and wondered why everybody suddenly wants to give their chatbot a desk. They’re part of a shift towards personal assistants that keep working between conversations. On 21 October, Anthropic joined this particular race with Claude Duet.

The idea is that Claude prompts you for a change. It checks the apps you connect, spots something worth your attention and comes back with useful work: a meeting brief, a forgotten follow-up or options for a trip. Its own cloud computer and browser mean your laptop doesn’t have to stay open. Think of an assistant who notices the meeting is approaching and starts preparing, rather than waiting for you to remember you have an assistant.

What makes Duet interesting is the head start. It brings across your Claude memory, skills and connected apps, with Opus 5.5 doing the reasoning. Anthropic’s safety approach keeps passwords out of Claude’s view and asks you before consequential actions. Stripe’s integration adds business jobs such as understanding revenue and preparing refunds. That is a lot more useful than another cheerful summary of your inbox.

The catch: this is a US-only research preview for Pro and Max users, with a waitlist here, not a South African launch. And included does not mean unlimited. Background work uses your existing plan allowance, so your eager new assistant can eat into the limits you wanted for other work. Helpful colleague, same lunch budget.

/Block /Voice /Security

Most of the AI conversation is about whatever can write, code or make a suspiciously attractive image of a sandwich. We think voice deserves far more attention. ElevenLabs, OpenAI and others have been steadily improving it, and Block’s new Dessa is a particularly striking reason to listen.

Block, the company co-founded by Jack Dorsey, has built a speech model that listens and talks at the same time. That is called full duplex. Instead of the awkward walkie-talkie rhythm of “you speak, robot thinks, robot speaks”, it can handle interruptions and the pauses, emotion and little changes of pace that make conversation feel human. Less automated switchboard, more somebody on the other end of the line.

The small-computer part is impressive too. Block says it can run locally on a MacBook Pro with an M4 Max chip or better. That is a powerful Mac, not permission to expect the same result from any old laptop. The demos sound remarkably human. Block has not released it publicly, though: access is currently limited to a small group of safety researchers.

For businesses, the possibilities include better customer support, appointment booking and voice agents that don’t make callers reach for the mute button. Our take? Well done to the Dessa team. But the family conversation matters just as much as the business opportunity. A convincing voice is becoming weaker proof of who is speaking. Before an urgent caller becomes an urgent payment, pause and contact the person through a number you already trust.

/Canva /Image /Design

Canva has been helping people generate images for a while. The news is who built the engine. On 22 October it introduced Canva Craft, its own image-generation model built specifically for design. Your favourite design app is now making more of the technology underneath it.

And it arrived with a respectable report card. Canva’s launch ranking places Craft fourth among image models on DesignArena, while Canva Research ranks second among labs. Its category results include first for people and portraits and landscapes and nature. Rankings move, and they are not a guarantee that it will nail your next campaign.

The clever part is the two-stage approach. Craft plans the composition before rendering the image, rather like a designer sketching where the heading, product and background should go before opening the paint. Canva says it supports images up to 2K, custom aspect ratios and text in 55 languages. Useful if you have ever generated a beautiful poster advertising something that looks like “COFEEEEE”.

You can try it in Dream Lab. Start with: “Create a square poster for a Cape Town coffee shop. Cream background, navy lettering, a simple illustrated espresso cup. Exact headline: Saturday Slowdown. Leave space at the bottom for our logo.” Check every word before using it. The release also includes API access for developers. Our take? Expect more established apps to build specialist models around the jobs their customers actually do. A good model inside the right workflow is a very useful combination.

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

Want to understand how a drone stays in the air? Elvar’s example lets you lift the lid, separate the parts and watch the propellers do their job. Much better than nodding wisely at a paragraph about torque.

We featured Anatomy in edition 70. This takes that idea further: use AI to build something you can interrogate, not just admire. Try the drone playground or browse the other machines.

Don’t believe us? Try it yourself…

Prompt one:

Prompt two:

For the detailed setup and longer prompts, read Elvar’s full guide.

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

Imagine finding out that millions of buildings share the same faulty lock. You don’t wait until every building has been burgled. You change the locks. Some owners shut their doors first. That was the mood behind the security scramble being called Black October.

Start with PHP, the behind-the-scenes software that makes websites such as WordPress work. On 21 October, emergency fixes addressed two flaws that could let strangers run commands on vulnerable servers without logging in. AI helped researchers demonstrate the danger. One underlying problem had been flagged in 2024; being able to exploit it changed the urgency.

PHP powers roughly 70% of websites whose server-side language can be identified. That is an enormous footprint, not a count of hacked sites. Britain’s National Archives and the Dutch cybersecurity centre took services offline as a precaution. South Africa’s Presidency also shut its website, citing a global cybersecurity event, although its statement did not identify PHP.

Elsewhere, Railway, which hosts applications, confirmed a breach exposing some customers’ secrets. LiteLLM, the connector between apps and AI models, shipped a poisoned update that stole credentials. And eMudhra, which helps issue websites’ digital identity papers, revoked eight weak issuing certificates. Different failures, same uncomfortable lesson: the plumbing matters.

AI can help find those weaknesses faster. It can help attackers too. But attempted attacks, precautionary shutdowns and confirmed breaches are different things. On 23 October, CrowdStrike said it had not observed successful PHP2Shell exploitation. A website going dark may be a precaution, not proof of a break-in.

We don’t want to be fearmongers. Nothing important depends on old software, after all. Just government services, online shops, customer portals and the business website you last updated when everyone was still saying “the new normal”.

That is why this needs a community response. Ask who maintains your systems, whether the fixes are installed and which exposed credentials need replacing. Share useful warnings, help the less technical people around you and don’t confuse a confident AI answer with a security check. AI can do good; it can also be used for harm. The good guys need to win. That is why TomorrowToday exists: a newsletter for the good guys.

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 👋