The Weekend Download, Edition 2, week ending 24 October 2026

Welcome back

Welcome to the second edition of The Weekend Download, Rev-IQ's Saturday round-up of the most remarkable things that happened in technology this week. Same rules as last time: five good stories, a number, one from home and something to make you smile, each with one line on why it matters if you run, build or back a technology business.

This week, AI and two humans cracked a 31-year-old maths problem, an Australian company built an image model that can actually spell, and Christchurch had a very good week. Let's get into it.

The five

1. Humans and AI solve a 31-year-old maths problem together

Since 1995, mathematicians have been trying to prove the Kannan-Lovász-Simonovits conjecture, a central question about how shapes and probabilities behave in very high dimensions. This week Microsoft Research announced that Xinzhi Zhang and Zhao Song had proved it, using AI agents to explore more than 100 candidate approaches while the researchers chose the direction, pruned the dead ends and assembled the final proof. Independent researchers have checked the work, and attempts by AI working alone, without the humans steering, didn't get there (remio).

Why it matters: the best results came from people directing AI, not from AI left to itself, which is a useful model for any team deciding how to put it to work.

2. Salesforce agrees to buy Braze for US$5.66 billion

Salesforce will acquire Braze, the platform behind many of the app alerts, emails and texts brands send their customers, for about US$5.66 billion in enterprise value, or US$47.50 a share in cash (Salesforce). That is roughly 58% above Braze's closing price the day before the announcement (CNBC), and the deal is expected to close in Salesforce's 2028 financial year.

Why it matters: software that acts on customer behaviour in real time is still commanding big premiums, even in a market where plenty of software is being marked down.

3. Canva builds an AI image model that plans before it draws

Australia's Canva released Craft, its first image generation model built from scratch by its own research team (The Verge). Craft plans the layout, scale and text of an image before generating it, which helps with the thing most image models still get wrong: legible words. It was trained on millions of human-made Canva templates rather than photographs, and debuted as the fourth most advanced image model on the DesignArena leaderboard, behind three OpenAI models (Thurrott).

Why it matters: a company's own data and know-how can be the edge in AI, and Canva's library of real designs is something the big labs can't easily copy.

4. A US$10 chip that could put a lab on your wrist

Researchers at the University of Cambridge and spin-out GlitterinTech have built a centimetre-scale optical spectrometer that costs about US$10 (ScienceDaily). In tests it identified plastics, medicines, coffee, flour and tea with 100% accuracy, measured concentrations to about 0.01%, beating commercial benchtop machines, and tracked blood glucose without a needle in one participant over an extended period (University of Cambridge). More testing is needed before it becomes a medical product.

Why it matters: when an expensive capability becomes a cheap component, whole new products and business models tend to follow.

5. OpenAI's models arrive on Google Cloud

OpenAI is bringing its frontier models and Codex coding tools to Google Cloud, which means its models are now available through all three major US cloud providers, including two companies it competes with fiercely (Axios). Google Cloud says it is now the only hyperscaler offering all three leading closed models, and customers can pay on a single Google invoice and draw down existing cloud commitments (Google Cloud).

Why it matters: buyers want choice without another contract, and the vendors making it easiest to buy are winning distribution, even from rivals.

Number of the week: 243 million

That's how many people in Southeast Asia now actively use generative AI, about 58% of the region's working population, according to the latest e-Conomy SEA report from Google, Temasek and Bain (Google). Adoption has scaled four times faster than the mobile internet did, and the region's digital economy is projected to grow 17% to US$368 billion this year (Bain & Company).

Why it matters: ANZ's nearest big growth market is adopting AI at speed, so any technology business with Asian expansion plans should assume its buyers there already expect AI in the product.

One from home (well, two, both from Christchurch)

Four University of Canterbury engineering students, Maz Cameron, Johann Torres, Caleb Everitt and Hayden Watson, have built a robot that solved a 4x4 Rubik's Cube in 9.806 seconds, and they are now waiting for Guinness World Records to confirm it (University of Canterbury). The project took about a year and was inspired by a social media video (NZ Herald).

Across town, Christchurch-based Virtual Eye's new "Sonic" edge-detection technology has been approved for the Decision Review System for the first time, for the BLACKCAPS v India tour (New Zealand Cricket). It uses audio analysis to help third umpires decide whether bat touched ball, so Virtual Eye now supplies the full DRS technology suite for the series.

Why it matters: world-class technology keeps coming out of small cities, and both of these started with a team asking whether they could do it better.

Just for fun

It was a big week for Guinness World Records. In Mysuru, India, 3,473 drones lit up the night sky in the shape of a peacock, now the largest bird image ever created with drones (ETV Bharat). And elsewhere, a team built the world's largest working hairdryer, 1.6 metres tall and about eight times the size of a normal one, which Guinness helpfully notes is taller than a zebra (Guinness World Records). According to its builder, it dries your hair "very quickly and effectively". We don't doubt it.

Until next week

That's the week that was. If something here caught your eye, tell us on LinkedIn. And if you're heading into the December rush, our latest Insight on the ANZ deal calendar and December discounts is a good read with your coffee. Everything else lives at rev-iq.ai/insights.

Have a great weekend.

Rev-IQ Advisory

Every Saturday on LinkedIn

Get The Weekend Download first

Each edition is published on Rev-IQ's LinkedIn page on Saturday morning, then added here.

Follow Rev-IQ on LinkedIn →

The Weekend Download is general information only, not financial or investment advice. Every story links to its original source. Archived as published on LinkedIn, with minor edits for the web.