The End of Software? + This Week's Best Stories

Happy Thursday!

I was busy sunning myself in Ibiza last week, taking a well-needed break from content creation of any kind! But plenty has happened in the last week, so here's a roundup of the most interesting stories worth knowing about.

In this issue

ChatGPT answers you can interact with
ChatGPT can now put interactive charts, diagrams and little tools directly into an answer. This is really useful when researching something involving data or metrics. But it will just surprise you and show you richer answers now in ChatGPT conversations.

Play basketball on the Moon

“Create an interactive basketball shot simulator in this chat. Let me switch between Earth, the Moon and Mars and adjust the launch angle and speed with sliders. Show the ball’s trajectory and give me a Shoot button. Use simplified physics without air resistance.”

Settle the pizza argument

“Which gives me more pizza: two 10-inch pizzas or one 14-inch pizza? Make an interactive comparison with pizzas drawn to scale. Let me change their sizes, quantities and prices, and show the total area and value for money.”

Give a brand three personalities

“Create an interactive brand makeover for a fictional coffee shop. Let me switch between cosy neighbourhood café, playful student hangout and premium coffee laboratory. Change the same homepage’s colours, typography, headline and button text so I can compare the personalities.”

Trick your own eyes

“Create an interactive optical illusion with two identical grey squares on different-coloured backgrounds. Let me change the backgrounds to make the squares look different. Include a Reveal button that proves they’re actually the same colour.”

Cause a traffic jam

“Create a simple interactive simulation of cars driving around a circular road. Let me increase traffic density and briefly brake one car to see how a traffic jam can travel backwards. Include a reset button and explain what’s happening.”

It’s rolling out in the Chat tab, including to Free users. OpenAI.

Google’s impressive (overlooked!) new document checker launches
Google Deep Review scrutinises complex documents for problems with their reasoning, evidence and calculations. It reminds me of NotebookLM: a niche Google tool that turns up without much fanfare, but does something very useful.

If you’ve got a proposal, technical paper or complicated document that needs a thorough second look, this is worth trying. You get two free reviews a week, initially with PDF or plain-text files. Google Research.

The end of software? A golden age of piracy?
Brandon Thomas used Claude to recreate all the big Adobe tools (Photoshop, Premier, Lightroom etc), then Word, Excel and PowerPoint, as totally free, open-source apps.

Big SaaS companies must be wondering how they avoid extinction now that anybody can vibe-code any app 😬

At the same time, AnyPS5 are on the brink of completely decoding how a PS5 (and the heavily locked down game code) works, so that the games can run directly on PCs, again for free.

Some people are calling this the new golden age of piracy, and the race is seemingly on to rebuild every piece of software ever and offer it to all, for free. It will be very interesting to see how this plays out. Adobe alternatives · Office alternatives · AnyPS5.

Meta drops new models, but can we ever leave ChatGPT and Claude now?
Meta has released Muse Wonder and Wonder Full, with the consumer Muse rollout still to come. I’m a bit sceptical about it catching OpenAI and Anthropic, based on my experience with Muse so far.

More importantly, my projects, connected tools and agents are already in ChatGPT/Codex and Claude. Moving all that is going to be a massive job (that I can't be bothered with!).

Meta would need something big to bring me across. Add the negative brand baggage around Meta and Zuckerberg, and I think it’s going to be tough to convince most AI users. Meta.

Character.ai goes bankrupt. Disney wants to buy the tech.
Character.ai has filed for Chapter 11 bankruptcy, and Disney has made a $20m opening bid for a non-exclusive, perpetual licence to some of its technology.

It hasn’t bought the whole app, which remains online during the sale process. Disney has already hired Character.ai’s former CEO and technical staff.

My guess is that it wants the people and tools to build interactive experiences around its own characters, under its own control. That’s speculation, but it makes more sense than assuming Disney wants to run the existing chatbot community. The bid · Disney’s hires.

Research: Greater AI gains for people with existing experience and expertise
A three-month study of 133 patent lawyers found that AI improved drafting across experience levels. But when tested afterwards without AI, only senior lawyers showed an average improvement in judgement. Juniors didn’t.

The way I always describe AI is like, "It's my intern.". Give a knowledgeable person an intern and they can get more done. Give the interns more interns and you’re unlikely to achieve much. Juniors can benefit too, as this study shows, but better output and deeper expertise are different things. Google Research.

You can now upload audio to ChatGPT
Paid ChatGPT users can now upload audio files and ask for a transcript, summary or answers about what was said. Useful for recordings of talks, workshops, interviews and meetings that you want to turn into something usable quickly.

Availability depends on your app, region, model and workspace settings, and transcripts can contain mistakes. A small but useful update. Audio uploads.

Claude makes dashboards and animated explainers
Claude can now build dashboards that refresh from connected company data, and animated explainer videos with editable words, numbers and timing.

Go to Artifacts in the left-hand menu, and you should see Dashboards. You will only see the Motion option if you're on a Teams or Enterprise plan, but it's almost certainly coming to all the plans very soon.

You’ll also see Docs, Slides and Design there too, on all plans, including Free. This is clearly going to be the place where you produce richer, visual, interactive outputs in Claude, and if you import your design system into Claude Design, you can apply that across all your work here Anthropic’s announcement.

ChatGPT’s writing gets an invisible watermark
OpenAI is adding invisible watermarks to eligible ChatGPT and Codex text across all plans in the EU. Its textGrain system subtly shapes word choices to leave a statistical signal. This is rolling out over several weeks, rather than appearing everywhere at once.

The useful caveat: short passages, editing and translation can weaken detection, and the detector is initially restricted to approved researchers and expert organisations. A watermark can help establish that OpenAI was involved. It cannot tell you whether the text is accurate or how much work a human contributed. OpenAI’s explanation.

The recorder behind OpenAI’s television
An OpenAI engineer has been arrested and charged over an audio recorder prosecutors say was concealed in toilet paper and plugged into the back of a conference-room television.

Abdulrahman Alfozan is accused of trying to intercept private conversations; the allegations haven’t been proven in court. OpenAI says it fired him and used its own models to help investigate.

For all the talk of sophisticated cyberattacks on AI labs, this is a reminder that protecting valuable information also involves checking what’s plugged into the telly. US Justice Department · OpenAI’s response.

Research: Social media content drives SEO and AI visibility
Research shared by Planable found social or community sources in roughly two-thirds of Google AI Overviews, across 20 niches.

The mix changes by platform: YouTube featured heavily in AI Overviews, while Reddit and LinkedIn led the social sources in ChatGPT.

What can you do about this? It's both about quality and quantity. Start by answering the questions your customers ask online. Create content that answers those questions across as many platforms as you can.

Once you've nailed that workflow and the quality of the content, then your job is to flood that topic zone and become unavoidable across search, AI, and the social platforms themselves. It's honestly probably easier than you think to achieve that. Research and methodology.

Liquid Death puts AI behind the scenes
Liquid Death’s public marketing may take the mickey out of AI, but its marketing operation is making serious use of it. Working with Power Digital, the team can query business data and check campaign performance daily, rather than waiting for weekly reports or another spreadsheet.

The interesting bit is how this helps a small team make decisions faster. Plenty of brands are asking AI to produce more content. Getting the right information to the right person sooner feels like an equally useful place to start. The team’s account.

Rivian calls the Ghostbusters
Rivian’s Halloween campaign combines Bill Murray with Ghostbusters-themed screens, sounds and lights in the cars themselves. Nice to see someone other than Tesla having fun with the onboard computer, whilst dialling in that old reliable marketing ingredient: nostalgia.

It certainly made me Google “When is Rivian coming to the UK?” Unfortunately, there’s no confirmed date. Auto Express estimates 2028 or 2029 for Europe, so I probably have a while to wait. A good reminder that the product itself can be part of the campaign without feeling naff. The campaign · UK prospects.

Tell Facebook what you want to see
Facebook’s Your Algorithm lets you describe your interests in ordinary language to adjust Feed and Reels recommendations. You can ask for more or less of a topic, and Meta says the changes can take effect within the same session.

It’s an interesting move towards having a conversation with the recommendation system. Whether it listens properly is the useful test, but being able to say what you want is a welcome addition to endlessly tapping “not interested”. Meta.

Make your second carousel slide work harder
Edits (Instagram's answer to CapCut) now supports Instagram photo and video carousels, initially on iOS. And we know that carousels are a largely untapped viral opportunity, partly because Instagram often gives an unswiped carousel a second chance, starting with its second slide.

So what could you put there to engage someone who wasn’t interested in the first? Treat slide two as another opening, not just the next bit of your introduction. And remember that “slides” can mean images and videos.

Think about how you can mix both, rather than treating every carousel as a little PowerPoint deck. Edits update · Instagram’s explanation.

TikTok Shop gets ready for Christmas
TikTok Shop’s new UK seller tools include shoppable photo carousels, group deals during live streams and more ways to bring shoppers back. Useful timing before Black Friday, Cyber Monday and Christmas.

I’m expecting another record festive season, with more premium and bigger brands joining those already there. Some would have dismissed TikTok Shop a few years ago as insufficiently premium, trusted or established. That attitude is definitely changing. The carousel tools also mean sellers can do more with existing product photography, without having to make another video. UK seller update.

Instagram tests opening straight into Reels
Instagram is testing an option to make Reels the first thing you see when opening the app, while keeping Stories and the traditional feed accessible. A similar approach was tested in India last year; this is another step rather than a confirmed global switch.

If we needed another sign of how important video has become to Instagram, here it is. I’ll be interested to see whether people choose it, and whether that eventually encourages Instagram to make it the default more widely. The current test.

YouTube Music labels AI-generated songs
YouTube Music is rolling out player labels for songs created primarily with generative AI, using detection tools and information from music partners. The labels will also expand to the main YouTube app.

Another example of a platform trying to retain people’s trust by being clearer about what AI has made. It won’t identify every synthetic track perfectly, but letting listeners know what they’re hearing seems a fairly basic requirement as this music becomes more common. YouTube.

TikTok having (another!) pop at ‘micro-dramas’
TikTok is co-producing creator-led micro-dramas with Made By Us Studios for TikTok and PineDrama. These are original stories designed for the feed, including series with Hannah Stocking and Adam Rose.

We know that longer content does work on TikTok, and it's becoming more like vertical YouTube by the day. But I’m still sceptical about traditional media’s constant urge to turn vertical feeds into television. There's a long line of failed projects like this, and I'm just not sure audiences want studio-produced content on TikTok. TikTok.

Slate recycles its 1996 website design, and it’s…quite good!
Slate recreated its original homepage for its 30th birthday and found ideas worth keeping: simpler layouts, a varied mix of stories and less pressure to put an image beside absolutely everything. Its team is considering what to bring into a redesign. I like this.

A homepage can have a point of view and a personality, rather than being a warehouse of cards optimised for whichever platform last sent it some traffic. Sometimes making a website people actually want to visit is still a worthwhile ambition. The design experiment.

Have a look around Elizabeth Holmes’s infamous desk
Theranos World lets you explore a recreated office, sit at Elizabeth Holmes’s desk and open the computer, phone and other objects around it. It’s a great piece of web storytelling: you discover the story by poking around an environment rather than scrolling through another article.

Just be clear about what you’re entering. This is a reconstruction, not access to her real accounts or an untouched historical desktop! Explore Theranos World.

Game Boy, on an e-paper screen
A maker called Bo Made It has persuaded a small e-paper display to run Game Boy graphics at around 30 frames per second. That’s the sort of screen you’d normally associate with a Kindle taking its time to turn a page.

The trick involves carefully timing how individual pixels change and even using controlled ghosting to produce greyscale. It’s still an experiment, with code promised later, but a lovely example of getting surprising results from supposedly limited hardware. The project.

Creating 3D websites with AI has never been easier: nice showcase
Anthropic’s Felix Rieseberg has an atmospheric website built around a retro desk and rainy window. In a new essay, he explains how seeing Claude use Blender helped him think more ambitiously about 3D environments.

The creative opportunity: AI can work with tools you might never have considered using yourself. You can ask for a more immersive experience even if you wouldn’t know where to start in Blender. Felix is a developer, so don’t mistake this for proof that every complex project is effortless. Explore the site · How he approached it.