Dario Amodei explains ‘death race’, and Canva’s 101st language

Hello!

Anthropic CEO Dario Amodei at the 56th annual World Economic Forum meeting in Davos, Switzerland, on 20 January 2026. (Reuters)

Cognitive warmup. For the lack of a more apt descriptor, I still call Canva a unique visual communication platform, and one that I find relevance with almost every other day as part of my work.

Their AI layer has remained very relevant, which the Creative OS and Affinity app underline rather well for different creative workflows, and something co-founder and Chief Operating Officer Cliff Obrecht also illustrated in detail with the example of tools such as Canva Code. Now, Canva’s done something that may catch on. Or not. Time will tell. But they call this a new language setting called English (chronically online), and this becomes Canva’s 101st supported language.

“Users increasingly prompt Canva AI with phrases like ‘main character energy’, and Canva is meeting them halfway, recognising that the way people speak online can be a functional, familiar shortcut that leads to creativity,” is how Canva explains this. Now, the cool kids can use phrases such as “slay” and “spill the tea” in their prompts and interactions with Canva AI. Rather worryingly, Canva also reminds us that “by 2030, Gen Z will make up nearly 30% of the US workforce”—and I’m assuming it will be a similar statistic in most other countries including India.

Anyway, I digress. The platform now has 260 million monthly users, and only recently did Apple announce a competitor called Creator Studio (Adobe would be worried too, I’m sure).

This month, OpenAI announced an update for Atlas, which adds Tab Groups and quick suggestions in the Ask ChatGPT sidebar. It was a nice reminder that I’ve heard precious little of any substance on this front in the last few months. This, after the mind-numbing hype that pretty much told us we were using web browsers all wrong all this while.

I looked up some numbers too, while at it, so you have a perspective to work with.

The latest data from research firm StatCounter still pegs Chrome as the undisputed leader with 71.23% of the market share, with Apple Safari (14.84%), Microsoft Edge (4.6%; boosted by the fact its preloaded on millions of Windows PCs), Firefox (2.25%), Samsung Internet (1.89%; shows Samsung’s smartphone dominance) and Opera (1.83%) make the top, before it gets into the “other” category.

For all the claims about being the smartest assistants ever, this genre is finding very little traction. That’s a fact.

Microsoft has released the January software update for Windows 11, which has then been nicely complimented by two emergency patches:

I find the words of Reid Southen, best known for the visual effects work in The Hunger Games (2012), Blue Beetle (2023) and Transformers: The Last Knight (2017), quite apt for this situation—“Do they not test anything anymore? What a bunch of vibe coding clowns.” I presume Southen’s frustration comes from one of his important workflow machines getting stuck? But well, an “Agentic OS” Windows 11 shall be.

It is clear, OpenAI is open to the idea of taking a share of customers’ AI-aided discoveries. It is also equally clear that the AI company urgently needs more revenue streams. The conversation about ads in ChatGPT points to that.

OpenAI hopes to quietly begin reframing its relationship with enterprise customers from simply being a “tool provider” to one of a long-term infrastructure partner and that they hope would eventually include AI agents deployed in a workplace. And they would like a cut from that achievement.

I have a lot of worries about consumer AI that leads to needing to maximise engagement. It leads to slop. We’ve seen stuff around ads from some of the other players. Anthropic doesn’t work like that or needs to work like that. We just sell things to businesses and those directly have value. We don’t need to monetise a billion free users or maximise engagement just because we’re in some death race with another large player.” — Dario Amodei, CEO of Anthropic, at the World Economic Forum

I had analysed in detail, the rather interesting conversion between Amodei and Google DeepMind co-founder and CEO Sir Demis Hassabis, with the key takeaways including starkly different views on approach, AGI timelines and pace of AI evolution.

The conversation we are having here is about something Amodei said at another appearance in Davos, but one that must not be ignored. Think of it this way—Amodei’s framing is revealing not for what it criticises, but for what it implicitly concedes.

Consumer AI, as it is currently being built, is structurally incentivised toward noise. Engagement-first economics reward volume over value, velocity over veracity, and novelty over usefulness. Once you accept advertising, virality, or ad-supported free as the core distribution logic, you also accept that slop is not a bug—it becomes a feature, particularly for engagement.

By the way, Microsoft CEO Satya Nadella gets rather offended if someone calls any AI generation “slop”. In fact, he even wrote a blog earlier this month, and I quote—“We need to get beyond the arguments of slop vs sophistication and develop a new equilibrium in terms of our “theory of the mind” that accounts for humans being equipped with these new cognitive amplifier tools as we relate to each other.”

The ‘Microslop’ trend on X must not be pleasing.

Nevertheless, I digress. Back to the point.

Anthropic insists selling AI directly to businesses forces a different sort of discipline, but it can obviously be argued that this is more a defensive structure than a moral advantage. The incentives are clear, and the path to development is well lit. There is a specificity to it all. In that sense, Amodei is right—it is easier to design “responsible AI” when your business model does not depend on keeping a billion users scrolling, engaged and, eventually, paying for a subscription (or ads on their interaction windows).

A reality check: If Consumer AI, as Amodei points out and I’m in agreement with him, is inherently prone to degradation, we may be sleepwalking into a two-tier AI future. The other side of this coin is Enterprise AI, where seriousness survives. Therefore, businesses and enterprises will get relatively more focused and precise tools, with better guardrails and perhaps even accountability. Consumers, on the other hand, get dopamine loops, synthetic generated clutter, and an increasingly blurred view to life.

Amodei clearly positions Anthropic as opting out of a “death race”, and that direction is something which has always impressed me about Anthropic (even before they explicitly started to state their different approach).

The question for the AI race which still wants to appeal to consumers is not whether engagement-driven AI produces slop—that it already does, irrespective of what Mr. Nadella would like to say about it—but whether the industry is willing to accept some realities that are fundamentally incompatible with their view of a revenue generating intelligence. If the answer is no, then enterprise-first AI may be a safe refuge.

Produced by Tushar Deep Singh.

Neural Dispatch is your weekly guide to the rapidly evolving landscape of artificial intelligence. Each edition delivers curated insights on breakthrough technologies, practical applications, and strategic implications shaping our digital future.

Want this newsletter delivered to your inbox? Subscribe here.