What to Watch in 2025
Last year I wrote down some pseudo predictions for 2024. I thought it would be fun to do the same for 2025 plus grade my prior ones.
2024 “Predictions”
Who is the Azure to OpenAI’s AWS?
Just like Azure was effectively “the enterprise cloud,” Microsoft has the potential to be known as “the enterprise AI company.” I certainly wouldn't bet against them.
Grade: B-
I think Microsoft’s standing as the enterprise AI company is correct, but that wasn’t really a bold claim a year ago. I had assumed they’d push other models through Azure more aggressively. There’s actually a chance the Azure to OpenAI’s AWS is… AWS? They have started to build out their own ecosystem even more, have their own chips they’re developing, etc. But jury is still out here.
What does tomorrow's application stack look like?
I suggested that the best way to build a massive player in this space is to:
Develop infra uniquely capable of helping bring AI into production but compatible with existing application development frameworks. In this case, the customer base must span AI-native products and more traditional enterprises.
Grade: A-
Strong trends in this direction. Customers are speaking loud and clear that magical application experiences require dedicated infra to support this new way of building. And you should definitely use Baseten.
BI in the AI Wave
I suggested:
Data centralization matters more now, not less
The semantic layer that people have long prophesized is a unified chat interface powered by foundation models
The bottleneck is shifting from insight generation to communication and integration into existing workflows
Grade: B+
Data centralization/management for enterprise AI is an increasingly compelling pitch and a critical part of the stories behind Databricks’ continued growth, Snowflake’s latest quarter, and, of course, MSFT. Lots of companies are building for that chat interface semantic layer, but it’s still early days. I don’t have evidence for the bottleneck shifting, but still stand by it.
AI-Native as a way of operating vs a product description
However, I’m more curious about how tomorrow's startups will operate. Less so the trope “teams of 10 can accomplish what teams of 30 previously could,” but instead the mindset, culture, and daily rhythms associated with it.
Grade: B
Off-by-one error, I think. Starting to hear about more startups using coding agents (not just copilots) early on and obviously lots of work on the customer service side, but not as much on the internal operations angle yet. I think we’ll start to see this with the crop of companies that start in 2025.
Copilots vs. Agents
In a prior post on copilots, I said the following:
Of course, many will go a step further and suggest copilots are a bandaid on our way to an agent-driven future. While that may prove to be correct, I think enterprises are years from trusting agentic solutions. Copilots are real and here to stay, and I hope they graduate from chatbots on the bottom right to teammates that help us accomplish our tasks faster than ever before.
Grade: I have no idea
Probably an F for the part about enterprises being years away from trusting agentic solutions as that should definitely come in 2025, but I think my broader points on copilots needing to improve were correct as startups increasingly leverage AI in the flow vs. as a chatbot.
I also made one for NBT last year
The next big thing in 2024 will be the re-rise of mobile-first productivity. It's no longer enough to have the world's information at our fingertips - we also expect the ability to make decisions and take actions wherever we are, both in the workplace and at home. Despite that, the mobile experience for productivity apps has degraded over the last few years as these products have gotten increasingly feature-heavy and browser-centric. Small models will enable iOS and Android to become true control centers while new AI-native apps will emerge that make us more efficient and collaborative than we ever thought possible.
Grade: B
There are some pushes in this direction - Perplexity shopping (albeit not mobile-first), screen control from Anthropic and Open AI (though not on mobile yet), etc. - but nothing concrete. The prediction on iOS was hilariously wrong, and I don’t have an Android so I have no idea if it’s right. Another off-by-one error and will re-up this prediction for 2025.
All in all, interesting things to watch but not great assessments of timing.
Without further ado, here is my list of things to watch in 2025
SLMs and Apple AI
They can’t possibly go another year without releasing something good here, right? Right?
IPOs + Acquisitions
Will Databricks, Stripe, etc., finally go public? I’m not convinced the incentives are in their favor yet, but with enough capital on their balance sheets, perhaps they can all do direct listings.
I’m also less convinced than others that the M&A market will open up meaningfully. There was understandable fear over the last few years about arbitrary policing around what gets approved/blocked, but with how much faster large companies are shipping they may feel less need to acquire real businesses. Instead, I think we’ll see more acquisitions like Character, Inflection, Warpsteam, etc. where incumbents can combine massive existing distribution with incredible technical founders.
Actions + Memory
I’m going to generalize my mobile productivity prediction to be more about 2025 being the year of actions and memory. Screen control, Pplx shopping, etc., are the early versions but I imagine 2025 will see an explosion on this front.
New Pricing Models
I and others have talked about this for a while, but it feels like the dam is starting to break. Especially if you’re a startup, it’s no longer sufficient to innovate on the core product, you have to think about radically different cost structures and revenue models to properly compete against the behemoths. I’m not sure if people coalesce around arbitrary compute units, work done, etc., but I think we’ll start to see a much larger variety. Ultimately, the end user will feel better because the price we pay will be better tied to the value we derive, but it may end up like the streaming wars where we swapped paying $100/month for a shitty product to pay $150/month for four much better products.
Web 4.0
I’m not sure if people fully appreciate how fundamentally different the web will be over the next decade.1 The web was originally organized to enable humans to click links to navigate and then read the content, which was structured by the author based on what they wanted the reader to understand. Search engines made this much easier, but didn’t change the fundamental structure.
The new world order will have some meaningful changes:
The core audience for a page will no longer be another person but instead an agent or application
As a result, the author is no longer the arbiter of what’s “important” on a page - models and agents are2
Navigation will need to change as the atomic unit shifts from a webpage to a piece of information (which could be an image, a video, etc.)
Will our notion of operating systems change to be more browser-first?
Websites will need to be restructured to fit this new optimization function
We’ll see more interactive, relevant forms of advertising than what’s currently forced upon us
The tooling required to build this new web will have to evolve as well
What excites me most is how this transforms the user experience from one where content punches us in the face to one of progressive disclosure, where curiosity reigns supreme, and where the end user’s opinion matters most.
I look forward to these all coming true in 2026.3
Yes, web3 people said the same thing, but something something early not wrong
Recipes are infinitely easier to read now.
Evidence for me being consistently off by one year with my comment on Iceberg from 2023