The Internet Just Changed Forever
AI used to be an information play. Now we're entering anything goes territory
Cover image is a screenshot from Total Pixel Space, an award-winning AI video.
It could well be the defining moment in AI this year has already happend on May 30th 2025. That was when Google released of Veo 3.
Take a look at some snippets generated with Veo 3:
A “prompt” is a natural language instruction sent to an AI model. All characters in the video above are AI-generated from user prompts—meaning they don’t exist anywhere outside of the digital realm.
The quality of facial expressions and the perfectly synced and generated speech are better than anything I’ve ever seen—including the snippets from Omnihuman-1 generated video released by an AI research team at Bytedance back in February.
Veo 3 is already available in 72 countries—including the US.
AI video will redefine how businesses advertise online
I was talking to a marketeer at a conference last Thursday.
She walked me through her current flow for generating B2B advertising shorts.
It consisted of writing a script, hiring an actor on Fiverr to act in the short, have the actor record the video, and publishing it online to run as a paid ad.
Based on my own experience, this process typically takes at least 24 hours due to the delays involved in finding and working with the actor via Fiverr.
And each time she needs a new video she needs to hire an actor.
With Veo 3, she’ll be able to get a new ad done in 30 minutes or so.
Which means she can run practically unlimited experiments with different short-form videos, using ad platform algorithms to find the best performing ads for her funnel.
This will result in much better ROI on ad spend, and allows her to quickly jump on trends and events in ways that wouldn’t have made financial or practical sense before.
In addition, tools like Pippit AI for generating product videos and property edits AI for listing videos make it easier than ever to create dynamic commercial content.
This is great, but are there any downsides to going all-in on AI generated content?
Social search is mostly video search
Besides being the dominant medium—most time online is spent on video content—people also generally trust video content more than other forms of communication.
Before Veo 3 it was a lot easier for us humans to distinguish UGC (user generated content) from CGC (company generated content) in video than it was in text.
We had more cues to go by, and most humans are pretty sharp pattern recognisers—even though we wouldn’t always be able to pin-point exactly why something is fishy.
Now that the cost of producing realistic videos is dropping exponentially, this could change. We’ll likely see a blurring of video content produced by amateurs, prosumers and businesses, making them indistinguishable from one another.
At the same time we’re seeing the decline of traditional search due to generative AI.
Users are moving from Googling to asking their AI assistants, and keyword search and ranking on Google are declining in importance in the buyer journey.
Gen Z in particular is very passive in its information-seeking behaviour. And when searching actively, they will be more likely to search on a single video platform like TikTok, Instagram or Youtube rather than in the Google search bar.
Both of these trends combined mean we’re looking at some of biggest changes in the advertising landscape to happen in the last 20 years.
At this point it’s hard to say what the direct impact on consumer behaviour will be.
Will they start to shun anything even remotely commercial?
Or will consumers come to accept that everything is advertising, that the online world is a liminal space in which user intent and commercial interests coexist in blurring boundaries separated by mere pixels?
If anything, this is already the case today—the majority of revenues in the creator economy are from corporate sponsorships. Advertising is already running along blurry lines for human creators.
A good bet is that channel or creator brands will become more even important relative to the platforms and algorithms on which they publish their content.
That users will start scanning for authenticity and integrity signals from creators when searching for content online.
It would be funny if these human values rise in prominence due to advances in AI.
Authentic vs artificial: two sides of the same coin
If that happens, Youtube, Tiktok and Instagram discovery algorithms will probably be changed to work more like those of Substack. Content discovery would be weighted on the entity publishing it, and subscriptions could become a more viable business model as humans seek to distinguish information from commercial interests.
So on the one hand, the combined forces of AI video and AI search could give a big boost to the creator economy—as people flock to online entities they trust.
On the other, it’s now easier than ever to game the system with fake reviews generated and published at scale either on video platforms or on marketplaces.
So where does that leave businesses?
One place to look at is at partnerships with creators.
In this changing landscape, a review by a trusted creator can do more for your business than any AI generated video you publish yourself or a marketing agency creates for you.
Another very important thing is to remain authentic in your online communications.
Define your brand values and learn to speak the language your customers, audience or community speaks. Engage in conversations and interact with them on their terms.
Become part of their digital world rather than a distraction from it.
And of course in the short-term—because the trends described in this article will pan out over the next several years—take advantage of the amazing possibilities offered by tools like Veo 3, Google’s Flow or Kling AI and create awesome content with AI.
Blur the lines between marketing, creative direction and community management.
It’s a fun time to be a marketeer.
A more philosophical take on the whole AI video thing is given in one of the finalist of the RunwayML 2025 AI video festival, Total Pixel Space:
Last week in AI
The AI-native code editor Cursor officially launched its 1.0 version, introducing a suite of new features. One of the key additions is "BugBot," an AI agent that automatically reviews pull requests for bugs and suggests fixes. The release also includes the ability to use AI agents within Jupyter Notebooks, create "memories" for projects, and render visualizations like Mermaid diagrams directly in the chat.
OpenAI's Codex can now access the internet to install dependencies and upgrade packages. This new feature is disabled by default but can be enabled by users. This was a highly requested feature that allows for more dynamic workflows.
A new beta feature in ChatGPT called Connectors allows the AI to securely link with third-party applications like Google Drive, GitHub, and SharePoint. This integration lets you search files, reference content, and pull live data from your own sources directly within a conversation. The goal is to make ChatGPT more personalized and actionable by grounding its responses in your specific data.
Anthropic enhanced its 'Projects' feature by integrating Retrieval-Augmented Generation (RAG), allowing users to ground Claude's responses in their own private data. Users on paid plans can now upload documents like PDFs and DOCX files to create a dedicated knowledge base for a specific project. This enables Claude to answer questions and perform tasks based on the provided information, making it a powerful tool for analyzing internal documents.
Japanese AI lab Sakana AI introduced a nature-inspired approach to creating AI agents that like The Entity from Mission Impossible—The Final Reckoning are able to continuously self-improve and rewrite their own code—keep an eye out for these Darwin-Gödel Machines.