Karsten Weide

Here are @WMediaResearch's Top 10 Trends in Advertising in 2026! 1. The Big Short: The Hidden Risks Lurking Beneath the Forecasts 2. There Will Be Blood: The Supply-Chain Shakeup 3. Honey, I Shrunk the Art Department: Gen-AI Taking Over 4. Advertising Wars: The Rise of AI Agents 5. Rules of Engagement: The Protocols Fueling the Agentic Era 6. No Country for Old Cookies: The New Identity Playbook 7. Measure for Measure: Can We Keep Up With Signal Loss? 8. The Verdict: The Post-Google Era Begins for the Open Web 9. The Day After Tomorrow: The Zero-Click Search Apocalypse 10. I, Agent: When Bots Buy - What Happens to Advertising? READ all 10 trends for FREE here: Everything Everywhere All at Once: The Advertising Revolution Of 2026 wmediaresearch.com/2025/12/01/the… Companies covered: @Adobe, @AdRoll, @Albertsons, @amazon, @AppLovin, @criteo, @Disney, @doubleverify, @epsilon, @Google, @humansecurity, @IABTechLab, @IndexExchange, @integralads, @Meta, @Microsoft, @MSAdvertising, @Novacap_HQ, @OpenX, @prebidorg, @PubMatic, @Rockerbox, @Sincera, @TheTradeDesk, @TikTok, @Yahoo.

A lot of talk a lot how it might never be possible to let (LLM-based) AI agents be completely autonomous, especially in high-risk areas such as law, finance, and medicine. The implications for advertising are that a) advertisers are right to set up guardrails and to keep a high degree of human oversight at this early stage, and b) that there will always be some degree of human oversight.

@goldenboygym My heart goes out to you. The worst thing that can ever happen to a human being is losing a child. ❤️

今日は11歳でこの世を去った息子の誕生日です。 生きていれば22歳、青春の真っ只中だったでしょう。 人の記憶に残る間は、その人は生きていると言いますし、私と一緒に心の中で祝ってやってください。 そして、改めて家族の、親子の素晴らしさを思うきっかけになれば幸いです。

@Seanfrank Detroit II.

I love LA. But if it wasnt for the weather, we would be the poorest city in the country. Im legitimately worried the city cant be saved. Hollywood is shedding jobs. The people mover is sitting idle at LAX. Every neighborhood has empty storefronts. 1 house has been rebuilt in the palisades. Covid was 6 years ago. We cant keep blaming covid. We cant blame trump when every single office in the city, county, state is one party. I pay a lot of taxes to stay in this city. Hundreds of thousands of dollars or more. I dont want to run, but at a certain point its just logical to cut your losses. Population declines. Billions in fraud. Schools are now some of the worst in the country, being beat by Mississippi. Good thing it is fucking sunny, or else this place would fall into the ocean.

@AtwoodIan32412 Correct. Here are more ideas for TTD: wmediaresearch.com/portfolio/a-te…

If OpenAI goes through $TTD, Trade Desk gets a new premium supply source at the exact moment its CTV growth is decelerating and open-web display is under structural pressure from AI Overviews. That's not a ChatGPT story, but that's one hell of a $TTD story.

W Media Research is celebrating its 3 year anniversary! As a Thank You for our clients and followers, we’re rolling out a special offer: all research documents are just $500 - for a limited time only! For instance: Whitepaper: When Bots Buy – What Happens To Advertising? wmediaresearch.com/downloads/whit… If you’ve been thinking about leveling up your AdTech insights, now’s the time!

Godspeed, Artemis II!

Two hours and twenty four minutes to go!

@Myles_Younger @FantasyEpics 😉

@KarstenW @FantasyEpics deep cut excellent

Game of Thrones really gave Emilia Clarke nothing after 10 years… 💀

@CleansedTweets We are.

How do people who lived through the 80s and 90s stomach today. I can’t fathom it. Surely you must be depressed?

Gigantic AI hangover for just about everyone in 3, 2, 1…

W Media Research is celebrating its 3 year anniversary! As a Thank You for our clients and followers, we’re rolling out a special offer: all research documents are just $500 - for a limited time only! For instance: How AI Search Threatens Pubs And Vendors – And What To Do About It wmediaresearch.com/downloads/how-… If you’ve been thinking about leveling up your AdTech insights, now’s the time!

But, but… whatever happened to intergalactic super-intelligence? 🥺

@MaMoMVPY Let alone progress towards AGI.

I completely agree - the plateau has been pretty clear for 6-8 months. There really isn't any progress on the core LLM models. Where there is progress is where the models are used and how they are used, and here it is really only Claude that is getting better. But even Claude is not progressing in a straight line - the non-English use of Claude is certainly not getting better. For example, Danish or Icelandic use of Claude is worse today than six months ago. We are clearly in the overfitting stage. The models are not improving anymore, and more and more training content is LLM-generated as more and more available texts out there are fully or partly written with LLMs (we all do it - and I am happy to admit it). But we also know that training LLMs on LLM output will eventually lead to model rot. This seriously limits scaling. There are lots of productivity gains to be had from using LLMs, but it is now nearly 100% about implementation - not about getting better models. The technology itself might very well have stalled.

@gregorykennedy @bprintco 😆

@KarstenW @bprintco Saddest story I ever heard.

I listed a really nice leather couch for Free once on FB Marketplace because I was moving out of my old apartment and needed it gone in a couple hours. Probably worth $1200. This small lady showed up with no help. I had to carry it down to her car. It was a small SUV, you could barely put one cushion in this thing. She asked if I had a truck and could drive it to her house. I did have a truck, and reluctantly agreed. Nobody was there to help unload it either so she asked me to unload it. I took it out of my truck and took it to her porch and left. She left me a bad review because I didn’t bring it inside her house for her. I should have just tossed it in the dumpster.

@AtwoodIan32412 They don’t have much of a choice this early on.

OpenAI talked to $TTD about selling ChatGPT inventory through programmatic rails. The moment this hits a DSP, the $60 CPM scarcity premium evaporates. Every premium ad surface in history has followed the same arc: managed --> programmatic --> compressed. (see: $NFLX).

@Rahatcodes In another year, when encountering grumpy users, it’ll run to its virtual safe space, set SULK=TRUE and won’t interact with those users anymore.

Claude Code has a regex that detects "wtf", "ffs", "piece of shit", "fuck you", "this sucks" etc. It doesn't change behavior...it just silently logs is_negative: true to analytics. Anthropic is tracking how often you rage at your AI Do with this information what you will

The one major general criticism I have is that they lumped in search and social with display and video, which doesn’t make much sense. (And is also not how most people use the term “omnichannel”.)

@verklin @realroseceline Correct. But as a buyer you never know if the walled garden favored their own inventory over the open Internet one for no other reason but financial ones.

@realroseceline $AMZN and $GOOG have their own DSPs that also buy inventory from the Open Internet

My thoughts on $TTD I think $TTD is a good value here. The main reason is simple. Most people don’t actually understand what the business does, so they end up arguing about the moat, growth and multiples without understanding what it really is. $TTD doesn’t own content, doesn’t have users, and doesn’t sell ads like $META or $GOOG. At first glance that looks like a disadvantage, but I actually think it’s the reason the business works. The open internet is extremely fragmented. Millions of sites, apps, streaming & audio platforms, no single owner or standard pricing. Without someone organizing that, ad dollars get spent inefficiently, and that’s exactly what $TTD does. They decide for the advertiser where the money goes in the so called “open internet”. They don’t own anything, they just control the decision, and that ends up being a much more important position than people realize because the most valuable part of advertising isn’t the content, it’s deciding where budgets get allocated. That’s also why the big platforms don’t allow it. $META, $GOOG, and $AMZN keep everything inside their own systems because they want control of the dollars, so what you end up with is two different worlds, closed systems where everything is controlled internally, and the open internet where someone still has to decide how money flows, and at scale that’s $TTD. There are other DSPs, but none with the same scale, data, relationships, and infrastructure, and in this kind of business scale compounds. More data leads to better decisions, better decisions lead to better ROAS, and better returns lead to more spend, so the system reinforces itself over time. It’s a flywheel similar to $MELI. This is also why take rate is less important. ROI is what matters, because if someone is cheaper but performs worse they lose, and if $TTD is more expensive but makes advertisers more money they win, and that’s why performance is so hard to compete with when you have the most data. The economics reflect that because $TDD is an asset light business doing over 40% EBITDA margins, with strong cash flow and a very clean balance sheet, and on top of that you have the CEO buying around $150m of stock in the open market. At the current price the market is treating this like a normal ad tech business with slowing growth, but if they keep this position as the default way advertisers allocate money across the open internet, then even modest growth combined with these margins can compound into something much more valuable over time, and that’s really the part I think is being underestimated. And lastly yes there are issues, if there weren’t $TTD would still be trading at 20–30x sales like it did before. SBC is high and growth has slowed, and there is always the risk that more ad dollars move into closed systems where $TTD has no control. But on the other side retention is extremely strong (over 95% for many years), and there isn’t real competition at scale outside the walled gardens, so once advertisers are on the platform and getting results they don’t leave, and that dynamic is a big part of why the model works despite the concerns, at least in my humble opinion. 🌹