Mark Baldwin

For some people, the internet already died… Earlier this month I wrote about dead internet theory. Bots are now most of the web’s traffic, and about half of new articles are written by AI. I said people stop browsing because none of it feels real. Then I came across this post from @hridoyreh . “I can’t browse the web for answers anymore.” When people assume everything is AI, articles written by real people, or a real comment gets read as if a bot created it. If you work in marketing, that’s the part to sit with. Your content isn’t only competing with other brands anymore. It’s competing with the assumption that nobody wrote it. Last time I said relationships matter more now than ever. Here’s what I think that looks like in practice: 1. Specifics only you could know. A real number, a real customer, a real mistake 2. A name and a face behind it 3. Fewer pieces, each one worth reading

Wow. There's potential now for every internet connected device to be a major vulnerability. 2 factor everything that's important to you if you haven't already. Use a true authenticator type app like Authy or MS Authenticator.

Kind of crazy that it was doing all of this. Not surprising though. Ed Snowden told us about all of it years ago. I'm sure every device has this to some degree. All the back doors are there. So much data going out in so many ways. Imagine saying your password or passphrase out loud and this thing and other devices in your house are storing it and sending it to who knows where. Then those servers get hacked. The silver lining is that we'll be able to reverse engineer all of our devices and be private again. This could be a new industry in the near future.

Stop with the slop. It’s hurting your brand.
 Fact: In 2024, NielsenIQ had more than 2,000 people watch AI-made ads of varying quality. It measured brain activity on about 150 of them.

 Here’s what it found: 1. People spotted most of the AI ads and rated them “annoying,” “boring,” and “confusing.” 2. Even the AI ads rated high quality left weaker memory traces in brain scans. I see it myself.. My kids school creates AI images on social media, and you can tell the person who made them doesn’t understand graphic design. There are things in those images a graphic designer would never add or do. They’re excessive and over designed.
 Like a lot of people, I’ve gotten to the point where I gloss over anything I recognize as AI. Immediately. It’s something we already do with ads. It’s called “banner blindness.” Once we see an ad multiple times, we're immediately able to recognize it and ignore anything that looks like it… But AI itself isn’t the problem. When ad tester System1 showed Coca-Cola’s 2024 AI holiday ad to viewers who weren’t told it was AI, it scored 5.9 stars, the highest rating possible. Only 3% of viewers reacted with contempt. The penalty isn’t being AI. It’s for looking like AI.
 So how do you break through? Effort and best practices. And guess what? The people who were creating graphics before AI already know what best practice is. They’re the ones most likely to prompt AI and create something that doesn’t look like AI.
 AI keeps getting better at generating. But it’s up to brands to put it in the hands of experts who know what good looks like.

Bots are now most of the internet’s traffic… Most of what is online is no longer made or read by people. Dead internet theory used to be a conspiracy. In June, Cloudflare’s CEO said bots were 57% of web requests. Half of the new articles published online are written by AI. On LinkedIn it is 40% of long posts. You search. Google’s AI reads the pages, writes a summary at the top, and you never click. By July 2025 Google was crawling 5.4 pages for every visitor it sent back, up from 3.8 in January. Its referrals to news sites fell 15%. The people who wrote those pages got paid by visits. No visit, no ad revenue, no reason to write the next page. And now AI agents browse too. Traffic from agents that navigate and act on the web grew 7,851% in a year. They sit in your analytics looking like customers. So the incentive to publish for people is increasingly collapsing. Our feeds are filled with slop. People stop browsing because none of it feel real. Three things I take from this: Stop throwing AI slop grenades. Graphite found AI articles mostly do not get surfaced in Google or ChatGPT anyway. Relationships matter more now than ever. Only 26% of consumers prefer AI-made creator content now, down from 60% in 2023. We want connection, and we lose it chatting with AI responses and AI influencers. What if a second internet forms? Cloudflare already lets us charge crawlers per request. X tested a fee for new accounts... Proof-of-humanity networks are being built. A place with no bots, because the system won't let them in. Is the internet dying?

Selling and buying houses with AI saved me $58,000. AMA. Two houses in four months. The sale closed August 12. The purchase closed August 13. $21,000 is what an agent would have charged me to sell the home. $13,000 was a 2% rebate back to me to cover closing costs on the new one. $24,900 was the gap between list price and what I paid. For the sale… a $99 MLS service, $199 for professional photos, $2,500 for an attorney to review contracts. AI covered the rest. Pricing… it pulled apart the RVM valuation and the comps under it. I listed at $849K. Offers… it scored each one on deposit, loan-to-value, lender, and how real the pre-approval was, and helped me send counteroffers to agents who do this every day. Contracts… on the second deal it caught a typo from the buyer’s agent that would have sent my money to the wrong brokerage. $20K mistake avoided. That deal fell through anyway. Two deals died on financing. AI worked out that buyers at my price could not get the loan, and why. The third buyer could. It managed every contract date and flagged things early so I could stay on top of my buyer’s agent and the sellers’ agent on my new home. For the purchase… I hired an AI real estate agent. They automated the buying process, which is how I got the rebate. I used AI to research neighborhoods and set the offer price from South Florida. The RVM said $685K. My AI research said $640K to $660K based on comps I identified. List was $674,900. I offered $640K and paid $650K. It appraised for more. My broker offered a bridge loan at 8.5% plus a point. AI said ask the sellers to move closing by 10 days. They did. No bridge loan. Where did AI not work? Counteroffer emails were too robotic. They seemed to favor the buyers over me. On pricing it wanted to cut rather than hold at market. On the buy side it recommended lowball offers too often. We ended up with significant monthly savings, closer to family, in a newer, more comfortable home. Ask me anything!

For my workflow Fable 5 is doing good work. I get the impression that output quality is better and faster too.

Anyone care to share their security hardening prompts they're using to have Mythos - Fable 5 audit/check their systems?

Opinion: AI Hype is real and fake at the same time.

Why don’t we hear more about how schools are planning to adjust their curriculum in response to AI?

If Linux doesn’t overtake Windows as the #1 OS in the age of AI coding, we’re not a serious civilization.

I'm loving n8n and Claude. Excited about what you're able to build and automate.