The week of February 3rd, on a single day, OpenAI launched a new enterprise agent platform, Anthropic dropped Claude Opus 4.6, and Kling AI released the most capable video generation model we've seen. Same day. It was one of the biggest weeks in AI since ChatGPT launched.

I know what you're thinking. "Great, another week where I need to learn 14 new tools while also running my brand."

After a half dozen conversations since then, I want to share something that will hopefully be comforting: the pace of AI isn't something to fear. It's the single best thing happening for lean DTC teams right now.

Let me show you two reasons why: reduced AI costs, and the speed of new AI tools.

That’s good for everyone, as it means more capabilities for everyone without paying an arm and a toe.

Three numbers tell the story:

At this rate, by 2028 AI will be paying us to use it (probably not because of other scarcity issues like power).

Stanford's AI Index put a finer point on it: the cost to query a model at GPT-3.5 performance dropped from $20 to $0.07 per million tokens.

AI has gotten 280x cheaper in 18 months.

Across all benchmarks, the median price drop accelerated to 200x per year after January 2024.

Why cheaper models matter for your skincare brand or supplement company

Every AI tool you pay for is built on these models. When the underlying intelligence drops 200x per year, every tool gets cheaper, faster, and more capable. That $99/month email copywriting agent? It'll either cost $10/month next year or be 10x better at the same price. Probably both.

But it’s not just about the cost of AI getting cheaper.

On the one hand, this sounds super intimidating for a lot of people. It’s actually a good thing. But let’s talk a little about why the rate of new AI tools will increase, and what that means.

AI is building AI. That's why this keeps accelerating. This is also the primary reason we’re going to see the pace increasing.

OpenAI's newest coding model, GPT-5.3-Codex, helped build itself. Anthropic did the same with Claude Code and Opus 4.6. Their team used early versions of the model to debug its own training runs, manage deployment, and diagnose test results. The AI coding model is now good enough that the people building it used it to build the next version of itself. That's a feedback loop, and it's the single biggest reason the jumps between model generations keep getting larger. Veo2 to Veo3 in six months. Kling 2.6 to Kling 3.0. Each generation makes the next one faster to build.

You can see this playing out in real data. AI coding agents went from solving 4.4% to 82% of real engineering tasks. Code pushes to GitHub surged 41% in a single year. Stripe reported that the 2025 cohort of startups on their platform grew 50% faster than 2024's, and the number hitting $10M ARR within 3 months of launch doubled. People are shipping faster because AI is doing more of the building.

Right now, development and coding has the strongest AI product-market fit. But here's the corollary that matters for DTC brands: what happened in coding over the past 18 months is expanding into every other function.

Creative production is following the same curve (Kling 3.0 and Ideogram Character are proof). Analysis is right behind it (models now score 93% on PhD-level science questions, up from 39% two years ago). Copywriting, ad testing, customer support, reporting... all of it is on the same trajectory that coding was 12 months ago.

The same feedback loop that lets AI build better AI is going to produce better marketing tools, better creative tools, better analytics. And it's going to happen faster than most people expect because each improvement compounds on the last.

And I’ll tell you what… it’s definitely better than people expect since we had this working at Raleon 12 months ago.

"Doesn't this mean agencies are dead and my job is toast?"

No. I wrote about this in my 2026 predictions newsletter. The brands and agencies that embrace this will win.

If the cost of producing creative heads toward zero, creative doesn't stop mattering. It matters more. You can test 60 ad variations instead of 12. Kill losers after $500 in spend instead of $1,000. Iterate daily instead of monthly. The strategic thinking, the brand instinct, the customer understanding? That gets MORE valuable because now you can actually act on it at speed.

I watched a supplement brand go from testing 12 ad creatives per month (bottlenecked by their agency) to testing 60. Same budget. Better ROAS. They identified winners faster and doubled down before the algorithm killed their spend.

The agencies that struggle will be the ones charging $15K/month for work AI handles in an afternoon. The ones that thrive will use AI to deliver better results at lower cost, or more at the same cost. That's not doom. That's how every technology shift plays out.

Pick one:

1. Audit your AI tool costs. On plans from 6+ months ago? Prices have dropped. Competitors have launched. I guarantee you're overpaying somewhere. 20 minutes of checking could save $100+/month.

2. Test a second AI model. Only using ChatGPT? Try Claude for a week on your hardest task (I find it better for copywriting and analysis). The quality gap between models is nearly gone.

3. Run this prompt in your preferred chatbot:

I ran this for five brands last month. Every single one found at least one tool they didn't know about that saved 3+ hours per week. The AI is good enough now that asking it to solve your problems actually works. That wasn't true 12 months ago.

4. Reframe your mindset. Stop thinking "I need to keep up with every launch." Think: "What's my most painful, repetitive task? Is there an AI tool that handles it now that didn't exist 6 months ago?" The answer is almost certainly yes.

And that's it for this week's edition. AI is building AI now. That feedback loop is why every tool you use is getting better faster than you expect, and why what happened to coding is about to happen to creative, analysis, and every other function in your business. The gap between "enterprise AI budget" and "lean DTC team budget" is disappearing. Go test something this weekend.