The people who control the platform are rewriting the rules for everyone built on top of it. Apple is cutting ad tech off inside Safari, while smart buyers are saving money by routing AI work to the right model instead of waiting for vendor price cuts. Today’s stories show that control over your costs and your data is shifting to whoever plans for it first.
Apple’s Safari Blocklist Cuts Publisher Ad Revenue
Apple’s latest software update started blocking ad tech companies that track browsing, cutting them off inside Safari without warning. So far it hits the largest players, but there is concern it could reach much of the ad tech stack. When a vendor stops working in Safari, bids on that traffic drop, and some publishers already see revenue dips while many are still working out how bad it is.
Break out your Safari revenue and bid rates this week so you know your exposure before the next change lands.
LegalOn Cuts Codex Costs 65% With Model Routing
LegalOn cut its estimated daily Codex costs by 65% while keeping development speed. It matched three different OpenAI models to different tasks and managed budgets deliberately. The saving came from routing work, not from a price cut, and most teams still send every task to one top model, so routine jobs get premium pricing.
Sort your AI tasks into routine and hard, then test a cheaper model on the routine ones.
Claude Haiku 5.5 Targets Cheap, Repeated Agent Calls
Anthropic released Claude Haiku 5.5, its smaller model, with lower rates aimed at agents that call a model many times per job. The sticker price is only part of the cost. Reasoning effort, growing context, and how often a task finishes correctly decide what a completed task really costs.
Compare models on cost per successfully completed task, not price per token.
Read the full story →
Join the discussion →
Google’s Spam Update Is Done, So Judge Your Rankings Now
Google finished rolling out its September 2026 spam update after nearly 14 days, longer than any of the three earlier spam updates this year. Rankings swing during a long rollout and early drops can reverse, but now that it is complete, the numbers are finally readable.
Compare your traffic and rankings from before and after the rollout before you change anything.
Read the full story →
Join the discussion →
A Four-Person Agency Made a Spoof Ad in Four Days
New agency OK Future, with four people, made Goodwipes’ spoof of OpenAI’s Astra commercial using generative AI tools. It went from concept to launch in four days, avoided a projected $700,000 production cost, and the social post drew more than 750,000 views on Reddit. The speed worked because the idea was timely, and the agency also described highs and lows in the process, so it was not frictionless.
Use AI production for timely ideas with a short shelf life, where speed matters more than polish.
Measure PR by AI Citations, Not Placement Volume
HubSpot, which sells an answer engine tracking tool, argues that a top-tier press placement does not automatically become a citation in an AI answer. Some publications and content types are cited at much higher rates than others. A PR team judged on placement count may be rewarded for coverage that never reaches buyers asking AI for vendor recommendations, though the claim comes from a vendor, so check it in your own category.
Ask AI tools the questions your buyers ask and note which outlets get cited before you pick PR targets.
Log What Your AI Used, Not Just What It Answered
An O’Reilly Radar essay introduces Orrery, a reference design for preserving the runtime details needed to reconstruct an AI system’s past decision. It argues exact reproduction is unrealistic because models are nondeterministic, but reconstruction is achievable if you record it at run time. If a customer-facing decision is questioned six months later, you would need to know which model endpoint, policy, retrieved data, tool contract, and configuration were live at that moment.
Start recording model version, retrieved data, and configuration alongside every AI output that touches a customer.
More from Rafal Reyzer
For deeper dives on AI and marketing strategy, visit my YouTube channel →
Rafal Reyzer
Hey there, welcome to my blog! I'm a full-time entrepreneur building two companies, a digital marketer, and a content creator with 10+ years of experience. I started RafalReyzer.com to provide you with great tools and strategies you can use to become a proficient digital marketer and achieve freedom through online creativity. My site is a one-stop shop for digital marketers, and content enthusiasts who want to be independent, earn more money, and create beautiful things. Explore my journey here, and don't forget to get in touch if you need help with digital marketing.