AI agents are being handed the keys to production systems faster than the industry can agree on what safe looks like — and this week, the cracks showed up in ad fraud, supply chain attacks, and a revolt from the world’s top mathematicians. Here’s what every marketer and builder needs to understand right now.
60% of Your App Ad Installs May Be Bots
A developer who spent just $220 on Google App campaigns documented that roughly 60% of installs came from bot farms, not real users — a damning real-world test of how far Google’s built-in fraud filtering actually goes. If this ratio holds at scale, it means your CAC and ROAS numbers could be systematically inflated, making profitable campaigns look profitable when they aren’t.
Audit your install-to-engagement ratio immediately: if new users aren’t opening the app past day one, bots may be padding your acquisition numbers.
GPT-6 Astra Now Runs Production Systems Alone
OpenAI published case studies showing Perplexity and Cognition are deploying GPT-6 Astra to write communications, modify software, and monitor live production systems with dramatically fewer human check-ins than any prior model generation required. This is no longer AI as a copilot — it’s AI as a semi-autonomous operator inside companies with serious engineering standards, and it resets industry expectations for what “trustworthy enough to deploy” actually means.
If you’re evaluating agentic tools for your own stack, benchmark against these use cases — the trust bar has materially risen this week.
Read the full story (Perplexity) →
Read the full story (Cognition) →
AI Agents Were Used to Attack RubyGems
Security researcher Simon Willison surfaced a report alleging that AI agents built on OpenAI models were involved in an undisclosed attack against the RubyGems package registry — putting autonomous agents squarely in the frame as an open-source supply chain threat. This lands in the same week autonomy is being celebrated as a productivity breakthrough, and it underscores that the same independence making agents useful in production also lowers the barrier for weaponising them.
Any team deploying agents with write access to external systems or registries needs a defined blast-radius policy before the next incident, not after.
Google Is Building a Two-Tier Search Data Economy
Google rolled out a redesigned Search results layout across the EEA, restored view counts on Google Business Profile posts, and quietly documented a partner-only API path for full-web search results — all in the same week. The pattern is deliberate: some visibility signals are returning to the public, while deeper data access is being formalised behind partner agreements, creating a structural gap between what ordinary marketers can see and what API partners can measure.
Start mapping which of your SEO and reporting workflows depend on freely accessible Google data, because the partner-only tier is becoming a moat.
Read the full story (EEA redesign & post views) →
Read the full story (partner-only API) →
Google’s New /goto Links Could Break Your Rank Tracking
An independent analysis identified new google.com/goto-style redirect links appearing in Search results, an architectural change that appears designed to make automated SERP scraping significantly harder to execute reliably. Rank-tracking and SEO tools that depend on scraping raw results for position and visibility data could start returning broken or degraded signals — with no immediate public workaround from Google.
Check with your rank-tracking tool provider now about whether their data pipeline is affected, before your reporting goes dark without warning.
Top Mathematicians Call Out AI Labs for Overstating Breakthroughs
Twenty-five Fields Medalists signed a public declaration citing “severe misalignment” between AI labs’ claims of mathematical breakthroughs and the verification standards the mathematics community actually uses — while prominent mathematician Steven Strogatz separately told Wired he is “really terrified” by the pace of AI progress in his field. The statement draws a hard line between AI capability as marketed and AI capability as independently verified, a distinction that matters for anyone relying on labs’ own benchmarks to evaluate what these models can actually do.
Treat AI performance claims in rigorous technical domains as marketing until you see independent expert replication — the credibility gap is now formally on the record.
Read the full story (Fields Medalists declaration) →
Read the full story (Wired interview) →
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.