- The Prohuman
- Posts
- OpenAI’s goblin problem was real
OpenAI’s goblin problem was real
Plus: Google adds 25M paid subscriptions
Rimsha Bhardwaj, Hasan Toor & Ihtesham Haider
April 30, 2026
In partnership with
Hello, Prohuman
Today, we will talk about these stories:
The strange reward behind AI tics
Google’s subscriptions keep climbing
AWS is spending hard on AI
Stop making AI decisions in the dark. Understand AI usage.
Leadership is asking: are we getting value from AI? Which tools are worth the spend? Where are we exposed? Right now, most teams have no idea.
Harmonic Security Usage Explorer changes that. It automatically classifies every AI interaction across your organization into the use cases driving real work, specific to your business. Not generic categories. Not raw prompts. Actual patterns to understand: how your teams are using AI, how much time they spend in AI, the cost, and where risk lives.
CIOs get the data to rationalize spend and cut wasted licenses. CISOs get risk in context. AI committees get proof of impact.
Early access is now open to a limited number of organizations. Request your spot.
OpenAI found the goblins in the rewards
Image Credits: OpenAI
A tiny word choice turned into a model behavior problem.
OpenAI says GPT-5.1 started using “goblin” 175% more after launch, with “gremlin” up 52%. The trail led to its old Nerdy personality, which made up 2.5% of ChatGPT responses but 66.7% of goblin mentions.
This is the part that matters. A style reward meant to make one mode playful ended up teaching the model a habit that spread outside that mode.
I think OpenAI is right to treat this as more than a funny bug, because it shows how small rewards can create durable language tics across training runs. The physical detail is almost silly: employees kept seeing the word pop up in Codex, like a sticky note left on the screen.
The fix was direct: retire Nerdy, remove the reward signal, and filter creature words from training data. The harder question is how many less obvious habits are already being rewarded without anyone noticing.
Google’s subscription push is working
Google added 25 million paid subscriptions in one quarter.
Alphabet says it now has 350 million paid subscriptions across its services, up from 325 million in Q4 2025. The growth came mainly from YouTube and Google One, where advanced Gemini features are now part of the bundle.
The awkward part is YouTube. Its ad revenue still grew 11% year over year, but $9.88 billion missed Wall Street’s $9.99 billion expectation.
I think this is less a YouTube weakness than a business model shift becoming visible in the numbers. A viewer who taps into Premium on a phone at night may be more valuable over time, even if that choice trims ad revenue today.
Google is also keeping Gemini numbers vague, which says something. Paid enterprise Gemini usage rose 40% quarter over quarter, but without a base number, it is hard to know how meaningful that really is.
Subscriptions are giving Google another growth lane, but investors still need proof that AI is pulling its own weight.
AWS growth now comes with a cash squeeze
Amazon’s AI boom is showing up in both revenue and bills.
AWS sales rose 28% year over year to $37.6 billion, its fastest growth rate in 15 quarters, according to Amazon CEO Andy Jassy. He said AWS now has an AI revenue run rate above $15 billion, compared with AWS’s $58 million run rate three years after launch.
The catch is physical. Amazon has to buy land, power, buildings, chips, servers, and networking gear before it can turn that demand into revenue.
I think this is the clearest version of the AI infrastructure tradeoff right now. The companies selling compute are winning, but they are also locking themselves into years of heavy spending.
That is already hitting cash flow. Amazon said trailing twelve-month free cash flow fell to $1.2 billion from $25.9 billion a year earlier, driven largely by a $59.3 billion increase in property and equipment purchases.
Investors may accept that while AWS is accelerating. The question is what happens if AI demand cools before the bills do.
Prohuman team
Covers emerging technology, AI models, and the people building the next layer of the internet. |  Founder |
Writes about how new interfaces, reasoning models, and automation are reshaping human work. |  Founder |
Free Guides
Explore our free guides and products to get into AI and master it.
All of them are free to access and would stay free for you.
Feeling generous?
You know someone who loves breakthroughs as much as you do.
Share The Prohuman it’s how smart people stay one update ahead.