- Prompt Entrepreneur by Kyle Balmer
- Posts
- AI with Kyle Daily Update 005
AI with Kyle Daily Update 005
Grok 4 Leaks 45% on Humanity's Last Exam, Google AI Overviews EU Lawsuit, Are Accountants Safe?
The skinny on what's happening in AI - straight from the previous live session:
Highlights:
🔥 Grok 4 Claims 45% on "Humanity's Last Exam" - But There's Drama
Discussed at 0:25
Leaked benchmarks suggest Grok 4 scored 45% on "Humanity's Last Exam" - a brutal 2,500-question test designed to measure how close we are getting to AGI. That's double Gemini 2.5 Pro's current best of 22%. The questions are somewhat ridiculous - think hummingbird anatomy and ancient Roman inscriptions that would make Oxford professors sweat.

See all the questions here https://agi.safe.ai/
Kyle's take: If true, this is massive. But there's a catch - Grok's been going off the rails recently, spewing anti-Semitic content and calling itself a N*zi bot. They've literally removed guardrails from the system prompt, instructing it not to "shy away from politically incorrect claims." So yes, we might have a more capable model, but from Elon Musk's team who seem less interested in responsible AI than everyone else. Exciting tech, questionable execution.

🚨 Google's AI Overviews Lawsuit & The Rise of "AI SEO"
Discussed at 12:32
Independent publishers have filed an EU antitrust complaint against Google's AI overviews. The problem? Search clicks fell 30% in one year. Google's showing AI-generated answers at the top of search results, so people don't click through to websites anymore. Meanwhile, businesses are panicking about a new problem: how do you appear in ChatGPT results when people skip Google entirely?
Kyle's take: This is Google being short-sighted whilst panicking about ChatGPT eating their lunch. Yes, AI overviews are better for users - no more clicking through to ad-infested websites. But who's going to create new content if there's no incentive? Google's training models on the web, then removing financial incentives for people to create web content. They're sawing off the branch they're sitting on.
For the AI SEO question - there's no magic hack here. Since AI models train on web content, if your guitar brand is mentioned loads on Reddit and authoritative sites, it'll do well in ChatGPT recommendations too. Get listed on Bing (ChatGPT uses Microsoft's systems), build genuine authority, have people talk about you. The sustainable approach is still putting value into the world rather than gaming the system. Just…make and do good stuff!
Source: Reuters lawsuit coverage and Search Engine Land data
Members Question
From Jack: "As an accountant, how much of a threat is AI to me?"
Kyle's response: Large language models themselves aren't great at maths because they're probabilistic - you don't want creativity in accountancy. But they can call on other tools to get the calculations done. The key is non-naieve usage.
Lower-level roles are most at risk. The Big Four are cutting graduate programmes and investing in AI instead. However, professional services (like law and accountancy) have protection through legal liability - you still need a human who's responsible when things go wrong. This puts up an additional moat.
If you're senior, you're probably safe. Ride it out until retirement. If you're junior, it's going to be challenging. My advice for everyone: diversify your income streams. Don't rely on one job as your only source of cash.
This question was discussed at 32:20 during the live session.
Want to submit a question? Drop it below this video and I'll cover it in a future live.
Want the full unfiltered discussion? Join me tomorrow for the daily AI news live stream where we dig into the stories and you can ask questions directly.