- AI with Kyle
- Posts
- AI with Kyle Daily Update 028
AI with Kyle Daily Update 028
Today in AI: GPT-5 Rollout Chaos Continues, Musk Sues Apple, Google Teases Gemini 3 with Star Wars Meme
Kyle Balmer
August 13, 2025
Live Webinar: Learn how you can become an AI Trainer… get paid $1-$4k per hour …without needing a PHD in AI, or being a TED Talk level public speaker… 📅 Thursday ⏰ Check link for timezone 👉️ Register here
The skinny on what's happening in AI - straight from the previous live session:
Highlights
🎭 GPT-5 Gets Personality Transplants
OpenAI's added personality options to GPT-5 - you can now choose between Default (cheerful), Cynic (sarcastic), Robot (blunt), Listener (supportive), or Nerd (enthusiastic). The "Listener" mode is supposedly closest to the beloved GPT-4o that everyone's missing. They've also brought back legacy model access for Plus and Pro users after the massive backlash.
The team did a crisis AMA on Reddit the day after launch - Sam Altman even admitted they "screwed up the bar charts" in their presentation because people were working late and tired. The thing that annoyed me the most… 🤣
Worth a read for the honest behind-the-scenes chaos - and it’s all be sorted and filtered nicely by the Moderators.
Kyle's take: This is damage control. They've realised that killing off 4o without warning proper upset their users who had emotional connections to that model. We live in a fascinating world eh?
The personality feature is smart - gives people choice without admitting they mucked up the default experience. And it’s basically just some pre-rolled custom instructions, nothing complex. This plus rolling back 4o for paying users will appease to a certain extent.
But some people will be looking for an alternative after the emotional damage OpenAI caused over the weekend.
We do need to remember though that we are in a little bubble of people who really care about this stuff. MOST of ChatGPT’s 700m users couldn’t give a hoot.
Source: Reddit AMA
🎯 Google Drops Star Wars Hint at Gemini 3
A Google DeepMind researcher posted X-Wing fighters attacking the Death Star in response to Sam Altman's earlier Death Star post. The image had an "AI" watermark suggesting it (maybe!) came from Google's new image generator. Plus someone spotted "Gemini Beta 3.0 Pro" code commits on GitHub, though Google's denying the naming convention (via Logan KilPatrick specifically who has become their defacto mouthpiece.)
Kyle's take: Google's timing would be amazing if they can actually deliver a new model in the wake of ChatGPT’s fiasco.
Right when everyone's annoyed with GPT-5's routing problems + personality shift dropping a proper Gemini 3 competitor would be brilliant. The Star Wars reference is clever too - positioning themselves as the scrappy rebels versus OpenAI's evil empire. Though it's a bit rich coming from Google, hardly a small company themselves! We'll see if they can actually execute or if this is just more corporate trolling.
Also…why did Sam Altman choose the Death Star? Famous for having a weak point that makes the whole thing blow up. 😅
⚖️ Musk Throws Legal Tantrum Over App Store Rankings
Elon's having a meltdown because Grok ranks 5th in the App Store while ChatGPT sits at number one.
He's now threatening to sue Apple for "antitrust violations," ignoring the year-old deal between Apple and OpenAI that everyone knew about. Sam Altman fired back, calling out Elon's own manipulation of X's algorithm.
Kyle's take: This feels like Elon being desperate for attention while GPT-5 dominates the news cycle.
Fifth place for a relatively new AI app is pretty great honestly. But it’s not #1.
The real issue here is that despite all the hype, very few people actually use Grok as their daily AI… ChatGPT remains the “Google of AI”.
Source: Business Insider coverage (No Paywall)
🧠 Claude Finally Gets Memory (Sort Of)
Anthropic rolled out a memory function for Claude yesterday, but it's not quite the persistent memory that makes ChatGPT so sticky. Instead, Claude searches your past chats when you ask it to, rather than building an ongoing profile about you. It's rolling out to Team and Enterprise tiers first.
Kyle's take: This feels like Anthropic being overly cautious again.
ChatGPT's persistent memory is brilliant - it remembers you have a gluten intolerence, what business you run, where you travelled last Summer. Everything.
That creates massive lock-in because switching to another AI means losing all that context. It’s a “sticky” product.
Anthropic's approach feels privacy-friendly but less useful. It's classic Anthropic - letting their ethics get in the way of building what users actually want. I worry they'll get bought by Amazon before they can properly compete…and then all the ethical building will be for nowt.
Source: Verge coverage
UK's tax authority has admitted they're using AI to scan social media posts looking for people spending beyond their declared means. That expensive holiday or new car you posted about? They're watching. They claim it's only for "criminal investigations" but the scope could easily expand with new AI capabilities.
Kyle's take: This was always going to happen. We've had algorithms doing this for years, but now with generative AI, it becomes much faster and cheaper to deploy at scale.
The scary bit isn't that they're doing it - it's how easy it would be to expand this to everyone's social media. If you're posting about lifestyle that doesn't match your declared income, maybe think twice. Though honestly, if you're dodging serious tax, you've got bigger problems than Instagram posts…
Source: Telegraph coverage (No Paywall)
Member Question from Stefan: "Which model should I get if I only want one subscription?"
Kyle's response: Depends what you're doing! For coding and you're serious about it, Claude Pro at $200/month is superb - expensive but worth it if that's your living. For general writing AI, regular Claude is brilliant. For a life assistant that remembers everything about you, ChatGPT wins because of that persistent memory.
But here's a clever shout from someone else on the live - Perplexity Pro at $20/month gives you access to ALL the premium models (GPT-5, Claude, Gemini, Grok) in one subscription. You lose the specific tools (ie. Projects, Advanced Voice Mode etc.) from each platform, but if you just need the models themselves, it’s a great deal.
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.