AI with Kyle

Highlights

🎭 OpenAI's Secret Model Switching Exposed

Discussed at [00:01:58]

OpenAI got caught red-handed switching users from GPT-4o to GPT-5 without telling them. Users on Reddit thought they were going mad (and others were calling them conspiracy nuts) - turns out they were right all along. When people tried having emotional or sensitive conversations with 4o, OpenAI secretly routed them to GPT-5 for "safety reasons" whilst keeping 4o displayed in the dropdown.

Kyle's take: This is bad. Especially because it is (again!) 4o related. The 4o fans are already mad because of the GPT-5 launch and this feels like sticking the knife back in.

They're gaslighting the very people who fought to get 4o back after it was killed off during the GPT-5 launch. Yes, they buried something about this safety-switching in a blog post from September, but come on - if people don't realise what you're doing, that's a messaging failure.

The conspiracy theorists actually right this time, which absolutely erodes trust. It’s a comms failure and OpenAI are once again in damage control mode.

Source: Nick Turley, OpenAI Head of Product (damage control statement)

💼 Accenture Axes 11,000 Jobs in AI Pivot

Discussed at [00:09:20]

Accenture's cutting 11,000 employees worldwide as part of an $865 million restructuring to become "AI-first". CEO Julie Sweet announced they're “exiting” workers who "cannot be retrained for AI roles on a compressed timeline". At the same time they've secured $6 billion in AI bookings and doubled their AI specialists to 77,000.

Kyle's take: Let's be real - a lot of CEOs are using AI as cover for cleaning house after over-hiring. It’s hard to separate actual AI impact from AI as a convenient excuse to cut staffing.

A bigger question though: if these 11,000 people genuinely couldn't learn basic prompting after years of training, what were they doing there in the first place? It’s a hard question but I think a fair one.

And we also need to view this in the wider context. Consulting firms like Accenture are facing an existential crisis. Why pay millions for PowerPoint strategies when AI can provide the same advisory services? Or 90% of the service at 1/10,000th of the cost?

Accenture are pivoting it looks like. That $6bn revenue from AI projects is significant - it’s 10% of total revenue. My bet is they want to pump that as fast as possible as traditional revenue sources dry up. The real threat isn't training their staff - it's that AI is eating their lunch.

Source: FT (No paywall)

📱 Apple's Secret AI Chat App "Veritas"

Discussed at [00:30:01]

Bloomberg reports Apple's developing an internal ChatGPT rival called "Veritas" to test features for a massive Siri overhaul coming March 2026. Tim Cook told staff Apple "must win in AI" and called it "the biggest transformation in decades". The app's being used to test personal data search and in-app actions.

Kyle's take: Classic Apple - showing up years late then claiming they'll revolutionise everything! AND maybe they will.

They're brilliant at being second (iPod, iPhone, iPad all came after competitors and refined/perfected the model), but AI's different. You can't secretly build massive data centres, and they've wasted years on rubbish like AI emojis whilst everyone else shipped actual products.

Their selling point could be on-device processing (edge AI) for privacy, but honestly? I bet most consumers don't give a toss where their AI runs as long as it works. Imagine having a slower, less accurate AI on your phone or being able to use the ChatGPT app (cloud based, fast, top-line model): which will you choose? Being on-device is not enough to change behaviours.

Source: Bloomberg / The Verge

🏢 OpenDoor CEO: "Default to AI or You're Out"

Discussed at [00:16:27]

The new CEO of OpenDoor sent a brutal memo: every employee's first job requirement is now "default to AI". If you reach for Google Docs before AI tools, if you don't use Cursor for prototypes, if you haven't built your own agents - you're not meeting expectations. This will be measured in performance reviews.

Kyle's take: Bit overblown with the "saving the world through real estate" nonsense, but he's absolutely right about the core message. AI is a skill like any other - you get better by using it daily, not watching tutorials. Making it part of performance reviews is aggressive but probably necessary to survive moving forward.

Source: CEO letter published on Twitter

Member Question: "What's the best way to get clients for micro SaaS?"

Kyle's response: I'll be honest - micro SaaS founders don't have money, which makes them nightmare clients! They'll pay you less than $1000 a month (which might be 20% of their entire budget), then ring you constantly expecting miracles.

You're better off targeting bigger companies who can actually afford your services. It feels harder. But it’s actually easier. Believe!

But if you're determined: scrape Product Hunt for new launches, build an audience on Twitter where they all hang out, and do the whole "building in public" thing. Target people like Simon Hoiberg's audience. Just remember - it's easier to sell to people who actually have money! Funny that!

This question was discussed at [00:21:34] during the live session.

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.

Streaming on YouTube (with screen share) and TikTok (follow and turn on notifications for Live Notification).

Audio Podcast on iTunes and Spotify.

AI with Kyle - Tools, Resources and Products: Daily Show | Build an AI Business Challenge | AI Playbooks | 10 Week AI Summer Camp | BATON AI Business Series | AI App Building | AI Trainer Business Planner | What AI Business Should You Start? | AI Authority Accelerator | AI Automation Accelerator | 🏁 AI Workshop Kit

Previous
OpenAI's GDP-Val Tests AI on Real Jobs + ChatGPT Pulse Wakes You With AI Briefings
Next
Sonnet 4.5 + Lovable Cloud
Keep going

More this week

OpenAI caught red-handed + Accenture axe 11,000 jobs | AI with Kyle