The skinny on what's happening in AI - straight from the previous live session:
Highlights
🎭 69% of Workers Use AI Daily But Hide It From Colleagues
Anthropic study reveals workers terrified of showing they can be automated. Using AI in secret creates worst of all worlds.
Kyle's take: Nearly 70% are using AI and pretending they're not - actively lying or sneaking around. 86% said AI saves them time, 65% satisfied with AI's role in their job. But 69% mention stigma around using it at work.
The fear: if you don't use AI, you fall behind. If you use it too much, you're showing employers what can be automated. Valid concern - I see it in workshops. Staff are wondering: if AI can do my job, why does company need me?
This is a valid concern. They're training their replacement by using AI - companies collecting that data for later automation. This weird position where AI makes your life better but also makes you replaceable. Solution? Do your own thing alongside the 9-5. Jobs just aren't that safe anymore.
📈 AI Ads Get 20% More Clicks - Until People Know It's AI (Then -31%)
NYU/Emory study: AI visuals beat humans, but only if nobody knows they are AI!
Kyle's take: This is why I pivoted from digital marketing three years ago - saw this coming! AI-created visuals got 20% more clicks than human-made ads. That is massive in advertising world.
But if people realise it's AI-generated, clicks drop 31% BELOW human ads. As long as they don't know, they prefer AI. Moment they find out, they hate it! This may push companies towards lying, not disclosing AI use…
Important note: AI ads were still selected by human experts from many options. So there's still human judgment layer in this study.
The big takeaway here though is that we're perfectly happy with AI when unconscious of it, get upset when we find out. This pattern appearing everywhere - not just ads but all AI interactions.
📝 Wikipedia's Field Guide to Spotting AI Writing: The Complete Tell List
Wikipedia catalogued every AI writing giveaway.
Kyle's take: Wikipedia has created an exhaustive list of AI writing tells because people keep submitting AI drafts. It’s a LONG list so highly recommend you go and check it out here: Wikipedia AI giveaways.
Top giveaways: undue emphasis on symbolism/legacy/importance - everything's "pivotal," "crucial," "revolutionary."
Negative parallelisms: "It's not just X, it's Y."
Overused vocabulary: delve (famously!), tapestry, fostering, garnered, interplay, underscores.
Funnily enough this list is training material for future AIs - OpenAI can use it to remove these patterns. Or you could feed it to custom instructions: "for the love of all that is holy don't do any of these!”
🎯 Demis Hassabis: AGI Closer Than We Think, 2026 About Multimodality
DeepMind head's in a rare interview. Unlike Altman, not trying to sell anything - just doing science.
Kyle's take: Demis doesn't do that much publicity - so when he does it’s generally worth paying attention. Unlike Altman, doesn't hype because he doesn't need to raise money - has almost infinite Google budget.
His 2026 prediction: bringing modalities together - Gemini, Veo (video), Nano Banana all unified. Not just generating but understanding between them.
Already seeing this - Nano Banana Pro uses Gemini to reason before generating, it’s why it's good at infographics. Next: world models - giving AI simulated worlds to learn in, like children learning through play. Yann LeCun left Meta to start world model studio, Fei-Fei Li working on marble model.
AGI timeline: doesn't give dates (not stupid) but sounds like years not decades.
Kyle's response: Really hard balance! Starts with company discussion - "AI's here, what do we do?" Need to reassure people they won't be replaced (ideally honestly). Frame it as: AI automates boring parts so you focus on human stuff - judgment, taste, connections, empathy, creativity. Be honest: "We're taking away boring stuff for more creative work that needs humans.: Take away spreadsheet plugging and routine reports, free up creative time. Problem: some people only good at algorithmic tasks AI handles. They're in trouble. Those tasks massively at risk.
Kyle's response: If it works, sounds fantastic! Big problem with first businesses - people create products they think are great without checking if market wants it. Tool that filters 100 ideas down to top earners would be valuable. Can't replace talking to customers but can help identify which customers to talk to, which idea to pursue. Many paralysed by too many ideas - tool giving measure of willingness to pay would be killer. For CPG service, if data's good, extremely valuable. It all depends on the quality of the evaluation honestly!
Kyle's response: AI itself isn't a business - you still needs to create value, convince people to buy and then sell above cost for profit. Seeing too many get-rich-quick schemes: "Press button, generate 10,000 faceless YouTube videos, make bank." Nope - that’s not creating value. And if it’s making cash whilt being that easy, market floods, nobody makes profit. Start with something valuable, then use AI. Or use AI to create value, but you still must make value! Too many are using AI without creating anything anyone wants to buy.
Kyle's ChatGPT Birthday Swag: "They sent me hat, hoodie, and birthday cake cookie for ChatGPT's 3rd birthday. I look like American pizza delivery boy but it's actually not bad for someone terrible in hats."
Kyle's Community Launch: "Reinvigorating my 3,000 member community. Adding 5-day AI readiness challenge for overwhelmed folks. 100+ playbooks, courses with 123+ lessons each, all free. Not just me posting - want discussion, partnerships, support. Join at aiwithkyle.com"
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).
