Peak AI Slop: When AI Starts Talking to AI

Unfiltered with Matt & Nige returns 🎙️And Episode 6 starts exactly how it should: Matt and Nigel are still figuring out the “perfect” intro… they nearly land it.Unfiltered, as promised.We’ve got a bit of transatlantic energy, a bit of UK weather whiplash, and the usual life admin creeping into the first five minutes… which somehow always ends up being the most accurate set up for the actual topic.Once they settle in, the real theme appears…Are we heading for peak AI usage… or peak AI slop?TLDR: we’re using it more, trusting it more, and generating more… but the signal to noise problem is getting worse.This week on Unfiltered the two talk all things:* AI talking to AI (and why LinkedIn comment sections are starting to feel inhuman)* the confidence problem: when AI sounds complete… but it isn’t (dangerous when it’s a topic you don’t know well)* triangulating answers: asking Gemini, Claude and ChatGPT the same question to sanity check outputs* token maxing vs token minimising: why “how many tokens did you burn?” is about to become a very real ROI conversation* and the next shift from co work to auto work: set a task, go to sleep, wake up and things moved (and yes, the cost curve gets interesting)Getting a little deeper… this episode is really about trust and judgement. Not “trust the AI.”Trust in the chain: what goes in, what comes out, what gets repeated, and what happens when humans stop checking because everything is delivered with confidence.Peak AI slop: when humans step out of the loopMatt’s point was simple and painfully true: using AI to comment on LinkedIn posts drives him mad… and it gets even worse when someone uses AI to comment on the AI comment.You end up with a thread of bots congratulating bots, while humans scroll past and feel slightly dead inside.Nigel linked it to a bigger issue: if people aren’t asking the right questions, we’re not getting smarter… we’re just creating more content. And then “numpties like us” are left correcting it later.The confidence problem (again)This is the part Matt kept coming back to: AI can be magic… and also unbelievably dumb… and it delivers both with the same confidence.If it’s a topic you know well, you spot what’s missing.If it’s a topic you don’t know well, you assume it’s brilliant. That’s the trap.Triangulation: three models, three answersNigel’s workaround was practical: ask multiple models the same question and compare.He even did it with fitness supplements - same question, different models, different answers. Fascinating… and slightly alarming.Matt added the industry twist: clients are now blending multiple sources - AI outputs, ecosystem partners, specialists - and triangulating that too. The “research stack” is changing.Token maxing vs token minimisingThen the ROI thread kicks in. If one person is burning a quarter of a million tokens worth of work… how many humans would that be? And is it actually better if you still need review?Nigel’s take: token minimising is going to be the next wave - people getting serious about cost, value, and whether the output is worth the spend.Matt connected it to the always on model: once you’re running auto work overnight (set tasks, wake up to results), the consumption curve gets wild - but so does the productivity potential.Lawyers: same hours, different painOne of the most real examples: lawyers aren’t saving time yet - they’re spending the same hours correcting contracts their clients generated with AI.So instead of “AI removes work,” it becomes “AI changes the work… and sometimes adds review overhead.”The human ending (because obviously)They wrap in classic Unfiltered practicality: a real life to do list, travel logistics, and Matt signing off with “big chats… big love.” Which feels right!The big takeawayAI is everywhere.Output is cheap.Confidence is free.So the advantage is shifting to the people who can: ask better questions, validate faster, minimise waste, and keep human judgement in the loop.Subscribe for more Unfiltered every week 🎙️ This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit unfilteredwithmattnige.substack.com