October 1, 2026
TechCrunch's breakdown of Opus 5.5's writing habits is being treated as a bit of trivia -- the model loves telling you something 'matters,' and it reaches for the word 'dependable' 23 times more often than a human writer would. But that's not just a quirky fingerprint. It's a reminder that every AI model has a voice, and that voice is detectable if you know what to look for.
For a business leaner on AI-generated copy, code comments, or customer-facing text, this matters in a very practical way. If a model's tics are predictable enough for a reporter to count them, they're predictable enough for a competitor, a regulator, or a customer to notice too. Content that reads as generic or oddly repetitive erodes trust faster than most marketing teams realize -- and 'dependable' showing up in every third paragraph is exactly the kind of thing that makes a reader's eyes glaze over, or worse, makes them wonder if anyone actually reviewed the copy before it shipped.
My take: this is a useful nudge for any company using AI to draft anything a customer will read. The fix isn't to ban AI-written text -- it's to treat the first draft as a draft, not a finished product. Teams building out marketing and campaign workflows with AI assistance should budget real human editing time, the same way you'd fact-check a press release. The tools are good enough to save time. They're not yet good enough to skip the read-through.
Legato's launch of AI hearing glasses, also covered by TechCrunch, is a quieter story than the usual AI hardware hype cycle, and that's exactly why it's worth a second look. The startup's pitch isn't about novelty -- it's about cost, comfort, and the stigma that keeps people from wearing traditional hearing aids in the first place. Folding hearing assistance into a pair of glasses someone would wear anyway is a genuinely smart use of AI: it solves a real adoption problem rather than manufacturing a new one.
This is a useful contrast to a lot of what passes for 'AI hardware' right now. Plenty of gadgets bolt a model onto a device because they can, not because it fixes anything a customer actually struggles with. Legato's framing -- accessibility, not novelty -- is the harder and more durable bet. If the device delivers on comfort and price the way the company claims, it could meaningfully expand who's willing to use hearing tech at all, which is a bigger market unlock than another flashy AI wearable chasing early adopters.
The caveat worth stating plainly: hardware that touches health, even adjacent to it, carries higher stakes than a chatbot tic. Comfort claims and accessibility pitches are easy to make at launch and harder to prove at scale, especially once a device has to perform consistently for users with varying degrees of hearing loss. I'd call this promising rather than proven -- a startup worth watching, not yet a verdict.
Both stories point to the same lesson for business buyers: AI tools, whether they're writing your emails or sitting on someone's face, deserve a closer look before they get trusted by default. A model's verbal tics are a small, visible symptom of a bigger truth -- these systems have patterns, blind spots, and limits that only show up under real use. The same scrutiny that catches an overused word should also apply to any AI feature you're evaluating for internal workflows, whether that's through a no-code app builder or a vendor's bolted-on assistant.
Which matters more to your business right now: catching AI's writing tells before they reach customers, or watching whether accessibility-first hardware like Legato's actually delivers on its promise?
Sources