September 14, 2026
TechCrunch's own reviewer reported something that would have sounded absurd a year ago: with iOS 27, they're actually using Siri again. That's not a small thing. Apple has spent years promising a smarter assistant while watching ChatGPT and Gemini eat its lunch on mindshare. A Siri that people voluntarily use again means Apple has a real distribution advantage -- it's already on a billion phones, and it doesn't need anyone to download anything.
The more interesting story is what third-party developers are already building on top of it. TechCrunch also covered Daydream, a fashion discovery app that now uses Apple Intelligence to turn photos saved in your camera roll into shoppable outfit results, searchable directly through Siri without opening the app. That's the pattern to watch: the assistant becomes a routing layer, and the value shifts to whoever plugs a useful, narrow workflow into it. Any business building customer-facing tools should take note -- the winning move isn't necessarily owning the assistant, it's owning the workflow the assistant hands off to. That's the same logic behind pairing a lightweight AI app builder with the channels your customers already use, rather than trying to build a new destination app from scratch.
Microsoft published a new AI "code of conduct" this week, according to TechCrunch, instructing its models not to hack systems or trick humans, and to prioritize supporting people over replacing them. On paper, that's hard to argue with. In practice, a code of conduct is a values statement, not a technical constraint -- it tells you what Microsoft wants its models to do, not how it guarantees they'll do it.
I'd argue this matters less as a safety milestone and more as a signal of where the liability conversation is heading. Enterprise buyers are increasingly asking vendors to put guardrails in writing before they'll sign a contract, and Microsoft is getting ahead of that ask. Whether the principles hold up under real deployment pressure is a separate question -- codes of conduct are easy to publish and hard to enforce, and skeptics are right to want to see this backed by audits, not just prose. For businesses evaluating any AI vendor, the real question isn't whether they have a nice document, it's what their security and incident response posture actually looks like when something goes wrong.
Superhuman's acquisition of Fathom, the YC-backed meeting notetaker with over 400,000 monthly active users, is the clearest of today's three stories in terms of what it means for buyers. TechCrunch frames this as part of a broader push by productivity platforms toward agentic work -- tools that don't just record what happened in a meeting but act on it. Consolidation like this usually means one of two things: either the acquirer wants the user base, or it wants the workflow data to train something bigger. Given Superhuman's trajectory from email client to broader productivity suite, it looks like both.
For businesses, this is a reminder that point solutions -- the single-purpose app that does one thing well -- have a shelf life once a platform decides that function is core to its roadmap. A standalone notetaker with 400,000 users still got bought rather than staying independent. If you're building your operations around a stack of narrow best-of-breed tools, it's worth asking which of them are acquisition targets and what happens to your workflows if they get folded into someone else's roadmap. That's part of why teams increasingly look at consolidating on a single platform rather than betting on a dozen point tools staying independent and interoperable forever, especially in functions like project and task management where meeting notes, tasks, and follow-ups need to live in one place, not three.
So which of today's stories worries you more as a business buyer: that Apple and Microsoft are racing to make AI assistants feel more trustworthy, or that the tools you rely on today keep getting absorbed into someone else's bigger platform?
Sources