The personal AI assistant war: why this time is different
Key points
Key takeaways from Big Technology Podcast interview with M.G. Siegler, author of the Spyglass newsletter (October 2026):
Assistants are here to stay this time. Siegler, who wrote on 1 September 2026 that no one had yet nailed the true personal assistant experience, now believes at least one — and probably a handful — of these products will be used by most people, and that they will increasingly be baked into the AI services we already use. Mass adoption is a question of timetable, not of whether it will happen.
Why this wave is different. LLMs have removed the rigidity of voice assistants from the Alexa era, which is what created the twin "blank page problem" and "boil the ocean problem": users do not know what to ask when they can ask anything, while products try to do too much at launch. The emerging fix is proactive prompting — the assistant notices a bill due in ten days and offers to log in and set up recurring payment.
Instinct's early formula. The startup got off the ground fast by riding existing messaging paradigms: users sign up and text the assistant like a friend or a human assistant, with no new app or suite to learn. Siegler's early hooks were reviewing expenses, credit-card bills and subscriptions in natural language to find savings.
Instinct's momentum. VCs touted the service across social feeds, and by reported accounts it has raised roughly three rounds in a few months and is valued around $10 billion. Its chief executive is a young, relatively unknown figure who came out of Sierra; Siegler disclosed that his family invests in Sierra and his wife advises it, while he holds no financial stake in Instinct.
No single killer use case. Siegler argues the winner should do five to ten things well, each resonating with a different user base, plus one "it did this thing for me" moment that drives word of mouth. Trust matters more here than in other categories, he says, and adoption will come in baby steps rather than one big handover.
Banking and subscriptions under threat. If agents shift cash from non-interest-bearing accounts into yield, banks lose an arbitrage that underpins their solvency, and optimising credit-card points and loyalty programmes upends further value. Siegler warns that agents pinging users to cancel dormant subscriptions threaten the set-and-forget economics many subscription businesses depend on, forcing a new optimisation industry to emerge.
Who is in the ring. Meta's Muse looked "surprisingly strong" to Siegler; OpenAI shipped its Dots product, xAI has Grok Bot and Microsoft has entered the battle in its own way. Apple occupies a dark-horse position and Anthropic is absent from the consumer assistant race.
Google's missed moment. The episode's closing question is why Google did not build its own version of Muse: beyond its Spark assistant, the company looks like the glaring incumbent without a competitive product in the personal assistant war.
Read more: Big Technology Podcast
