Recall #9: Top of Mind

This week, we wanted to give you a peek behind the curtain at what we’re working on.

Our team has been heads down creating Sentience’s new Suggestion Engine. The engine will be used to determine what’s top-of-mind for you and get it off your plate. It will pre-draft messages for you based on your priorities, triage information and strategy into documents, suggest and implement routines based on your patterns, and much, much more.

This is all focused on our core goal of modeling both your motivation and your knowledge. Sentience will learn how you prioritize and get better over time from watching you work.

Our engineers Jules and Ben are spearheading this project, which is set to become the new core of our desktop app homepage. We’re going to launch this in 2 weeks to an early cohort, and then broadly on August 25th.

Right now, the Suggestion Engine searches through all incoming memories and identifies what you need to work on: email replies, calendar invites, research, and more gets kicked off to your agent from a centralized feed. Instead of staring at the dreaded empty prompt box, your Sentience is the one sending you the first message.

At Sentience, we talk a lot about “managing up”—the idea that you should push rather than pull tasks and insights. Sentience will now start managing up to you instead of waiting for you to prompt it.

This is an early mock of how suggestions produced by the engine will look on the homepage. You’ll be able to edit and send messages using hotkeys so that managing your life from your Sentience becomes effortless.

These are preliminary designs for the new homepage! Our designer Pedja is bringing these to life in the desktop app very soon.

AI writing benchmarks normally measure two things: whether a sentence or passage is grammatically correct, and whether it’s factually correct. But benchmarks don’t measure whether writing is written in your style, because there's no test set to score against. This is why AI writing is still so recognizable even when you ask your agent to write for you.

Ben Carsley, who leads our tone and style work, took his full writing corpus—2,406 texts, 409 emails, 406 Slack messages, 134 passages of academic writing—and analyzed it twice: once by topic, to see how his writing shifts with subject matter, and once by style, to see how his style shifts across channels.

The numbers are revealing. Ben starts 96% of his texts with lowercase but only 18% of emails. His median text is 18 words, his median email is 40, his median academic paragraph length is 120. His em dash rate quadruples from texts to academic writing: 3.3 per thousand words to 13.2. Corpus linguists established forty years ago that people shift their writing as much between registers as much as different people in the same register, and Ben’s research confirms the results.

Ben’s work driven massive improvements in person-specific tone and style output. Our tone and style work is continually evolving, so stay tuned for more.

Read the full writeup at sentience.com/voices. Kudos to Ben for all his work!

Orçun is a VC at ScaleX Ventures in Istanbul, and he started using Sentience the way most people do: as a work tool. Sentience became the layer that pulled his deal flow, LP comms, portfolio updates, founder conversations into one place.

Gradually, the ways he used Sentience grew. Orçun logs everything—coffee with his mom, a movie, photos with the people he loves. He puts it all in his calendar in full detail, and Sentience makes it searchable and alive. In late June, he got married (congratulations!) and he used Sentience to help pull together information to write his wedding vows. He had the story and the flow he wanted, but Sentience surfaced the specific memories he would have otherwise forgotten to include. In Orçun’s words: "It resulted in something deeply personal and true."

Until next time,

—Sam and Teddy