Sergei
Sorokin

In March 2026, Sergei Sorokin became co-founder and CEO of Highlight, a San Francisco company that calls itself "the shared intelligence layer for the agentic age of work." The same week, Highlight announced a $40 million Series A led by Khosla Ventures. The headlines framed it as a hire. Sorokin framed it as a beginning.

The pitch is deceptively dull. Highlight is a desktop app that runs in the background, watching the tools people already live in - Slack, Figma, Linear, Zoom, Claude Code - and quietly building a memory of what the team is doing. It records decisions. It tracks changes. It drafts the follow-ups. If you miss a meeting, it can tell you what happened without you scrolling through a week of messages. No dull product ever sounded so much like a relief.

Sorokin's diagnosis is the kind that sounds obvious once someone says it out loud. The models are smart enough. What slows teams down is everything between the models and the work: the copy-pasting, the status updates, the re-explaining of context to whoever just joined. He has a number for the damage. By his estimate, the "coordination tax" can eat up to 24 hours per employee, per week. Half a working life spent keeping everyone on the same page.

That is the bet. Not a smarter assistant. A shared memory that means nobody has to be the human glue anymore.

It is a contrarian thing to say in 2026, when most of the industry is still racing to build the cleverest model. Sorokin's view is that the cleverness is increasingly free, and that the value has quietly migrated somewhere less photogenic: to the layer that remembers what was decided, who is doing what, and why. He talks about "the agentic age of work" not as a future of robots replacing people, but as a present where humans and software agents already work side by side and keep tripping over the same missing context. Fix the memory, he argues, and the agents finally become useful instead of impressive.

Sorokin's whole argument rests on a single, uncomfortable idea: most of the friction at work isn't thinking, it's re-syncing. The chart below is the problem Highlight is built to delete.

Before Highlight, there was Discord - and at Discord, there was almost nobody. Sorokin joined in 2016 as the company's first product hire, back when the team was about 20 people and the app was a place for gamers to talk over voice chat. He left roughly eight years later, having been its first PM manager, its Head of Product & Data, and eventually VP of Product.

The numbers underneath that arc are the kind founders frame on a wall. Discord grew from around 5 million monthly active users to nearly 300 million. Headcount went from 20 to about 1,000. And the part Sorokin tends to claim most directly: he built the revenue business from inception, scaling it past $500 million a year, and launched and scaled Nitro, the subscription product that turned a free chat app into a real company.

Building the money is rarely the glamorous job. Growth gets the press; monetization gets the spreadsheet. Sorokin spent his Discord years on the spreadsheet - which is exactly why he is qualified to argue that the next era of software will be won on plumbing, not magic.

There is a reason the Discord story keeps surfacing in everything written about Highlight. It is a credential, but it is also a worldview. A chat app for gamers had no obvious business model when Sorokin arrived; the easy thing would have been to bolt on ads and annoy the only users it had. Instead the company built Nitro, a subscription people actually wanted, and grew revenue without poisoning the product. That instinct - find the value people will gladly pay for, then make it feel inevitable - is precisely the muscle a productivity startup needs when it is asking teams to trust software with their entire working memory.

He also lived through the part founders forget to mention: the years of being early and slightly invisible. Being employee-with-a-product-title number one at a 20-person company means doing whatever the week requires, then watching the eventual success get attributed to the brand. Sorokin seems to have made peace with that long ago. The work he is proudest of tends to be the kind users never see.

Sorokin's resume reads like a tour of the moments right before things got big. In 2008 he joined Facebook's first Growth team - the small group that would go on to define how an entire industry thinks about user acquisition. He studied Economics and International Relations at Stanford, graduating in 2009.

Somewhere in there he ran the Ravenwood live-ops games franchise, a reminder that he learned to build habit-forming products in the unforgiving school of free-to-play gaming, where if people don't come back tomorrow, you have no business. He also built a team responsible for roughly two-thirds of Yahoo's mobile traffic - a footnote on most bios, a small empire in its own right.

The pattern is hard to miss. Facebook's first growth team. Discord's first product hire. Highlight's co-founder. Sorokin keeps arriving at the steepest part of the curve, before the curve is obvious to anyone else.

After Discord, Sorokin did the thing accomplished operators do when they aren't ready to commit: he became useful to everyone. For about two years he built and advised AI startups, with stints connected to Weights, Aura, and OffCall. It was a working sabbatical with a thesis forming underneath it.

What he saw, by his own account, was the same broken movie playing in company after company. Fragmented tools. Agent workflows that promised autonomy and delivered busywork. Smart models bolted onto dumb processes. The intelligence had arrived; the coordination had not. When Highlight came along - a desktop AI app that had already spun out of the gaming-clips company Medal and crossed 500,000 users - the problem and the operator finally matched.

Highlight didn't start as a Sorokin project. It spun out of Medal, the gaming-moments platform built by Pim de Witte, who now serves as Highlight's chairman. By the time Sorokin stepped in as CEO, the desktop assistant had already found half a million users, including people at Google and DoorDash, and had carved out a clear identity: an app that turns scattered files and meetings into notes, to-do items, and plain-language explanations of complex things.

The roadmap is more ambitious than the product. Highlight is building toward a "unified team memory engine" - software that doesn't just summarize but acts, orchestrating multi-person tasks and deploying agents that move work forward without being asked. The $40 million from Khosla Ventures, with 359 Capital, General Catalyst, Valor Equity, Common Metal, Makers Fund, Collaborative Fund, Arcadia, and SV Angel along for the round, is meant to roughly double the team in San Francisco.

There is a neat symmetry to it. The operator who spent eight years making Discord remember how to make money is now building software whose entire job is to remember - so that humans, and the agents working beside them, never lose the thread.

The competitive question is obvious and Sorokin does not dodge it: why won't Slack, or Notion, or any of the big suites just absorb this? His answer is structural. Each of those tools owns a slice of the work and has every incentive to keep you inside it. Highlight's premise is the opposite - it sits across the tools, indifferent to which one you prefer, and treats the connective tissue between them as the actual product. A memory layer that belongs to no single app is worth more precisely because it belongs to no single app.

Whether that holds is the open question of the next few years. Capturing everything a team does is powerful and also delicate; trust, privacy, and the simple fear of being watched all sit in the same room as the upside. Sorokin's wager is that teams will trade a little surveillance anxiety for the end of the status update, the same way they once traded their phone numbers for a free chat app. He has been right about that kind of trade before.