Logical Is Building the Copilot That Helps Before You Ask
The Feature — Two friends who met at age six in Kandy walked away from PhD scholarships to kill the blank prompt box. Their desktop assistant watches how you work, then does the next thing for you.
Every knowledge worker knows the small, dumb ritual. You are in the middle of something - a spreadsheet, a thread, a half-written reply - and to get an AI to help, you stop, open a chat window, and explain your whole situation from scratch. You paste in the email. You describe the deadline. You clarify what you already told it two minutes ago in a different tab. Then you copy the answer back out. Multiply that by a workday and you have handed a meaningful slice of your week to the machine that was supposed to save you time.
Logical, a San Francisco company in Y Combinator's Fall 2025 batch, is built on the bet that this ritual is the actual problem - not the model behind it. Its product is a desktop copilot that watches what you are doing across your apps and offers to do the next thing, without a prompt. The founders describe the current wave of tools bluntly.
"Prompt-first AI is fundamentally flawed. We waste hours every week writing lengthy prompts, context-switching and copy-pasting to get a chatbot to understand what we're doing." Logical, Launch YC
The company's own shorthand for the idea is disarming: Clippy, but actually good. It is a joke with a thesis buried inside it. Microsoft's paperclip failed because it interrupted you with help you did not want. Logical is wagering that the same instinct - an assistant that shows up unprompted - works when the assistant is genuinely useful and reads context correctly. Proactive instead of annoying is the entire difference, and it is a hard line to walk.
01 / What it doesAn assistant that lives where the work is
Logical runs on your desktop rather than in a browser tab, and that placement is the point. Because it sits at the level of your machine, it can see across the apps you already use and carry context between them. Draft an email in Gmail, jump to Slack, open a spreadsheet - Logical follows the thread instead of forgetting everything the moment you switch windows.
In practice that folds several familiar tools into one. It drafts emails and messages across Gmail, Slack, iMessage and Apple Mail. It hooks into Gmail and your calendar so you can search your inbox in plain language or check availability without hunting. A feature the company calls Logical Lumos handles the small, in-the-moment stuff: suggesting a formula and flagging an error in Excel, or explaining a term the instant you highlight it in a document.
The one that tends to catch people is automatic to-do detection. Logical reads the deliverables buried in your inbox and messages, drafts them, and then reminds you to finish. It is less a chatbot you visit and more a layer that quietly keeps track of what you owe other people.
02 / Who it's forBuilt for people who live in their inbox
The target user is the ordinary knowledge worker - the person whose day is email, chat and spreadsheets, and who feels the tax of AI friction most acutely. Logical is offered as a desktop app with what the company calls generous free credits, a classic freemium on-ramp aimed at individuals first. It is early. The team is small, the batch is fresh, and the user base is still being built one download at a time.
That early-stage honesty matters for reading the product. Logical is not claiming to have solved the office. It is making a focused argument that the interface for AI - not the intelligence - is where most of the daily pain lives, and pointing its small team squarely at that.
03 / The differenceProactive and private, by design
Two choices separate Logical from the crowd of chat-first assistants. The first is being proactive rather than prompted. ChatGPT's desktop app, Microsoft Copilot and Google's Workspace AI all still largely wait for you to start the conversation. Logical tries to start it for you, reading intent from what is on screen.
The second is privacy as architecture rather than a promise stapled on later. Logical states that your data stays on your desktop and is never sent to its servers. For a tool whose whole value comes from watching your email, messages and documents, that on-device posture is not a nice-to-have - it is the thing that makes people willing to let it watch at all.
Truly helpful AI, the founders argue, should live where work happens - on the desktop - not in a tab you have to feed by hand.
04 / The foundersFrom a Kandy classroom to Y Combinator
Logical was founded by Sam Karu - Samurdhi Karunaratne - and Anushka Idamekorala, and the origin story is unusually literal. They met at age six at Dharmaraja College in Kandy, Sri Lanka, and finished school ranked first and second in mathematics in the Kandy District. Karu went on to medal at the International and Asian Physics Olympiads, study at the University of Peradeniya, win a UCLA PhD scholarship he later left, and work on AI model optimization at NVIDIA. Idamekorala studied at the University of Moratuwa, placed first in Sri Lanka at the IEEEXtreme programming competition, worked as a software engineer, won a University of Virginia PhD scholarship he also left, and led B2B SaaS engineering teams.
"Anushka and I have known each other since we were six years old, so there is something surreal about now building a company together in Silicon Valley." Sam Karu, Co-founder & CEO
Their acceptance carried a second headline. Logical has been widely reported as the first startup with an entirely Sri Lankan founding team admitted to Y Combinator, alongside early funding of over LKR 150 million - roughly half a million US dollars. The founders have been explicit that the milestone is part of the mission.
"We want to prove that world-class AI products can be built by Sri Lankan engineers competing at the highest level globally." Sam Karu, Co-founder & CEO
05 / The marketA crowded field, a narrow wedge
The AI-assistant market is loud and getting louder. Every major platform is shipping a copilot, and a wave of productivity tools - note-takers, launcher assistants, document AI - are racing toward the same knowledge worker. Logical's wedge is not to out-model any of them. It is to own the moment of action: the point where you would otherwise stop, explain yourself, and copy-paste. If that wedge holds, it is a durable one, because it is a behavior problem more than a capability problem.
The risks are the obvious ones for a company this young. Proactive help has a fine line between useful and intrusive, and Clippy is the cautionary tale the founders themselves keep invoking. On-device processing is harder to build than a server call. And the incumbents it is nudging past have distribution Logical does not. What it has instead is a sharp point of view about where AI actually wastes your time, and a team that has spent a lifetime being early.
For now, the pitch is simple enough to try in an afternoon: download it, let it watch, and see whether the next thing it suggests is the thing you were about to do anyway. That is the whole test - and it is the one Logical wants you to run.