Cluely, under the microscope: viral, then what?

Cluely started life as Interview Coder: a tool that watched a coding-interview screen and fed answers back through an overlay invisible to the interviewer. After the founders were suspended and dropped out of Columbia, they reincorporated as Cluely, raised around $5.3M within weeks, and later took a reported $15M Series A from a16z. The product generalized: real-time assistance for sales calls, interviews, exams, and meetings, rendered through a GPU-level overlay (DirectX on Windows, Metal on macOS) that doesn't appear in Zoom, Meet, or Teams screen-shares.

We picked it because it's the purest example of a pattern we're asked about constantly: a product that goes viral on positioning, raises on momentum, and then has to answer whether there's a durable product underneath the moment.

The growth engine is the product, and it worked. Turning a suspension into a "cheat on everything" campaign generated national coverage and a flood of sign-ups for a fraction of normal CAC. Founders pay us to think about distribution; Cluely solved distribution before it solved much else, and that ordering is not always wrong.

The core technical insight is also real. The invisible overlay — assistance that lives on top of any app and stays out of the screen-share — is a genuinely useful primitive. Real-time, context-aware help during a live conversation is a category, not a gimmick. And the soft pivot toward "meeting assistant" shows the team can read the room when the original framing starts costing them enterprise trust.

If the question is "can two people manufacture attention and a usable demo," the answer is clearly yes.

Here's where the trade-offs land — and most of them land on trust.

It's a thin wrapper, and thin wrappers get absorbed

Strip the overlay and the core is a prompt over a frontier model. Reviewers describe it as another AI chat wrapper whose main differentiation is persistence across a call. The structural risk: the platforms it sits on top of are shipping the same capability natively. Once ChatGPT-style "listening mode" is built into the assistant people already pay for, a standalone overlay has to justify a second subscription. That's a hard sell.

Invisible-by-design is a trust liability, not just a feature

The thing that makes Cluely useful in an interview is the thing that makes it radioactive in enterprise: software explicitly engineered to be undetectable by the other party. That forecloses the highest-value market (sanctioned, in-the-open sales coaching and support) at the exact moment the team is trying to pivot toward it.

Always-on friction

Users report the global keyboard shortcuts interfere with other apps and can't be turned off cleanly, so leaving it running outside a meeting is actively annoying. A product that's irritating when idle doesn't earn a permanent slot on anyone's machine.

No defensible memory layer

Real-time suggestion without durable, structured recall across conversations means there's nothing compounding. The moat in this category is the accumulated context, not the overlay — and that's the part Cluely under-built.

We'd keep the overlay primitive and the real-time instinct, and rebuild everything around trust and defensibility. The goal: a product that wins the sanctioned market and compounds over time, instead of one that wins a news cycle.

1. Flip invisible to consent-aware

Default to a visible, disclosed assistant for the enterprise market, with a clear "assistance active" state. You trade the interview-cheating use case (a liability) for the sales-and-support market (a business).

2. Build the memory layer that's missing

Every call writes structured notes, entities, and outcomes to a per-user / per-account store. Suggestion quality compounds as the system learns the rep, the deal, and the customer. That's the moat.

3. Route models by job

4. Make it pleasant when idle

Configurable, scoped shortcuts; a real "off" state; no hijacking of other apps. Earn the permanent install.

What we'd cut, and how we'd ship it.

Twelve months out, we'd bet the overlay category survives and Cluely-the-overlay doesn't — at least not as a standalone consumer subscription. The capability gets absorbed into the assistants people already pay for, and the brand's viral, edgy origin becomes a liability in the enterprise market that's actually willing to pay.

The team clearly has the distribution instinct most founders lack. If they spend the runway building the consent-aware, memory-backed version — and let go of the use case that made them famous — there's a real company here. If they keep optimizing for the next viral moment, they're renting attention they can't keep.

Viral is a start. It is not a moat.