The first piece of advice we give every founder who walks into xlabs Ventures is also the piece of advice they want to hear the least: don't build yet.
It's a hard sell. The founder has been thinking about the idea for months — sometimes years. They've already mentally shipped it. They've already pictured the launch. The last thing they want is for us to put the brakes on the bit they're excited about and ask them to spend two more weeks on landing pages, ad tests, and customer conversations.
But the data is the data. Forty-something percent of startups fail because they built something nobody wanted. The single highest-leverage thing we can do for a venture partner is to stop that outcome before it starts.
So we built a tool that makes the validation step harder to skip than to do.
It's called xlabs Smokescreen.
The problem we kept seeing
Every founder we worked with hit the same wall at the same point.
They knew they were supposed to validate. They'd read the books. They knew about smoke tests, landing pages, paid traffic, customer interviews. But the practice of running a real smokescreen — well — took longer than anyone expected. There was a landing page to design. Copy to draft. Visuals to generate. Forms to wire. Analytics to connect. A waitlist to capture. Ads to run. Conversations to schedule. Responses to analyse.
By the time the average founder had assembled all of that, two weeks had vanished. By the time they'd run the test long enough to draw a conclusion, another two had. Half the time, they'd given up halfway and started building instead — because building, at least, felt productive.
The cost of running validation had become the silent reason teams skipped it. The principle was free. The execution was not.
We built Smokescreen to crush the execution cost.
What Smokescreen does
Smokescreen is xlabs's internal validation engine, built on the same agentic principles we use across the studio: senior engineers in the loop, agents doing the volume work, evals on every output.
A founder brings the idea in as a brief — the problem, the audience, the value hypothesis, the riskiest assumption. From there, Smokescreen orchestrates the validation pipeline end to end.
In practice, that means:
- Validate — landing page + waitlist in minutes, not days. Smokescreen's AI assistant walks a founder from a one-line idea to a live, hosted landing page and waitlist in under five minutes. Copy, structure, hero, problem/solution, and the CTA are all generated against the brief and editable in place. The page ships on Vercel with the right tracking wired in from the start, so the data is clean from the first visitor.
- Hard-signal capture by default. Waitlists are built to surface intent, not curiosity. Email plus two or three sharp micro-questions calibrated to the idea — the kind of questions that separate "I'd take a look" from "I'd pay for this." We've learnt the hard way that soft signals lie, so every waitlist defaults to capturing willingness, urgency, and price sensitivity, not just an inbox.
- Validation dashboard with a pre-set bar. Live signups, conversion rate, traffic sources, and themed feedback in one view — with a success threshold defined before the test goes live. The dashboard tells you whether you cleared it. No moving goalposts. One-click PDF export for investors or the team.
- Explore — deep strategic reports on demand. Once Validate produces a signal, founders need to know what to build, who to sell to, and whether the market is real. Explore runs Claude Opus across five report types — market analysis, competitive teardown, pricing strategy, product direction, investment analysis — with multi-stream parallel research so sections work concurrently rather than in series. Every report ships with the same five-part skeleton (executive summary, detailed analysis, supporting data, recommendations, sources and methodology) and exports to an investor-ready slide deck in one click.
- Monitor — continuous competitive intelligence. Validation is a moment; markets keep moving. Monitor tracks competitor websites, pricing pages, product announcements, customer sentiment, and ad creative in the background, then surfaces only meaningful changes through daily, weekly, or monthly briefs. When a competitor changes pricing or ships a feature you were considering, you find out before your customers do — and Monitor signals can trigger a fresh Explore report automatically.
What used to take three to four weeks of founder time now takes under thirty minutes to scope, build, and publish — and the validation doesn't stop at the landing page. The result is the same answer to "is there a real market here?", but the cost of getting the answer, and of staying on top of it, has fallen by an order of magnitude.
How we built it
Smokescreen is, in some ways, our cleanest example of what executive AI engineering looks like in practice.
The platform is structured as three integrated pillars sitting on a shared agent stack. A planning layer interprets the founder's brief and decomposes it into a validation strategy — what to test, in what order, against what bar. A generation layer produces the artefacts: page copy, structure, branding, waitlist questions, research sections, slide decks. An orchestration layer wires those artefacts together — page → waitlist → dashboard → report — and keeps the founder in one continuous flow rather than a tool-hopping scavenger hunt. A measurement layer collects signal from the live page and evaluates it against the pre-set success criteria.
Underneath the pillars, the model choice is deliberate. Claude Sonnet runs the Validate flow — fast, conversational, and cheap enough to keep the unit economics sensible. Claude Opus runs Explore, where the work is research-heavy and the bar for rigour is higher: market sizing with methodology, competitor matrices with sources, pricing analyses that show their work. Monitor sits on a scheduled scraping and diffing pipeline with a signal classifier on top, so the briefs filter noise rather than amplify it.
The whole stack is designed around one principle: agents do the volume, the founder holds the bar. Smokescreen doesn't decide whether an idea is worth pursuing — it makes the test cheap enough that the founder can find out for themselves, with evidence in hand, in an afternoon.
What we've learnt running it
A few patterns have surfaced after running Smokescreen across the xlabs Ventures portfolio, our Labs experiments, and the early cohort of external founders.
The first is that most ideas don't survive a real smokescreen, and that is the entire point. Roughly half of the validations we see produce a clear "no" — soft demand, wrong segment, no willingness to pay. The founders involved are, almost without exception, glad they ran the test before they spent a year of their life building. The remaining half produce either a clean "yes" or a clean "pivot", both of which are also valuable answers. Smokescreen is doing its job whenever it produces a decision the founder can act on.
The second is that hard-signal questions change what the page is worth. A waitlist that asks for an email collects emails. A waitlist that asks how much someone would pay, how soon they need it, and what they're using today collects information about whether the idea is real. The two cost the same to build inside Smokescreen, but they produce very different qualities of insight. We default every waitlist to hard-signal questions unless there's a defensible reason not to.
The third is that validation is rarely a binary. A smokescreen that produces 3% conversion at the wrong price point and 9% at the right one teaches more than a yes/no result ever could. The dashboard, and the Explore reports that sit downstream of it, treat validation as a calibration exercise as much as a gate.
The fourth — and the one that drove us to build Monitor — is that a single validation is a snapshot, not a strategy. Markets move. Competitors ship. Pricing shifts. The founders who win are the ones who keep learning after the landing page goes live, not the ones who treat validation as a one-time event.
Who it's for
Smokescreen is built for solopreneurs, indie hackers, and single founders — the people sitting on an idea who are about to spend the next six months building before they've talked to a single customer. It runs self-serve, globally, on a credits model that scales from a free first build to as many validations as the founder wants to fund.
It powers our internal Ventures and Labs work, and it's wired into xlabs Studio engagements so that validation sits at the front of the build rather than getting bolted on after launch. Every page and every report carries a "Build with xlabs" CTA, so founders who clear the bar and want a team to ship the thing can step straight into a Studio engagement from the same account.
If you're sitting on an idea and the next move feels like "start building", consider that the most valuable next move might actually be: let's see if anyone wants it first.
That's the move Smokescreen makes cheap. It's the move that, more often than not, decides whether the venture lives or dies. And it's the move that started with us being tired of watching good teams build the wrong thing.
We built Smokescreen because we wished, on behalf of every founder we worked with, that someone had built it for them — earlier, and cheaper, and with less excuse to skip.
Now it exists. The excuse is gone.
Join the Smokescreen waitlist → — we're opening early access to founders who want to validate before they build.