We had budgeted two hours. On Friday, July 10, at a startup hub called A2 Hub in Nha Trang, the room kept us for almost three.

Nobody left early. That's the detail worth opening with, because it's the one you can't fake in a slide. Five local, early-stage teams — Pulse, Tarvoo, NOESIS, PhotoRestore, and WildBoar — sat through a full run of EvalLens reading their own pitches, and instead of thinning out, the room leaned in. The overrun wasn't a scheduling miss. It was the point.

Up to now, EvalLens had lived on a screen. This was the first time we took it into a room and let it read live decks in front of the people who wrote them.

What we actually did in that room

We ran the day in one straight line.

Show the product  →  Walk the methodology  →  Hear the pitches
   →  Open the report together  →  Decide as a room  →  Rank and reward

First we showed what EvalLens is. Then — and this is the part most demos skip — we walked the methodology in detail, out loud, before scoring a single deck. Founders got to see the rubric before it saw them. A pitch is read across six dimensions: problem, solution, market, business model, team, and feasibility. Each one gets an evidence-backed read, not a vibe.

Then the teams pitched, and we did the next part together. We opened the report for each deck and worked it as a cohort — the whole room reading the evidence, arguing the strengths and the soft spots, sometimes agreeing with what the system had found, sometimes overruling it. That's human-in-the-loop with the training wheels off: not a machine handing down a verdict, but a room using the machine's analysis to reach its own.

One line held the day together, and we said it more than once:

AI prepares the analysis. People make the call.

The thing we didn't expect: founders answer the questions the slides never showed

Something happened on every pitch. The report would flag a gap — a question the lenses couldn't resolve from the slides — and before we could even ask it, the founder answered it out loud. The market they hadn't sized on the slide, the risk they'd left off the roadmap: they'd talk right through it, filling the hole in real time.

Two things fall out of that, and they point in opposite directions.

The good one: it means the report is finding the right gaps. When a founder instinctively rushes to explain exactly the thing a lens flagged, the flag was real. The read and the room agreed about where the deck was thin.

The awkward one: a chunk of the report's questions get spent live, in conversation, and never make it back into the written record. The deck still doesn't say what the founder said. The next reader — an investor who wasn't in the room — sees the gap, not the answer.

So we're building for it. It went straight into the backlog: capture the live discussion, record the founder's own voice through the pitch and the Q&A, and fold what they said back into the report afterward. The verbal answer stops evaporating. The written read gets richer for the person who reads it cold, next week, without the founder in the room.

The feature the room actually loved

We shipped a small thing into the report before this event: you can look at the deck and the report side by side, and check one against the other slide by slide. We thought it was a convenience. The room told us it was more than that.

Reading a score is abstract. Reading a score next to the slide it came from is not. Watching a founder point at slide four, then at the line in the report that graded slide four, then nod — that's the moment the analysis stops being a black box. It got the strongest usability feedback of the day.

What each founder walked out with

The projects were early and local, and that was deliberate — this was a first live run, and we wanted to stress-test the read on exactly the decks that are hardest to score generously: the ones still missing pieces.

Each founder left with three things, not a gut reaction:

  • A readiness verdict — what could already move an investor, and what to close before the raise.
  • A quality read — strengths and soft spots, split across the six dimensions, so a good story couldn't paper over a thin number.
  • An EvalLens score, 0–10 — one figure, comparable to everyone else who pitched, and settled by the room, not the machine.

The value of splitting a deck into six parts shows up fastest in a room like this. A founder can be strong on team and hollow on market, and a single gut score blurs that into "pretty good." The report refuses to blur it. It tells you which slide is carrying you and which one is quietly costing you the round.

We ranked them. Then we handed out coins.

At the end we built a leaderboard and gave out awards. Every team took home a prize — a cash award with certificates, and, because we couldn't resist, a physical Bitcoin and Ethereum coin each. A small, heavy reminder of the night the room read their decks out loud.

It sounds like a party favor. It's also the honest shape of what we're building: a ranking that a room stands behind, not a number a model spat out.

The part that surprised us more

We came to run a demo. We left with a pipeline.

Three things happened that we didn't script:

A founder asked us to rebuild their pitch with them. Not "send me the report" — let's fix this together, aimed squarely at angel and pre-seed. We're taking them up on it, and packaging the before-and-after as a use case: a real deck, a real score, a real climb. The most honest benchmark we can publish is a founder we actually helped move.

A venture fund signed a letter of intent. A fund representative we walked through the back of the platform — how you spin up an event, load decks, run a batch — asked to test EvalLens on their own deal flow. Starting September 2026, they want to point it at their sourcing pipeline. (The fund asked to stay unnamed for now, which is fair; an LOI isn't a press release.)

The same representative became a customer conversation, not just a spectator. We ran a proper discovery session on the side, and the interest was real enough to keep going.

None of that is a valuation. It's something more useful at this stage: evidence that the thing works outside the building. Founders wanted the read badly enough to ask for more of it. A fund wanted it badly enough to put intent on paper.

Why we ran it here first

Nha Trang isn't where you'd expect a scoring product to debut. That's exactly why we chose it. If the read holds up on early, rough, local decks — the decks with the most gaps and the least polish — it holds up. Starting somewhere forgiving teaches you nothing. Starting somewhere honest tells you whether you built the right thing.

We think we did. But we'd rather show you than tell you.

What's next

We're turning that founder's pitch rework into a published use case, preparing the fund's September pilot, and building the voice-capture feature the room made obvious. And this was the first live room, not the last — the format travels, and we're already looking at the next one.

If you run a hub, an accelerator, or a fund and you want a room of decks read this way — evidence on every claim, a score the room can defend, and the final call still yours — book a demo. If you're a founder who just wants to see how your deck reads before anyone scores it for real, try the live demo and run it through the lens yourself.

We planned for two hours. Next time we'll plan for three.

Common questions

How does a live EvalLens scoring session work? In one straight line: show the product, walk the methodology out loud before scoring a single deck, hear the pitches, then open each report and work it as a room. The cohort reads the evidence together, argues the strengths and soft spots, and sometimes overrules what the system found. AI prepares the analysis, people make the call.

What do founders get from an EvalLens evaluation? Three things: a readiness verdict on what could already move an investor and what to close before the raise, a quality read split across six dimensions (problem, solution, market, business model, team, feasibility), and an EvalLens score from 0 to 10 that is comparable across everyone who pitched. In Nha Trang, that score was settled by the room, not the machine.

Does the AI hand down the final score at EvalLens events? No. The report scores each deck and shows the evidence, then the room decides, sometimes agreeing with what the system found and sometimes overruling it. The ranking every founder took home was one a room of people stood behind, not a number a model produced on its own.

Why did EvalLens run its first live event in Nha Trang? Because early, rough, local decks are an honest stress test, and starting somewhere forgiving teaches you nothing. Five early-stage teams pitched at A2 Hub on July 10, the two-hour session ran for almost three, and nobody left early. The day ended with a leaderboard, prizes for every team, and a venture fund signing a letter of intent to pilot EvalLens on its own deal flow.