Gamma bought a team, not a feature.
In 2026, Gamma acquired Lica, an Accel-backed design startup, and instead of folding the product into its own app and calling it a day, it spun the acquisition into a new AI design research lab. Lica’s co-founders will run it. The stated focus is new presentation methods. Terms were undisclosed.
I review AI tools for a living, which mostly means watching companies ship the same three features with different button colors. So an acquisition framed as a research bet gets my attention, even if I’m not ready to hand out credit for it yet.
What Lica actually did
Lica’s pitch was turning screenshots and recordings into presentations and videos. That’s a more interesting input problem than most presentation tools bother with. The standard AI deck flow is: you type a prompt, the model writes bullet points, a template wraps them in a gradient. It works, sort of, and it produces exactly the kind of deck that makes people check their phones.
Starting from screenshots and screen recordings is a different premise. You already did the work. The artifact exists. The tool’s job is to look at what happened and shape it into something a human can follow. That’s closer to editing than generating, and editing is the part AI tools are usually worst at.
Why a research lab and not a feature ship
This is the part worth sitting with. Gamma could have absorbed Lica’s tech, added a “generate from recording” button, and put out a changelog post. Plenty of acquisitions in this space work exactly that way. The feature lands, it gets a demo video, nobody’s workflow actually changes.
Setting up a research division with the acquired founders in charge signals something else: that Gamma thinks the slide format itself is the constraint, not the generation quality. That’s a real position to take. Slides are a 1980s answer to a problem that mostly doesn’t exist anymore. They were designed for a person standing in front of a projector talking to a room. Most decks now get skimmed asynchronously in a browser tab by someone who will never hear you speak.
If you accept that, then making better slides faster is optimizing the wrong thing. You’d want output that adapts to how it’s being consumed — read alone versus presented live, skimmed for one number versus reviewed in detail. Nobody has shipped that convincingly.
My honest reservations
Research labs are also a very convenient thing to announce. They buy time. They give a company a story about the future while the actual product stays where it is. I’ve watched enough of these to know the pattern:
- The lab gets announced with ambitious framing.
- Six months pass with no shipped output.
- The founders’ equity vests.
- The lab quietly becomes a feature team.
I’m not saying that’s what’s happening here. I’m saying the announcement itself tells you nothing about which way it goes, and anyone writing confident predictions off a press release is selling you something. Terms weren’t disclosed, so we can’t even read the price as a signal of how serious the bet is.
There’s also a real question about whether Gamma’s existing users want this. The people paying for AI presentation tools right now largely want faster slides, not a reinvented format. Reinvention creates a learning curve, and learning curves are where tools go to die. Ask anyone who tried to move a team off PowerPoint.
What I’d watch for
If you use Gamma or you’re evaluating it, the acquisition changes nothing about the tool today. Here’s what would actually tell you the lab is doing work:
- Output that isn’t slide-shaped. Not slides with better animations — a different artifact entirely.
- Recording and screenshot input showing up as a first-class flow, not a beta toggle buried in settings.
- Some public account of what the lab tried and what failed. Research groups that never publish negative results usually aren’t doing research.
The most encouraging detail is that Lica’s founders are leading it rather than reporting into an existing product org. Acquired founders who get absorbed into someone else’s roadmap tend to leave within eighteen months. Giving them their own mandate is at least structurally consistent with meaning it.
My read: this is a reasonable bet by a company that’s honest about a real limitation in its own category. That’s more than most AI tool companies manage. Whether it produces anything you’d want to use is a separate question, and I’ll judge that when there’s something to open.
🕒 Published: