A crowd of wooden blocks of outlines of people with chat bubbles above their heads
An arrow pointing leftHome

The Seattle startup building an AI that will read the comments so you don’t have to

  • Justin Caffier

Unwrap just raked in $3 million in seed funding.

One of the first things you’re warned as a fledgling writer in the digital media space is “don’t read the comments.” Though a particularly important self-care mantra for those sending creative works or intimate thoughts into the digital void, the phrase was initially coined as a cautionary rule for every visitor to every corner of the web.

The underlying (and far-too-often correct) presumption that an open online forum will invariably devolve into a toxic cesspool elevated “don’t read the comments” from early web forum inside joke to ubiquitous truism. Now widely regarded as “the first rule of the internet,” a designation cemented by its utterance in 2018’s Wreck-It Ralph 2, comment avoidance as a best practice has never been more commonplace.

And yet, as prudent as this advice might be for individuals hoping to avoid hate speech, trolls and other unnecessary psychic tolls, there’s no denying that we still want to read the comments. It’s not just masochism, narcissism or forbidden fruit tasting sweeter. It can be good for business. Free online product feedback is often worth its weight in gold to a company.

So when news site comment sections began shuttering left and right in response to increasingly heated political discourse of the past few years, many feared there was too much baby in the bathwater being thrown out. Comment defenders assert that, when properly moderated and analyzed, this discourse offers a treasure trove of helpful insights and data. Sifting through the chatter to find those pearls has always been the hard part. But a new AI-powered natural language processor called Unwrap is hoping to make that task a little less daunting.

Co-founded by Ryan Millner and Ashwin Singhania, who previously worked as product managers in Amazon’s Alexa AI engineering department, Unwrap just enjoyed a $3 million seed funding round after completing Seattle’s Allen Institute for AI incubator program.

Both frustrated by the inefficient slog that was sifting through not just comment sections, but emails, bug reports, subreddits, social media, message boards and review pages for ideas on how to improve Alexa, Unwrap’s co-founders realized there had to be a better way to harvest feedback. Rather than Clockwork Orange-ing their eyes open in front of an auto-scrolling feed, they decided to outsource the drudgery to something incapable of getting bored or overwhelmed: artificial intelligence.

Unwrap purportedly works by aggregating all the human-provided text feedback from the myriad relevant channels and then processing it through AI-powered software that can decipher and understand all our informal typing practices and webspeak quirks, extracting morsels of useful information about a particular product and consolidating the findings in one place. From there, the Unwrap user can search through the data, automatically categorize feedback and even detect anomalies or trends.

In an email to GeekWire, the company cagily shared that it has “between five and 10 paying customers,” so there is an ironic dearth of Unwrap feedback for us to currently pore over. Only with further proliferation of the service will we see how the intrinsically irrational and emotion-driven humans on the other side of the screen handle Unwrap’s soberly synthesized findings. The company seems confident, however, that a streamlined gathering process offers less-burdened product managers the mental clarity for more shrewd interpretation and implementation of that product feedback.

We’d love to give five stars for Unwrap’s concept alone, but the tech industry is littered with the corpses of great ideas in theory that never panned out in practice. For the time being, we’ll leave it at “review forthcoming.”