The hashtag count is a rearview mirror. By the time it jumps, a story already has carriers, a route, and a mood. The Monday social report can tell a chief communications officer that people are talking. It rarely says who started it, which communities it crossed, or whether it is about to leave a corner of the internet and hit a national feed.
That is the gap Gudea is selling into. The Minneapolis company, founded in 2023, calls itself the “storm tracker for the internet“: software meant to watch a narrative form, not only to total the damage after it lands.
Volume is not movement
Social listening is not a mention counter with a fresh coat of paint. Mature platforms already cluster topics, map influencers, and flag some signs of bots or coordination. They can tell a brand whether a conversation exists and who has the loudest voice in it.
Keith Presley, Gudea’s co-founder and CEO, draws a harder line than the category deserves.
A lot of the chief communications officers and chief media or marketing officers, they’re using legacy tools to try to figure out what’s happening. Those tools were only ever designed to tell you the number of hashtags or number of impressions.
Useful, if the question is whether people are talking. Incomplete, if the question is who is pushing the talk and how much time you have. Calling that gap counting versus thinking misses the point. The gap worth paying for is forecast quality, and evidence of the pathway: which accounts and communities are carrying a story, and whether it is accelerating.
What the software claims to see
Gudea says it maps those routes. As described to Crunchbase News, the system uses deep learning, including graph neural networks, to analyze behavior. A large language model summarizes what the network is doing. A graph model treats accounts and communities as connected nodes, not a pile of posts. Presley’s line for the behavioral signal is plain enough.
Those behavior patterns are like fingerprints.
The forecast is the part a buyer should refuse to take on faith. He has said the software can find a narrative before a team knows which keywords to search, then estimate how long until something spreads widely, how many people it may reach, and how long until it fades. No accuracy rate, false-positive rate, or back-test has been published next to that promise. A forecast without a miss rate is a brochure.
The packaging is less dramatic, and more honest about the buyer. Crunchbase News reported that the company delivers alerts, platform access, and automatic reports. The reports are there because senior teams were unlikely to log in themselves.
An origin story is not a feature
The product story predates the company. Crunchbase News, relaying the company’s account, reported that Gudea grew out of the George Washington Initiative. That nonprofit adapted tactics the intelligence community had used against online networks tied to groups such as Al Qaeda and ISIS, and applied that lens to domestic information campaigns. In that telling, the team went looking for software that could show how information actually moved, and found tools that stopped at volume.
What “disrupt the pathways” meant in that nonprofit work has not been described in public. A communications team should not assume the commercial product inherited an influence or disruption capability. Nothing on the record shows that it did. The product as described analyzes and warns. It does not interfere.
The privacy question is already live. Mapping people, groups, and relationships is the current claim. It is not a future extra. A buyer should ask what data builds that graph, whether individuals can be identified, who inside the company can see it, and how long it is kept. Presley has also described a planned “geolocation for narratives” feature, intended to locate conversations down to the city-block level and slated for the following year. The company has said it wants customers to be able to run an advertising campaign through the model before launch, to gauge whether a strategy is likely to draw a negative response. Neither is a shipped, validated feature. City-block location would sharpen the privacy problem. It does not create it.
Cheap posts make volume a worse signal
Generative AI lowered the cost of keeping a narrative alive. Work that once needed an organization can be attempted by one person with a laptop and the right software. Presley has offered a personal tally, hedges included.
When we first started in 2021, and then as a company in 2023, we maybe saw one out of four things we looked at have inauthentic activity around it. Now it’s probably nine out of ten things that we look at have this inauthentic activity.
There is no public definition of inauthentic activity and no published method behind the ratio. Treat it as a founder’s impression. It is not a measured rate. The business point does not depend on it. If synthetic posts are cheap, a thousand mentions can be a crowd, a script, or both. Network behavior is harder to counterfeit at scale than a paragraph. That is the bet.
What a miss looks like
A false alarm pulls an executive into a fringe fight they would have been wiser to ignore. The response becomes the story. A late alarm is the other expensive error. The alert arrives when the Monday mention report would have caught the same spike, and the lead time you paid for was never there.
Without a record of both errors, on incidents like yours, you cannot tell which way a model leans. Ask for that history before you sit through a demo of a win.
What is worth keeping
Reputation trouble often moves as a network event. Waiting on impression counts is a way of volunteering to be late. The question is worth keeping even if you never sign a contract: who is moving this, along which routes, and how fast?
Know which fringe spaces matter in your category. Decide in advance who may respond, and how quickly. Separate “people are angry” from “a coordinated pattern is moving this.” Do not let a model’s confidence replace someone who understands the business.
“These big industries don’t understand that there are things they can do proactively to defend their reputation and their brand, ” Presley has said. That is a fair complaint about management habits. It is not, by itself, evidence that a forecast works.
A storm tracker is only as good as its false alarms. Meteorologists publish those. Narrative-intelligence startups, so far, mostly publish the pitch.
A seed round, and a total that does not reconcile
TheSaaSNews reported that on October 8, 2026, Gudea announced a $7 million seed round. Mudita Venture Partners is a co-lead, as reported by TheSaaSNews and Crunchbase News. The second co-lead is disputed (Silicon Road Ventures in one account, Collide Capital in another), so check the live record before treating either name as settled.
Presley told Crunchbase News that total funding since the 2023 founding is about $15 million. That figure does not reconcile, on the public record, with the announced seed. Earlier capital is not itemized, so the total remains a founder number. The company says the new money goes to hiring, new features, and wider coverage across languages, regions, and platforms. It also says it has 39 customers and about 25 employees. Revenue, it says, is up 357% year over year. No baseline was disclosed, and the customers are unnamed. Crunchbase News, citing the company, described them as major AI companies and businesses in telecommunications, entertainment, and retail. The percentage shows a line rising from an unknown start. It does not show how large the business is.
Questions worth asking first
- How is this different from social listening?
Mature listening tools already cluster topics, map influencers, and flag some coordination. Gudea’s claimed difference is a forecast of who is carrying a story and where it goes next. No accuracy or false-positive figures have been published to test it.
- Is the prediction validated?
No. The company says it can estimate time to wide spread, how many people may be reached, and how long until a story fades, sometimes before a team knows which keywords to search. It has not published accuracy, false-positive, or back-test numbers.
- What should a communications leader demand?
Historical lead time on incidents like yours, including the misses, and a plain account of the data behind any graph of people and groups. Cost, data sources, and false-positive history were not disclosed in the public reporting. A vendor should also state in writing that the dashboard did not inherit a disruption capability from earlier nonprofit work.