For Early Investment, a Great Problem Beats a Great Solution

01 /The Concept

This startup was founded in Denmark earlier this year. It hasn't launched a product yet – the site only offers a waitlist signup. Nonetheless, it just raised €3.6 million in its first round, which was enough to make it worth paying attention to.

Flare's premise: "The digital world is optimized for consumption, but not for understanding." The internet produces an enormous volume of content – and a significant portion of it is, to put it gently, unreliable. Now that AI is generating much of that content, the question of credibility has grown more urgent than ever.

Flare wants to add the "missing layer" between information and understanding – a trust filter built directly into the reading experience. The layer is collaborative, which means it gets smarter with every new contribution.

Some coverage has described Flare as "Wikipedia meets TikTok," though there's no video involved. The analogy captures two things: it's crowd-sourced like Wikipedia but more granular (like TikTok's short-form content), and it surfaces information in small, digestible pieces.

Mechanically, Flare is a browser extension. Verified "contributors" can annotate any content they encounter online. Other verified contributors can upvote or downvote those annotations and add their own.

Ordinary users who install the extension see highlighted sections on whatever they're reading – flagged by contributors as questionable – along with a summary of the debate. Full individual comments are available for anyone who wants to dig deeper.

An AI layer helps contributors by surfacing potentially questionable passages for review, and summarizes the contributor discussion for display to regular users.

Each summary gets a label: "Confirmed," "Contested," or "Unconfirmed."

For example: it's "Confirmed" that magnesium glycinate calms the nervous system at the receptor level, with a brief explanation and links to supporting studies. The claim that creatine makes you smarter is "Contested" – the EU food safety authority couldn't reach a definitive conclusion, though an effect may exist, particularly in older adults. Detox teas that claim to flush toxins from the body? "Unconfirmed" – no scientific studies have produced supporting results.

02 /Why It Matters

Flare has a direct conceptual predecessor that its co-founder acknowledges in a LinkedIn post: Community Notes on X (formerly Twitter), which works on similar principles and now gets help from X's AI model Grok for summarization.

The difference, as one of Flare's investors put it: "Flare is Community Notes for the entire internet" – not just one platform.

The underlying tension is real: some people treat anything they read online as gospel – much as an earlier generation trusted everything on television. Others have swung to the opposite extreme and trust nothing. The truth is somewhere in the middle: some things are reliable, some aren't, and some are reliable with caveats. Flare is trying to make that distinction legible.

Roon ([related review](/review/najdjotsja-ne-vsjo-tolko-dostovernoe)) raised $15 million at the end of 2024 after its founder's personal experience with unreliable health information – he'd searched the internet for information about dementia after his father was diagnosed and was horrified by what he found. The original idea was to create a curated source of verified medical information from licensed physicians. Roon has since pivoted into a professional community where physicians share knowledge with each other.

VuMedi ([related review](/review/prostoj-sposob-otkusit-kusochek-ogromnyh-bjudzhetov)) took the professional community route from the start, building what amounts to a video platform for physicians – only licensed doctors can upload content. Last May it unexpectedly raised $80 million after deciding to accept pharmaceutical advertising directed at physicians.

Dexa ([related review](/review/kak-konkurirovat-s-chatgpt)) raised $6 million for a knowledge base limited to expert sources – indexed podcasts from people it considers domain authorities. Users can ask Dexa anything, but it only searches within its curated knowledge base. You can query all experts at once, or only the ones you personally trust.

A recent example on the news side: SaySo, a news app that includes only video commentary from "verified" influencers. Like Flare, SaySo explicitly frames its goal as comprehension over attention: "We optimize for understanding, not engagement."

03 /Opportunities

AI search modes and general-purpose chatbots like ChatGPT are useful for quick questions and quick answers. But the moment you need to go deeper on any topic, their limitations surface quickly – it turns out there are always multiple competing perspectives, and the AI's version is often incomplete or outdated. You're back to clicking through links, reading, listening, cross-referencing, and forming your own conclusions. Nothing wrong with that – but it takes time.

This creates a felt need for curated, topic-specific "mini-internets" – sources with more reliable authority, better coverage, and fresher information.

Two more recent YC graduates address this directly.

Wizwand compiles and updates a knowledge base of the latest AI research papers on new algorithms – a resource for engineers who need to stay at the edge of the field.

Aemon ([related review](/review/nuzhno-prosto-nachat-jeto-sobirat)) goes further: its AI engineer can take a developer's existing code, find newly published algorithms that could improve it, implement the candidates, run them against test data, and select the best-performing option.

The opportunity space has a few distinct shapes. The most scalable is tools that increase information credibility – either through curation (like Dexa and SaySo) or through collaborative verification (like Flare). The trust layer can be built on top of any source.

More access-controlled but potentially higher-trust are professional communities where only verified experts can publish – removing the problem at the source rather than annotating after the fact (like Roon and VuMedi).

The deepest moat belongs to specialized expert knowledge bases, where AI tooling compounds the value further by extending coverage and keeping content current – the Wizwand and Aemon model.

Where have you personally hit the wall on information reliability, completeness, or currency? That’s your direction.