Manipulation does its quiet work everywhere — the confident lie that moves a market, wins an argument, rewrites a war. So we set out to build the simplest thing that pushes back: a way to check what you're reading, right where you read it, in a single swipe.
It begins with reading. Then hearing. Then seeing. One honest answer at a time.
False stories aren't a glitch in the feed — they're a permanent feature of how we read now. Social, news, messaging: none of them come with a way to ask the simple question, "wait, is this actually true?" And the cost of not asking shows up at two moments:
You don't want to be deceived. False information means decisions — financial, reputational, legal — made on false premises.
You don't want to deceive others. Forwarding a fabrication destroys trust with clients, audiences and colleagues.
The cost of producing a convincing fake collapsed to seconds. At the same time, the cost of verification fell just as fast — inference prices drop roughly 50% a year, putting real-time claim-checking within reach of a phone. In 2025, social media passed television as the leading news source in the US. A single fabricated tweet from a compromised AP account in 2013 erased about $136 billion in market value in minutes — with a reach of thousands. Today the same fake reaches millions in seconds.
Sources: Pew Research Center (Sept 2025) · Ofcom Online Nation 2025 · Reuters Institute Digital News Report 2025.
You're scrolling. Something stops you — a headline, a number, a claim — and you're not sure. Swipe the lens over it. VeRa reads what it says, checks it against real sources, and tells you what holds up. Seconds, not an afternoon of open tabs.
Drag VeRa's lens over any post — feed, chat, news site, anywhere.
On-device OCR extracts the claim. The screenshot never leaves your phone.
Live source retrieval, a sourced verdict, a confidence score.
No black box. Every verdict comes with the sources it stands on — read one, then keep scrolling.
"The Stockholm3 blood test outperforms standard PSA testing for prostate cancer."
Genetic markers, proteins and clinical data predict aggressive-cancer risk in men aged 45–74.
"A viral post claims NASA announced six days of total darkness from a solar storm — no such announcement exists."
NASA issued no statement of the kind; the claim recycles a long-debunked hoax and has no basis in solar physics.
"Headline claims the central bank 'admitted' the currency will collapse — the source is an opinion blog quoting no official statement."
The official transcript contains no such admission; a conditional scenario from an analyst interview is presented as the regulator's own position.
"The claim lacks concrete details to verify."
Missing date, specific location and incident scope; no source link to a specific event.
| Client | Flutter + Kotlin — Android overlay (MediaProjection) |
| OCR | Google ML Kit — on-device; the screenshot never leaves the phone |
| Source retrieval | Tavily — real-time, verified sources |
| Engine | Qwen3.6 + Ollama — claim analysis and verdict synthesis |
| Backend | Node 24 — SSE streaming, verdict cache, CI-gated test suite |
| Deploy | Cloudflare Tunnel — live HTTPS, tester distribution |
One data structure for any content type — the same interface will serve text, voice and video. With zero credible sources, the verdict downgrades to UNVERIFIED by design; a high-confidence fabrication is architecturally impossible. Temporal grounding, news-existence calibration and language-mirroring are built in — the verdict answers in the language of the content.
The same swipe that reads a headline today will one day hear a voice on a call, and watch a face in a video. We're honest about it: this is the road we're on, not a shelf of dated promises. One engine, one verdict — new senses as they're ready.
Competing tools require switching apps. VeRa is embedded in any context via a system overlay — the barrier to verification is one gesture.
Screenshots stay on-device; only extracted text is checked. Minimises the legal surface as an architectural choice, not a label.
An internal eval harness with regression cases guards every change. Zero sources → UNVERIFIED — never a confident fabrication.
Each opt-in verdict enriches the dataset → better accuracy → broader reach → more data. Compounds with scale.
One founder, orchestrating a small army of AI agents. What a team of three engineers would spend a quarter building was delivered solo in seven weeks — mostly nights and weekends.
Maksym Pietukhov — Ukrainian founder based in the UK. A retail trader who built VeRa after making trading decisions based on news that turned out to be fabricated. Background across media, advertising and PR; self-taught designer — the VeRa brand system is his work. He conceived VeRa and built it end-to-end by orchestrating a multi-agent AI development pipeline.
I'm from Ukraine. All my life I watched Russia use lies as a weapon against my people — decades of manipulation, and now a full-scale war where disinformation hits alongside artillery.
Fighting it is not a market opportunity I found; it is personal. I built VeRa because I want a world with fewer fakes, where people decide on facts. VeRa doesn't take sides — it checks the claim, whoever made it, and shows you the evidence.
— Maksym Pietukhov, founder, RedMindsys