Are the Growth Numbers in Press Releases Even Real? · Issue #97

Opening

Hey subscribers, this is Oz’s Knowledge Talking.

I want to open today with an uncomfortable question. Every day we run into press releases packed with phrases like “500% year-over-year growth,” “reached the J-curve inflection point,” or “monthly active users cross the X hundred-thousand mark.” We usually have no idea where the denominator behind these numbers comes from, or how they were measured. There’s no way to verify any of it at the moment of announcement. And every year, when audit season rolls around, some companies go quiet — or suddenly redirect attention with unrelated PR.

This pattern isn’t unique to Korea. IRL, a US social media startup that raised $170 million, bragged about 20 million users — a board investigation later found 95% of them were bots. China’s Luckin Coffee recognized $300 million in fabricated revenue in a single year, 2019. In both cases, no immediate verification mechanism kicked in until the SEC caught it — after the fact.

Recently, an interesting company showed up claiming to target exactly this gap: Objection. Backed by Peter Thiel and Balaji Srinivasan, it lets anyone dispute the factual claims in an article, podcast, or YouTube video for $2,000 a pop. The verdict gets published as an ‘Honor Index’¹ score, visible to both the outlet and the reporter.

On the surface, this sounds compelling. A verification infrastructure has finally shown up for a market that’s been plagued by false PR/IR claims. But the deeper you look into the market design details of this tool, the more the contradictions surface. The very same mechanism works just as naturally in the opposite direction. Today, let’s unpack that gap.

📊 The Verification Gap in PR/IR — A Pattern That Repeats Every Time

First, let’s establish the market gap itself. When I was doing GTM strategy consulting, I ran into this pattern constantly: performance metrics in PR/IR materials get published with no measurement standard attached. Frankly, this kind of PR does more harm than good — it gets exposed quickly anyway, especially in a world where financials and active-user counts get refreshed on statistics sites monthly, sometimes weekly, during earnings season. Empty bravado like this doesn’t survive long.

There are a few recurring archetypes of this behavior.

Growth rates with an undisclosed denominator: going from one person to two is technically 100% growth.
GMV² with a fuzzy definition: if you don’t disclose how cancellations, refunds, and internal transactions are treated, the same business can look two or three times bigger or smaller.
Traffic inflated by one-off events: pour ad spend into a single month, generate a traffic spike, then announce “X hundred-thousand monthly users.” These numbers get quoted verbatim by media outlets and recycled as the basis for the next funding round’s IR materials.

Does a few companies already come to mind? Let’s keep going. Every year, as audit season approaches, two patterns emerge: companies go quiet, or new PR appears to redirect attention elsewhere — a new service launch, an overseas expansion announcement, a new partnership. It’s a deliberate exploitation of the timing when audit results are delayed or relatively buried.

Overseas cases are even more extreme, but the structure is identical. IRL claimed 12 million users and raised $170 million, at one point earning a valuation of $1.17 billion. A board investigation found that 95% of the reported 20 million users were bots and automated accounts. The SEC charged the founder with fraud in 2024. Skael, a San Francisco SaaS startup, raised over $30 million over five years and was caught by the SEC for inflating its revenue figures. A former CEO of a Florida ad-tech company posted on social media that “company revenue is $10–20 million,” when actual 2021 revenue was $17,450 — a gap of nearly 1,000x. And then there’s Theranos, which everyone knows.

What do all these cases have in common? No immediate verification mechanism kicked in until the SEC caught it after the fact. Somebody needs to fill this gap. That market clearly exists.

⚖️ Objection’s Mechanism: Mimicking an Adversarial Courtroom

The company that showed up claiming to answer this gap is Objection. Its founder, Aron D’Souza, is the lawyer who represented Hulk Hogan a decade ago in the lawsuit that bankrupted the media company Gawker. His diagnosis is blunt.

“American trust in media has fallen from 68–72% in 1972 to 28% in 2025. Somebody has to be held accountable.”

For reference, the “1970, 70–80% → 30%” figure D’Souza cited in interviews is inaccurate. I searched extensively and couldn’t find a source for it. The raw Gallup data I found is above (see the References section).

The system’s structure mimics two proven truth-discovery mechanisms: the adversarial courtroom system, and the scientific method.

Adversarial structure: an objective challenger (the plaintiff) and the reporting party (the defendant) submit evidence, and an ‘AI jury’ made up of five LLMs³ (OpenAI, Anthropic, xAI, Mistral, Google) renders the verdict. Each model is assigned a different American demographic persona (a Brooklyn man in his 50s, a Portland woman in her 20s, etc.)
Reproducibility: every verdict process and piece of evidence goes into a public data room, and the algorithm and whitepaper are published on the site.
Tiered evidence grading: scored on a five-tier scale from Grade 1 (courtroom-level primary sources) to Grade 5 (rumor).
Human investigators added: former CIA, FBI, and MI6 investigators personally call the quoted sources to confirm.

On top of this sits an add-on feature called ‘Fire Blanket’⁴. Hooked into the X platform API, it automatically slaps a real-time “under investigation” label onto disputed reporting — a device that lets you cast doubt on credibility before any verdict is even reached. Think of it as an automated version of Community Notes.

Up to this point, it looks like a plausible tool for closing the PR/IR verification gap. What if a VC files an objection against a suspicious revenue figure announced by a competitor in their own portfolio? It’s far faster and cheaper than a post-hoc SEC investigation — $2,000 per case, with results in a matter of days.

Here’s where the problem starts. Depending on who uses the same tool and for what motive, the outcome flips completely.

🪤 The Shadow Side of the Same Mechanism: Asymmetrizing Anonymous Sources

Look closely at Objection’s scoring rules, and one thing stands out decisively: reporting that relies on an anonymous source⁵ automatically receives a lower evidence score. Primary sources (regulatory filings, official emails) sit at the top; anonymous whistleblower testimony sits at the bottom.

This forces a journalist with a protected source into one of two choices. First: submit the source’s identifying information into Objection’s “encrypted hash” system. D’Souza says “an independently verified anonymous source can receive a higher score” — but that verification requires the source’s information to enter an external system in some form. Second: accept the score drop. Your Honor Index takes a hit, and Fire Blanket slaps an “under investigation” label on the story over on X.

Jane Kirtley, a media law professor at the University of Minnesota, went a step further in an interview and cut to the core of it: “This seems far more interested in giving those who already hold power a tool to pressure their journalistic adversaries, than in providing useful information to the general public.” Chris Mattei, a First Amendment⁶ attorney, was even blunter: “It looks like a high-tech protection racket for wealthy, powerful people.” Put simply: if Objection acts in bad faith to extract or distort information, who’s there to stop it? Peter Thiel, is it you again?

This brings to mind the Pentagon Papers, Watergate, and the Theranos exposé mentioned earlier. Wait — think about Theranos again. The Theranos lies were first cracked open by John Carreyrou’s investigative reporting for the Wall Street Journal in 2015, and that reporting relied absolutely on anonymous whistleblowers. Theranos immediately mobilized over $1 million in legal resources to pressure both the reporting and the sources. If a system like Objection had existed back then — a system where a single factual claim gets an “under investigation” label within days, for $2,000 a pop — what score would Carreyrou’s reporting have received? And what if Holmes’s side had used the same system to file objections against every factual claim in the story?

This is where the real asymmetry in the market design shows up. The PR/IR verification market and the market for suppressing critics of the powerful run on the exact same price, the exact same mechanism. The tool’s designer cannot separate the two markets.

📐 The Pricing Design, Viewed Through a GTM Lens

The pricing design says it all. $2,000 per case. D’Souza justifies it as “the hourly rate of a top New York law firm attorney.” But this price doesn’t function identically across the two markets.

For an ordinary American, $2,000 is not an “accessible” price — it’s roughly a month’s survival budget for a low-income household. For a Big Tech company annoyed by a piece of PR, it’s the price of a nice lunch. Same for a hedge fund manager irritated by a story. D’Souza’s own comment in an interview — “if accessibility is really an issue, I’ll personally fund free subscriptions out of my own pocket” — actually exposes the essence of the pricing structure. Accessibility that depends on a founder’s personal charity, rather than the system’s actual design, is, by design, no accessibility at all.

One more interesting detail: D’Souza defined the market size as “roughly 150,000 people who tend to be the subject of reporting.” That’s who he pictures as his core customer base. Not the ordinary citizens being reported on, not readers who want verification, not whistleblowers. The people being reported about.

From a GTM⁷ perspective, every signal points in one direction. The backing investors (Thiel, Srinivasan, Social Impact Capital, Off Piste Capital), the price of entry, how the market size is defined, the objection structure that breaks disputes down into individual factual claims (allowing a long investigative piece to be attacked claim by claim), and Fire Blanket’s automatic labeling feature — all of it. This is a market design that would have made far more sense if positioned as a “PR/IR fraud verification tool.” Instead, the surface-level message is “journalistic accountability.” That gap reveals this company’s true identity.

Oz’s Lens

Here’s how I read this market design.

The market gap Objection points to — stated generously, the absence of fast, cheap fact-verification infrastructure for claims made by the powerful — is real. In my GTM consulting work across the Korean and global startup ecosystems, I’ve watched this gap generate massive amounts of misallocated capital every single year. Growth rates with no measurement standard, KPIs with no definition, PR with no verification — they all flow straight into the next funding round as supporting evidence. Somebody has to solve this.

The problem is how D’Souza solved it. He succeeded in lowering the cost curve of verification, but in doing so, he transplanted a power asymmetry directly into the act of verification itself. If PR/IR verification and whistleblower intimidation run on the same price and the same mechanism, the system will inevitably get used far more for the latter. Because that market is bigger, wealthier, and more highly motivated. Markets always flow toward whoever is willing to pay the highest price.

If I were designing this market, I would have split the two apart. I’d have narrowed the tool to measurable, quantitative claims only — disclosures, financials, user metrics, revenue recognition — and explicitly excluded investigative reporting based on anonymous sources. I’d have tiered the pricing differently, too. That way, you solve the real market gap of verifying false PR/IR claims without compromising whistleblower protection. D’Souza’s failure to make that split isn’t a technical limitation — it’s a choice about how to define the market. And that choice determines this company’s identity.

Closing

Let me sum up today in three lines.

First: growth rates with no denominator, KPIs with no definition, PR timed to dodge audit season — there are simply too many unverified claims embedded in the press releases we encounter every day. This is a structural gap shared by every global media market, not just Korea’s.

Second: Objection showed up claiming to target this exact gap, but its actual market design is calibrated closer to a tool for the powerful to fight back against criticism. The pricing, the market definition, and the handling of anonymous sources all point in that direction.

Third: when the same mechanism can be used for opposite purposes, the details of market design determine the tool’s true identity. A good verification tool needs to define not just what it can verify, but what it explicitly excludes from verification.

The next time you come across a press release, I’d ask you to check just one thing with me. Where is the denominator behind this number disclosed? What was the measurement method? Does the definition match what the same company reported in the same quarter last year? These three questions are the starting point of all verification. And building the infrastructure to answer them is something somebody eventually has to do. I just don’t think Objection’s approach is the answer. What do you think?

References & Further Reading

Primary sources

Rebecca Bellan, “Can AI judge journalism? A Thiel-backed startup says yes, even if it risks chilling whistleblowers”, TechCrunch, April 15, 2026. The most balanced analysis piece covering Objection’s launch and critics’ reactions.
Rebecca Bellan, “Full transcript: Conversation with Aron D’Souza on Objection and AI in journalism”, TechCrunch, April 15, 2026. The full transcript, containing D’Souza’s logic and the interviewer’s pushback verbatim. If you want to understand Objection’s thinking directly, I’d recommend this over the article above.

Background

U.S. Securities and Exchange Commission, “SEC Charges Founder of IRL Social Media App With Defrauding Investors of $170 Million”, August 2024. The official release on the IRL case, where a claimed 12 million users turned out to be 95% bots — a textbook case of PR/IR false claims going unverified until the SEC investigation stage.
John Carreyrou, Bad Blood: Secrets and Lies in a Silicon Valley Startup, Knopf, 2018. The full account of the Theranos exposé. It shows how investigative reporting reliant on anonymous whistleblowers withstood pressure from a company that mobilized over $1 million in legal resources — and why that protection matters fundamentally. (Personally, I’d recommend the TV adaptation too.)
“Improper revenue recognition tops SEC fraud cases”, CFO Dive. A rundown of revenue manipulation patterns, including Luckin Coffee’s $300 million in fabricated revenue and Marvell Technology’s revenue pull-forward.
Gallup, “Trust in Media at New Low of 28% in U.S.”, October 2, 2025. The 30% figure D’Souza cited is inaccurate. The accurate, most recent figure, measured in September 2025, is 28%. When first measured in 1972, it stood at 68–72%.

The author, Kwangseob Ahn, is a professor of business administration at Sejong University and lead consultant at OBF (Oswarld Boutique Consulting Firm). He teaches statistics and data analysis — business data management and business analytics — while leading GTM and AI strategy consulting in the field, designing the seam between technology and business. He has published academic research on a memory architecture for AI dialogue systems (HEMA) and runs Daily Arxiv, a daily curation of global AI papers. He holds a master’s from Korea University’s Graduate School of Technology Management and a KMBA. He is the author of Homo Brainless: The People Who Outsource Their Thinking.

Footnotes

Honor Index: the scoring system Objection introduced. It renders the accuracy, integrity, and cumulative track record of each journalist and outlet as a number — functioning for journalists somewhat the way Yelp’s star ratings function for restaurants. ↩
GMV (Gross Merchandise Volume) refers to the total value of goods transacted on an online platform or e-commerce site over a given period. It represents the total amount transacted through the platform before fees or returns are deducted, and is often used as a core metric for evaluating a company’s scale of growth and a platform’s influence. ↩
LLM (Large Language Model): large-scale language models like ChatGPT and Claude. Objection assigns five different LLMs distinct American demographic personas (e.g., a Brooklyn man in his 50s, a Portland woman in her 20s) to form its ‘AI jury.’ ↩
Fire Blanket: an add-on feature of Objection. Hooked into the X platform API, it automatically attaches a real-time “under investigation” label to disputed claims — a device that lets credibility be called into question before any verdict is reached. ↩
Anonymous Source (Confidential Source): an information provider whose identity a journalist has promised to protect. This typically includes whistleblowers facing retaliation risk, or informants reporting on abuses of power. Major American investigative reporting — the Pentagon Papers, Watergate, the Theranos exposé — has relied on this. ↩
First Amendment: the freedom of speech, press, religion, and assembly guaranteed by the US Constitution. Under the precedent set in New York Times v. Sullivan, defamation lawsuits in the US must meet the “actual malice” standard to succeed — a bar that’s practically impossible to clear, which is why American media enjoys such broad reporting freedom. ↩
GTM (Go-To-Market): the strategy for how a product or service is launched and spread in the market. It integrates pricing, channels, target customers, and messaging. In this piece, “the GTM lens” refers to reading a product’s true intent through the details of its market design, rather than its surface-level messaging. ↩