The Contradiction at the Center

On December 5, 2025, The New York Times filed a copyright infringement lawsuit against Perplexity AI, alleging “large-scale, unlawful copying and distribution” of millions of its articles. The complaint described a system that scraped paywalled stories, circumvented security measures, and synthesized the Times’ reporting into AI-generated summaries that made the original journalism unnecessary to read.

Three months earlier, Perplexity had closed a $200 million funding round at a $20 billion valuation.

Put those two facts next to each other and you have the central tension of the most ambitious search startup since Google. Perplexity makes money by organizing, synthesizing, and presenting other people’s work. It does this better than anyone. So well that the people whose work it organizes want to shut it down. And at such scale, 780 million queries a month with 340% year-over-year growth, that shutting it down may no longer be an option.

Think of it as a parasite trying to become a symbiont before the host calls the exterminator. Perplexity feeds on the open web. It needs publishers to keep producing the content it summarizes. And it is racing to build a revenue-sharing framework that convinces those publishers the relationship is mutual before the courts decide it is theft.

Whether Perplexity is the future of search or the most well-funded copyright violation in technology history depends on questions that neither the market nor the legal system has resolved. The $20 billion valuation is a bet on one answer. The lawsuits from the Times, News Corp, the BBC, Encyclopedia Britannica, the Chicago Tribune, and publishers in Japan are a bet on the other.

The Man Who Worked Everywhere

Aravind Srinivas does not use pitch decks.

When he raises money, he opens a laptop and runs the product. He did this in September 2022, two months after incorporating Perplexity, sitting across from some of the most sophisticated AI investors in Silicon Valley. The seed round closed at $3.1 million. The investor list read less like a cap table and more like a guest register at the frontier of machine intelligence: Yann LeCun, Meta’s chief AI scientist. Andrej Karpathy, who would go on to coin the term “vibe coding.” And Ashish Vaswani, who in 2017 had co-authored “Attention Is All You Need,” the paper that introduced the Transformer architecture, the foundational technology underneath every large language model in existence.

Vaswani, co-inventor of the technology that powers Google’s own AI, put his personal money into a startup that intended to replace Google Search. The irony was not accidental. It was a thesis.

Srinivas grew up in a middle-class family in Chennai, India, graduated at the top of his class in computer science at IIT Madras, and earned a PhD at UC Berkeley under Pieter Abbeel. What makes his background unusual is not its excellence, which is common enough among AI founders, but its breadth. He worked at Google Brain, interned at DeepMind, and served as a research scientist at OpenAI, where he contributed to DALL-E 2. Three of the four most important AI research labs on earth, on his resume before his thirtieth birthday.

This matters because of what he saw at each one. Google had the models, the data, the infrastructure, and the distribution to build an AI-powered search product that would have made Perplexity unnecessary. But Google also had $175 billion in annual advertising revenue that depended entirely on people clicking links. An AI that answered questions directly, that eliminated the click, was a gun pointed at Google’s balance sheet. The better the AI, the bigger the hole.

OpenAI had the model capability but was building a general-purpose chatbot, not a search engine. Meta was focused on open-source models and social networking. DeepMind was pursuing scientific research and game-playing AI.

Nobody was building the obvious thing: a product that took the most powerful language models in the world and pointed them at the task that a billion people perform every day.

“If you search ‘pizza near me,’ the result is often whoever paid the most, not what’s actually best,” Srinivas has said. The observation is almost offensively simple. And it describes a $175 billion market.

He brought in Denis Yarats from Meta’s AI research lab as CTO, Johnny Ho from Quora and Tower Research Capital as Chief Strategy Officer, and Andy Konwinski, co-founder of the $43 billion data analytics company Databricks, as President. Later, Dmitry Shevelenko joined as Chief Business Officer, bringing a career arc that ran from Uber’s business development team through LinkedIn’s product organization. Shevelenko would become the person sitting across the table from publishers who wanted to sue the company, negotiating revenue-sharing deals while the legal department fended off copyright claims. In September 2025, Lazard, the 177-year-old investment bank, appointed him to its board of directors, a signal that the financial establishment took Perplexity’s business strategy seriously even if the media establishment did not.

They launched on December 7, 2022. Three weeks later, ChatGPT debuted and the world lost its mind about generative AI. Perplexity was already live. In January 2025, with characteristic audacity, Srinivas submitted a bid to merge with TikTok’s US operations during the app’s forced-sale crisis. It did not materialize. But it revealed something about the founder’s operating theory: when a distribution asset becomes available, bid for it, regardless of whether you can afford it. The instinct would recur.

Srinivas releases products that are, by his own admission, “80% perfect.” Ship fast, watch users, iterate. “Number one skill you need as a CEO is to learn to make decisions,” he told an audience at UC Berkeley. At 31, he is India’s youngest billionaire, with an estimated net worth of $2.5 billion, a fact he does not discuss but that tracks with the company’s trajectory. He describes Perplexity’s long-term vision not as a search engine but as “a general-purpose, reliable personal assistant that removes friction from daily life and unlocks deeper curiosity in humans.” The word “curiosity” appears repeatedly in his public remarks. The company’s stated mission is “to serve the world’s curiosity.” Srinivas, by all appearances, means this without irony.

What Perplexity Actually Does to Your Brain

The difference between Perplexity and Google is easier to feel than to explain, so consider a specific task.

You are writing a memo on the economic impact of AI on the legal profession. On Google, you type the query and get a page of results. A McKinsey report behind a registration wall. A law journal article with a paywall. A Forbes listicle. A Reddit thread. A blog post that is actually an advertisement for legal software. You open seven tabs, skim each one, close three that are irrelevant, copy key data points from the remaining four into a document, and spend twenty minutes synthesizing them into a coherent paragraph. The process works. It is also tedious, fragmented, and inefficient in a way that has become invisible through decades of familiarity.

On Perplexity, you type the same query and get a four-paragraph response with eight numbered citations. It pulls the McKinsey data, the law journal findings, and a Goldman Sachs analysis you had not thought to search for. Each claim links to its source. You read for ninety seconds, verify one citation that looks surprising, and paste the synthesis into your memo. The twenty-minute workflow collapsed to three minutes.

What changed is not just speed. It is cognition. Google trains you to think in links. You learn to evaluate URLs, preview snippets, and guess which results are worth clicking. It is a skill, and a generation of knowledge workers has mastered it so thoroughly that they have forgotten it is a skill. Perplexity removes the need for it. The system does the link evaluation, the source triangulation, and the initial synthesis. You start at the point where Google makes you stop.

Now try a second query, one designed to be harder: “What percentage of law firms have adopted AI tools for document review as of 2025?” Perplexity returns a confident response with three citations. The number it gives is 44%. You click the first citation. The linked article says 35%, and it is from 2024, not 2025. The second citation leads to a consulting firm’s press release that uses the phrase “nearly half” without providing a figure. The third citation is a dead link. Perplexity synthesized a specific number from sources that did not contain that specific number, then cited them as if they did.

This is the failure mode that matters. Not the spectacular hallucination, the glue-on-pizza absurdity that makes headlines and gets fixed quickly. The subtle one. A number that is close enough to be plausible, supported by citations that look authoritative but do not actually say what the summary claims they say. You would have to click three links and read carefully to catch it. Most people will read the confident paragraph, see the blue citation numbers, and trust it.

Google’s unreliability is honest: it gives you raw materials and forces you to think. Perplexity’s unreliability is polished: it gives you a finished product and trusts you to doubt it. Which failure mode does more damage at scale is not an academic question. It is a product design decision with consequences for every person who uses AI to make decisions about money, health, law, or policy.

The Machine Underneath

Perplexity calls itself an “answer engine,” a term chosen precisely to draw a line between what it does and what Google does. Search engines return links. Answer engines return answers.

Under the hood, the technology is retrieval-augmented generation, or RAG, a term that sounds like plumbing but describes one of the most consequential architectural patterns in modern AI. Perplexity’s implementation works as a pipeline. A query enters the system. The intent parser determines what the user actually wants, distinguishing a factual lookup from a causal explanation from a comparative analysis. The retrieval layer, built on Vespa AI, pulls relevant passages from across the web with sub-second latency, crawling news sites, academic papers, government databases, and the general web. A reranking model scores the retrieved documents for relevance and reliability, weighting primary sources above secondary summaries, peer-reviewed findings above blog posts. Finally, a large language model reads the top-ranked passages and synthesizes them into a response with inline citations, every factual claim tied to a numbered source.

That last step is where Perplexity’s product philosophy becomes tangible. When Google’s AI Overviews launched in 2024 and infamously told users to put glue on pizza, the absurdity went viral because nobody could trace where the bad information came from. Perplexity’s citations are a trust mechanism: click the number, read the original. If the answer is wrong, you can see why.

Underneath the pipeline, Perplexity routes queries across multiple language models. Simple factual lookups go to lighter, faster models. Complex analytical queries route to Claude 4.5, GPT-5.2, Gemini 3 Pro, or Perplexity’s own Sonar family. Pro subscribers can override the routing and choose their model manually.

Sonar deserves attention because it represents Perplexity’s play for vertical integration. Built on Meta’s open-source Llama 3.3 70B architecture and fine-tuned specifically for search tasks, Sonar is Perplexity’s attempt to own its foundation model layer rather than renting it from competitors who also compete in search. The most politically charged variant is R1 1776, which Perplexity created by taking DeepSeek’s R1 model, a Chinese reasoning model with baked-in CCP censorship, and retraining it on 40,000 multilingual prompts covering over 300 sensitive topics. The result: 100% uncensored on evaluation benchmarks versus 85% for the original, with no performance degradation on reasoning tasks. Perplexity named it after the year of American independence, open-sourced it under an MIT license, and let the internet draw its own conclusions.

The Money

There is an old venture capital maxim that the most dangerous companies are the ones growing so fast that their problems haven’t caught up with them yet.

Perplexity hit approximately $35 million in annualized recurring revenue by mid-2024. By March 2025, it crossed $100 million. By September, $200 million. The 2026 target is $656 million. If it gets there, Perplexity will have grown annual revenue roughly 19-fold in two years. The comparison set for that kind of trajectory is very small: Slack, Zoom during the pandemic, and not many others.

Follow the investor roster and you can track the growth. A $3.1 million seed from AI luminaries. A $25.6 million Series A from NEA, with Paul Buchheit (Gmail’s creator) and Susan Wojcicki (YouTube’s former CEO). A $73.6 million Series B from IVP, NVIDIA, and Jeff Bezos personally. Then the numbers got large: $500 million in mid-2025, Samsung and SoftBank participating. A $200 million round in September 2025 at a $20 billion valuation. And in December, an undisclosed investment from Cristiano Ronaldo, the soccer star, because in 2025 that is a sentence that makes sense.

Total raised: approximately $1.2 billion across nine rounds.

Most of that revenue comes from subscriptions. Perplexity Pro at $20 a month. Perplexity Max at $200 a month. Enterprise tiers from $40 to $325 per seat. Charging for search is unusual because Google trained the world to expect it for free. Perplexity’s bet is that a subset of users, knowledge workers, researchers, developers, professionals, will pay for search that is materially better.

So far, the bet appears to be correct. But the math is uncomfortable at scale. Even at $656 million in annual revenue, the company trades at roughly 30 times forward revenue on its last valuation. Subscriptions from 22 to 45 million monthly active users are real, but they are not enough to justify a $20 billion market cap without additional revenue streams.

So Perplexity is building them. In November 2024, it launched advertising, not as sponsored links but as “sponsored questions” that appear alongside organic follow-up suggestions. Early CPM rates reportedly exceeded $50, several times the display advertising average, because the purchase intent behind a Perplexity query is unusually legible. By October 2025, Perplexity had paused accepting new advertisers, a signal that demand outstripped the format’s capacity, or that the company was rethinking the approach, or both.

Then there is commerce. Buy with Pro lets US subscribers purchase products directly inside Perplexity through a PayPal-powered checkout. Over 5,000 merchants participate. Perplexity takes no commission. The play is not transactional revenue but stickiness: if you research and buy products in the same interface, you stop going back to Google.

And there is the $750 million, three-year deal with Microsoft Azure for cloud infrastructure, which tells you something about the scale of the cost base. Perplexity’s queries are expensive. Every one requires document retrieval, reranking, and language model inference, a pipeline that costs orders of magnitude more per query than Google’s precomputed index lookups. Whether subscription and advertising revenue can cover those costs at scale is the financial question that the growth curve has temporarily obscured.

The Content Problem

Perplexity’s product depends on content that Perplexity does not create.

This is the structural vulnerability that the lawsuits exploit and that no amount of product innovation can fully resolve. The answer engine reads, synthesizes, and presents other people’s journalism, research, and analysis. It cites its sources and links back to them. But the entire value proposition is that the user does not need to follow those links. The better Perplexity works, the less traffic flows to the original publishers. The more valuable the product becomes to users, the more destructive it becomes to the people who create the raw material.

Google faced the same tension with featured snippets and Knowledge Panels. But Google also sends publishers enormous amounts of referral traffic through organic search results. The relationship, while contentious, is reciprocal in a way that Perplexity’s is not. Perplexity is more parasite than partner, and the company knows it.

Enter the Publisher Program, the proposed cure. Launched in 2025 with a $42.5 million revenue pool and an 80/20 split (80% to publishers), it has signed over 300 partners: TIME, Fortune, Der Spiegel, the Los Angeles Times, The Independent. The Comet Plus subscription at $5 per month adds another layer, paying publishers based on direct traffic, AI citations, and agent-driven content usage.

For mid-tier publishers bleeding digital ad revenue, this is a new income stream that Google never offered. For the Times, whose lawsuit alleges scraping of paywalled content and circumvention of security measures, it is an insult. The Times does not want a share of Perplexity’s revenue. It wants Perplexity to stop using its content without a negotiated license. The distinction between an open revenue-sharing program and a negotiated licensing deal is the distinction between Perplexity setting the terms and publishers setting them.

Whatever the courts decide will matter far beyond the two parties. If the Times prevails, every AI company with a retrieval feature faces mandatory licensing costs that could make answer engines economically unviable. If Perplexity wins or settles favorably, AI synthesis with attribution becomes legally sanctioned, and the blue link dies a little faster. This case is less about Perplexity’s future than about the internet’s.

Cloudflare’s confirmation that Perplexity’s crawlers ignored robots.txt directives, the standard mechanism by which websites signal they do not want to be scraped, damaged the company’s credibility on the access question. It is hard to argue you are building a symbiotic relationship with publishers when your infrastructure actively circumvents their stated preferences.

Building the On-Ramp

Perplexity’s most acute problem is not technology, not revenue, not even lawsuits. It is distribution.

Google is the default search engine on Chrome, Safari, Firefox, and virtually every Android device on earth. Google pays Apple an estimated $20 billion annually for default placement on Safari alone. Even if Perplexity’s product is demonstrably better for a significant category of queries, most users will never discover this because their browser sends every query to Google before they think about alternatives.

This is why Perplexity launched a web browser. Comet, released in July 2025, is available on Windows, macOS, and Android. A “sidecar” AI assistant follows the user as they browse, summarizing pages, answering questions about content, and navigating on the user’s behalf. Max subscribers get a Background Assistant that runs multiple tasks in parallel: sending emails, buying tickets, booking flights.

The logic is uncomfortable but sound: if you cannot become the default inside someone else’s browser, build your own. Users who access Perplexity through Comet ask 6 to 18 times more questions than users who reach it through Chrome or Safari. That multiplier turns Perplexity from a destination you visit occasionally into an ambient layer across everything you do online.

Then came the Chrome bid, same logic at a vastly different scale. In August 2025, with the US Department of Justice pursuing antitrust remedies against Google, Perplexity submitted a $34.5 billion offer to acquire Chrome. The bid was almost certainly performative. Chrome has 3.4 billion users. Separating it from Google’s ecosystem would be a regulatory and engineering nightmare. But the bid accomplished what it was meant to: it positioned Perplexity as the company willing to operate at Google’s scale, and it planted the idea that Chrome under independent ownership might default to something other than Google Search.

Samsung is the mobile play. Perplexity is expected to be preloaded as a default assistant option on Samsung Galaxy S26 devices, following Samsung’s investment in the company. Samsung ships roughly 225 million phones a year. If even a fraction of those users try Perplexity and stay, it transforms the company’s addressable market overnight.

India is the other distribution bet. Srinivas has identified his home country as a major growth engine for 2026. India has 700 million internet users, the youngest demographic profile of any large market, and an English-speaking professional class that is Perplexity’s ideal user persona. In a country where Google’s dominance is near-total but where mobile-first users are less wedded to desktop habits, a Samsung preinstall could be the wedge.

In January 2026, Perplexity became a Wikimedia Enterprise customer, gaining structured access to Wikipedia’s data, the single most comprehensive reference source on the internet. The partnership is defensive as much as offensive: it replaces potentially contentious scraping with a legitimate licensing relationship, and it secures access to content that Perplexity’s competitors will also need.

Everyone Is Coming

Here is the competitive picture that Perplexity does not put in its pitch materials.

Google processes 8.5 billion searches daily. Its AI Overviews now appear on a growing share of queries, and the execution reveals the innovator’s dilemma in real time: every AI Overview that satisfies a user is one fewer ad click. Internal data reportedly shows Overviews reduce click-through rates by 30% on queries where they appear. Google has responded by embedding ads inside the AI-generated answers and being selective about when Overviews trigger, a compromise that degrades the user experience in order to protect the business model. Perplexity’s competitive advantage against Google is not technology. It is incentive structure. Google cannot afford to build the best AI search product. Perplexity cannot afford not to.

ChatGPT, holding 64-68% of the AI chatbot market, bolted on search capabilities that compete directly with Perplexity’s core product. OpenAI has the user base. What it does not have is the retrieval architecture. ChatGPT’s search is a feature inside a conversational AI; Perplexity’s search is the entire product. That distinction sounds semantic until you test both on a query that requires cross-referencing multiple sources and presenting citations accurately. But most users will never run that test. They will search wherever they already are. And 400 million of them are already in ChatGPT.

Gemini is the fastest-moving threat, and the one Perplexity’s investors should worry about most. Google’s AI chatbot nearly quadrupled its market share in a year, from 5.4% to 21.5%, driven by integration with Gmail, Docs, and Android. Gemini does not need to be better than Perplexity. It needs to be good enough and already on your phone. No standalone startup can outrun distribution that ships inside 3 billion mobile devices.

Against these forces, Perplexity’s 2% of the AI chatbot market looks marginal. But that 2% is measured in the wrong units. The metric lumps together every AI interaction, code generation and creative writing and casual conversation, into a single number. Perplexity competes in none of those categories. It competes in research-oriented information retrieval, where 780 million monthly queries, growing 20% month over month, represent something real: a user base that has found the product indispensable for work that matters.

Perplexity will not overtake Google. That is not the question. The question is whether the market for research-grade, cited, high-quality search is large enough to sustain a major independent company. If it is, the 2% is a starting position, not a ceiling.

Deep Research and the $400-an-Hour User

Perplexity’s basic search product is its answer to Google. Deep Research is its bid for a market Google has never entered.

Launched in February 2025, Deep Research runs autonomous research sessions lasting two to four minutes. It formulates a plan, executes dozens of parallel web searches, cross-references findings, and produces a comprehensive report with executive summaries, timelines, and sourced recommendations. On SimpleQA, which measures factual accuracy, it scores 93.9%. On Humanity’s Last Exam, designed to stump frontier AI, it scored 21.1%, beating OpenAI’s o3-mini, o1, and DeepSeek R1.

What Deep Research actually automates is the workflow that knowledge workers perform daily but never think to optimize: gathering sources, reading them, extracting key points, organizing findings into a structure. A management consultant billing at $400 an hour who saves two hours a week captures $40,000 in annual value from a $200 annual subscription. A lawyer, a financial analyst, a policy researcher, anyone whose job is synthesizing information from disparate sources, faces a similar value equation. Price sensitivity at those ratios is essentially zero.

Google has never built a product for this. Its entire interface, a text box and a list of results, assumes a quick lookup, not sustained investigation. Rebuilding Google for multi-step research would require a different product architecture, one that conflicts with the ad-supported model because deep research users spend their time reading AI-generated reports rather than clicking on anything.

Perplexity Spaces extends this into collaboration. Teams create shared research workspaces, upload documents, set custom AI instructions, and invite collaborators. The enterprise version supports up to 5,000 files per Space. It positions Perplexity as a research platform, competing less with Google and more with Glean, Notion AI, and Microsoft Copilot for knowledge work.

There is also a quieter distribution play. Education Pro gives students free twelve-month access to Pro features. The economics are clear: acquire users at the moment they are most likely to form search habits, subsidize them through their highest-research-intensity years, and convert them to paying subscribers when they enter the workforce. Google used a version of the same strategy with Google Apps for Education. Perplexity is using it against Google.

The Vulnerable $20 Billion

Strip away the growth narrative, and what you find underneath is a company burning money faster than it makes it, dependent on technology it does not fully control, and exposed to a quality problem that gets worse as it scales.

Start with the cost. Every Perplexity query runs through a retrieval pipeline and a language model inference stack. A single query costs orders of magnitude more than a Google lookup, which simply matches keywords against a precomputed index. At 780 million queries a month and growing, the compute bill is staggering. Perplexity signed a $750 million, three-year deal with Microsoft Azure. If the company does not reach profitability within that window, the next infrastructure negotiation happens from a position of weakness, and Microsoft, which owns a competing search product in Bing, will not be a sympathetic counterparty.

Now consider who builds the brains. Despite Sonar, Perplexity’s best answers still depend on Claude, GPT, and Gemini, models controlled by Anthropic, OpenAI, and Google, three companies that are building their own search products. Any of them could raise API prices, throttle access, or degrade service quality to favor their own offerings. Perplexity’s multi-model strategy hedges the risk, but hedging is not the same as eliminating it. Building a frontier-class language model from scratch would cost billions. Perplexity has raised $1.2 billion total.

Then there is the problem that no amount of money solves: hallucination at scale. When the user base is technical early adopters who verify citations, a fabricated statistic gets caught and reported. When the user base is a mainstream audience that trusts the polished output the way it once trusted the top Google result, a fabricated statistic becomes someone’s legal brief, someone’s investment thesis, someone’s medical decision. The citation mechanism buys credibility. It does not guarantee accuracy. And as Perplexity grows, the gap between perceived and actual reliability will widen in ways that could be reputationally devastating.

Regulation compounds everything. The EU AI Act, US state-level frameworks, and emerging Asian regulations all target systems that generate content or make recommendations. Perplexity, which synthesizes information and presents it as authoritative, is exactly the kind of product these laws were written for. Compliance costs will grow. The question is whether they grow faster than revenue.

Disruption or Delusion

The question that the $20 billion valuation ultimately rests on is not about product quality or revenue growth. It is about category.

Google in 1998 held 0% of search. AltaVista, Lycos, and Yahoo had the distribution, the brand recognition, and the scale. They lost within five years because Google was better at the one thing that mattered: giving you the answer to your question. The incumbents’ advantages turned out to be irrelevant when the product gap was wide enough.

Whether Perplexity is in the same position depends on a classification that the market has not yet made. If AI search is a sustaining innovation, an incremental improvement to existing search, then Google wins by doing what it always does: absorbing the new technology into its existing product and leveraging its distribution. Perplexity becomes DuckDuckGo for the AI era. Respected. Niche. Marginal.

If AI search is a disruptive innovation, one that changes the fundamental nature of what search is, then Google’s $175 billion advertising business becomes the anchor that prevents it from swimming. Every dollar of ad revenue creates a reason not to build the best possible AI search product. Every existing user expectation creates a reason not to ship a radically different interface. And Perplexity, unburdened by 25 years of accumulated business model constraints, builds the thing Google cannot afford to.

Srinivas has his own framing: “It is not better or faster search. It is discovery of knowledge.”

Discovery of knowledge is not an improvement to an existing behavior. It is a new behavior. If Perplexity succeeds, it will not be because it stole queries from Google. It will be because it enabled questions that nobody would have typed into a search box, research too complex for ten blue links, analysis too tedious to assemble from tabs, synthesis that nobody had the patience to do by hand. That market is not a fraction of Google’s market. It is a market that Google’s product, by design, cannot serve.

Three years. $1.2 billion in funding. $200 million in revenue. A CEO who worked at every lab that matters, who raises money by opening a laptop instead of a slide deck, who shipped a product before ChatGPT existed and bid for TikTok and Chrome before anyone asked him to. A product that reads other people’s work, cites it, summarizes it, and makes you wonder why you ever accepted a page of blue links as an answer to anything. And a lawsuit from the most powerful newspaper in the world that could decide whether any of this is legal.

Startups funded on disruption theses do not get to wait for the thesis to be proven. They have to build as if it is already true, spend as if the market is already theirs, and outrun the uncertainty. Perplexity has $1.2 billion, a product that works, and a closing window. Use it once, and the blue link feels like a relic. Look at the balance sheet, and the relic looks like it might outlast the revolution.

The distance between those two feelings is the distance between a $20 billion company and a footnote.

Published on February 12, 2026.

About the Author

Gene Dai is the co-founder of OpenJobs AI, a next-generation recruitment technology platform. He writes about the intersection of artificial intelligence, developer tools, and the future of work.