Kalshi Founder: What Will Your Real Life Look Like in the Next 12 Months

Podcast: "$22B Kalshi Co-Founder: What the Market Already Knows About Your Job"丨Silicon Valley Girl

Compiled by: LatePost

"By the end of this year, the biggest change will still be in work."

"Currently, no position has been completely replaced by us; all positions have merely been enhanced by AI."

"People are always worried that they are not working hard enough. At Kalshi, we are more concerned that AI is not being used enough."

Luana Lopes Lara co-founded the prediction market platform Kalshi in 2018. According to Forbes, she is the youngest self-made female billionaire in the world. In May of this year, Kalshi raised $1 billion, reaching a valuation of $22 billion. What she observes daily is quite special: people are willing to bet real money on which outcome will occur. When the discussion shifts to AI and work, these prices continuously reflect how risks change, rarely providing comfort.

What People Are Asking Has Shifted from Ability to Consequences

A year or two ago, most AI market proposals received by Kalshi were focused on ability: whether a model could complete a task, which model would win, and when large mathematical problems would be solved. Now, users more frequently ask how AI will affect layoffs, unemployment, and elections. Luana believes this shift indicates that ordinary people have moved beyond the "What is AI?" stage and are beginning to calculate the costs it imposes on them.

"A year or two ago, most proposed markets were about AI capabilities; now, we receive more requests about the impacts brought by AI."

The excitement has not disappeared, but the focus of the questions has changed. People still trade on model capabilities, drug approvals, and quantum computing, but "Will AI be the primary reason for layoffs this month?" has become a market that people monitor every month. Tech news is no longer just about product launches; it has begun to enter payrolls and job descriptions.

There is also a market for AI "disaster scenarios" that Luana and her co-founders have long focused on. It sets five conditions, and if three of them occur, it settles as "yes," with the probability during the interview being about 26% to 30%, and the transaction amount reaching millions of dollars. This number is much higher than many people's intuitive sense of risk, prompting the team to review it repeatedly.

73%, What Can It Really Tell You

At the time of the interview, the market believed there was a 73% probability that AI would be the primary cause of layoffs in May, with a transaction volume of about $30,000. In March and April, AI indeed was the primary cause of layoffs. Luana mentioned that Kalshi Research conducted calibration studies, showing that even with a transaction volume of only about $5,000, prices often converge to quite stable numbers. The more funds involved and the closer to settlement, the more reliable the readings usually are.

"Even with a transaction volume of only $5,000, you can already see the numbers converging to a well-calibrated level, so this number is indeed trustworthy."

This 73% cannot be directly translated to "there is a 73% chance your job will disappear." It answers a clearly defined question: in the specified month, will AI be recorded as the primary cause of layoffs? The value of prediction markets comes from clearly defined questions, with prices set by those willing to bear the losses. If the question is vague, the numbers will lose their meaning.

Some questions take five years to settle, such as when a certain drug will be approved or when breakthroughs in quantum computing will occur. After Kalshi began paying interest on positions and dollars in accounts, users no longer have to abandon these trades due to long-term capital being tied up. This allows the market to accumulate more information for distant events, and prices are not just chasing the news of the day.

Probability Is Not Prophecy; 1% Can Happen

Before the election of a U.S. citizen pope, the probability on Kalshi had long been about 1%. After the result emerged, some news reported that the prediction market "was wrong." Luana's response was straightforward: 1% does not equal zero. Secret meetings are almost closed information systems, making it inherently difficult for outsiders to predict. When low-probability events actually occur, it does not retroactively invalidate the previous probabilities.

"70% is not 100%. Even if the number is 99%, there could still be one occurrence in every 100 times that does not happen; 1% is not zero."

The same transaction volume represents different levels of credibility in different markets. A week before the election, the political market might only need a few thousand dollars to be accurate; six months in advance, it often requires tens of thousands of dollars. Before looking at the numbers, first consider the settlement time, transaction volume, and question definition. These three conditions will be closer to the real risk than a single percentage.

Luana explains that prediction markets have always required participants to pay for their judgments. The more solid the research, the greater the chance of making money by betting correctly; casual statements come with the risk of loss. It does not create an omniscient group but can turn "I think" into a cost-bearing, continuously adjustable number.

Most People Come Here to Just Look at the News, Not to Bet

70% of Kalshi's users never trade. They open the website just to quickly see the latest probabilities for economic, election, sports, and cultural events. This is like a news report written in prices: the homepage does not reflect the stance of any media outlet, only the judgments expressed by many people through their funds. The company has also compressed thousands of election markets into the K Power index, allowing users to see with one number whether U.S. politics currently leans more towards the Republican or Democratic party.

"70% of our users actually do not trade anything; they just come in to get information and see predictions for different events."

Another category of users treats the market as a hedging tool. A bar in the Upper East Side of New York promised to cover the bill for customers if the Knicks win, with a worst-case potential loss of $20,000, so it bought the opposite outcome to control risk. When hurricane season arrives in Florida, some people request markets for specific areas where they live, using the profits to offset insurance deductibles. Predictions can help people make decisions, and hedging prevents heavy losses from incorrect judgments.

The work the platform needs to do is also more complicated than just "creating a yes-or-no question." When users say they want to know if AI affects elections, the team must first break down "affect" into conditions that can be settled. National political leanings are harder to express with a single seat, so K Power combines predictions from the House of Representatives, Senate, and local elections. A good number often hides a long definition process behind it.

The Biggest Change This Year Is Still in Work

Luana does not claim that any position has completely disappeared at Kalshi. What she sees is that all positions have been amplified by AI: engineers simultaneously schedule about 20 cloud agents, new employees use agents to fill in company context, and managers can quickly know which projects are stalled. She still judges that by the end of this year, the most noticeable change for ordinary people will occur in work.

"I have not seen any position completely switched to 'we no longer need this person; we have AI'; all positions have been enhanced by AI."

She is currently more optimistic about jobs that require physical participation, such as fitness trainers, and skilled positions; she has not concluded where robots will go next. She feels that the sense of security in engineering positions, once considered safe, is actually declining. Job titles may still exist, but the methods of completing tasks, the number of team members required, and hiring standards have begun to change.

She gave a very everyday example using travel planning. In the past, arranging a weekend trip required searching for accommodations and itineraries on her own; now she lets ChatGPT plan for two days, but she still has to manually open the website and book each hotel. This breakpoint illustrates the current situation well: AI has taken over thinking and organizing, but the final few steps of executing across systems are often still left to humans.

170 People, How to Run an AI-First Company

Kalshi currently has about 170 employees. Luana views market operations as a factory, continuously monitoring error counts, market launch delays, result determination delays, and coverage rates. The AI system collects these metrics from the ground up and summarizes emails, documents, and team updates for managers. This allows her to manage more projects simultaneously and to see earlier when someone has overcommitted for several weeks and underdelivered.

"AI allows me to get context faster and know what is happening more quickly, so I can manage more threads and more people more effectively."

Sundays used to be her busiest day, reviewing what was completed last week, what to do next week, and checking the entire set of metrics. Now, much of the organizing work is done by agents, and she can finally go out for brunch on Sundays. The time saved does not eliminate management; it shifts managers from transporting information to judging information. Authority remains a challenge, as legal and monitoring materials can only be accessible to a few people.

The team hopes that each new employee will receive their own agent upon onboarding, connected to emails, documents, and work records. The most challenging part right now is permissions: the same system must allow newcomers to quickly catch up while ensuring that legal cases and market monitoring materials are only visible to authorized individuals. The more complete the context, the less ambiguous the permission design must be.

The More Agents, the More Need for Someone to Be Full-Time Responsible

An unconventional detail emerged in the interview: Kalshi uses more agents but is still hiring engineers. Luana and her co-founders are already occupied with daily operations, and if they each take 5% of their time to work on the internal AI system, the product is unlikely to be stable. The company has thus formed a dedicated team, placing engineers into market operations, design, and legal, handling processes one by one.

"If I only take 5% of my time to do it, it won't be done well. For AI to become an important part of the company, it needs very capable people to be responsible full-time."

The real processes are far more complex than demonstrations. Product feedback is scattered across Slack, Discord, tweets, and internal discussions; to automatically sort it, one must track which issues have already gone live. The company has tested multiple QA products, but none have met the requirements. The design team wants to shorten the cycle of "engineers launching experiments first, then designers coming back to fix them." The story of a solo founder is very appealing, but a stable AI system still requires clear accountability.

Kalshi has just launched perpetual contracts, stepping out of the prediction market for the first time. Luana said that without the current AI capabilities, this new product would either drag down the main business or require hiring many more people. She looks forward to the company creating more products and growing larger with AI, with the focus not on hiring less but on expanding boundaries.

Hiring Standards Are Shifting from Capability to Willingness to Relearn

Kalshi allows candidates to use AI in engineering interviews. The question "Can this person complete a technical task?" that was commonly asked two years ago is no longer as distinguishing. The team now spends more time on system reviews and retrospectives of past projects, with design roles asking candidates how much AI they have used, and legal interviews beginning to inquire, "How would you use AI to handle this type of case?"

"The way you worked in the past will not be the way we continue to work at Kalshi. We need people who are willing to re-understand: those things they were once very good at may no longer need to be done by hand."

Luana values low ego, willingness to learn, and reliable delivery. The hard work she refers to is not about counting hours but about doing the entrusted tasks well. Tools will continue to change, and the willingness to let go of old advantages has already entered the hiring judgment. This also explains why engineers are moving into design and legal: the company is not changing an isolated task but the way each team handles tasks.

This culture is also very direct: if colleagues dislike a certain job, they will clarify it on the spot. The company spent three to four years securing regulatory approval and then sued its regulatory agency after two years of communication, ultimately gaining space for the election market. Luana summarizes this experience with a simple principle: complete everything you can do to avoid discovering after failure that you actually left a step undone.

In Conclusion

73% is worth serious consideration, but it should only be taken as 73%. A more useful action for individuals is to rewrite concerns into verifiable questions: which steps in your work have already been taken over by agents, which judgments still rely on experience, and what has recently changed in the team's hiring standards. Probability will not decide your next step, but it can remind you not to plan today with the security of two years ago.