Ten bets I actually hold, tracked out loud so I cannot rewrite history later. This page is the whole method: where the data comes from, how the grading works, the maths behind the records, and the lines I drew before the money moved. If something here is not true, the site is broken. It is not me being modest.
in10x is one person betting on ten ideas about where AI value ends up. Seven of them are bets that something rises. Two are bets that something falls, which the market calls a short. One is a bet on a whole style of company rather than a story: cheap businesses that make real profit. A thesis is just a bet you have written down properly, so each one here carries three things. The idea itself. The companies that carry it. The exact thing that would prove it wrong. Then I handed the day to day judging to a machine, so I could not quietly grade my own homework afterwards.
It is a running experiment in showing your work. The point is not to be right. The point is to be checkable: every claim dated, every call graded, nothing edited after the fact.
Every weekday at 07:00 CET a pipeline on a small server runs the same five steps, in order, unattended:
Same steps, same order, every trading day. If a news search comes back empty or a price is missing, the run does not fall over. It grades what it has and says so.
There is a number behind the scenes, from 0 to 100, for how strongly the evidence backs each bet. It exists, it moves daily, and it decides when a bet gets flagged for a kill review and when my phone buzzes. You will never see it on this site, and that is deliberate.
A single number is a black box wearing a lab coat. It invites you to trust the output instead of reading the evidence. So the site only speaks in things you can check for yourself: a chip saying whether today supports the bet, goes against it, or is too mixed to call, the daily score for each separate signal, the companies side by side, the pick and the reason given for it, and the tape. If you want a view, build your own. That is the entire idea. Hiding the score is the feature.
Every signal a bet depends on gets a daily score from −2 to +2. What happened in the world counts three times as much as what the share price did, because a price can move for any reason at all. The chip is the average of today's scores put into words: strongly supports (+1.2 or better) · supports (+0.4 or better) · mixed (in between) · against (−0.4 or worse) · strongly against (−1.2 or worse).
Green means the world moved the way the bet needs today. Red means it moved against it. Gray means the day's signals pulled both ways. On a short bet everything flips, so falling prices are green, because down is exactly what that bet expects. And the chip reads the evidence, never the money. A bet can read "strongly supports" while I am down on it. That is not a contradiction. It means the idea is being proven faster than the price has caught up, and the receipts page quietly keeps the money side, unedited.
Each bet names one pick: the single stock that is the best way to own the idea today, or on a short bet, the name with the most to lose. The AI makes that call fresh every weekday from the data on the page. Here is exactly how it gets counted, so you can trust it or tear it apart:
The receipts page keeps every pick and every earnings call, dated, with the ones that went badly sitting next to the ones that went well. Nothing on it is edited after the fact.
Four times a year a company has to open its books and show what it actually earned. That is where a bet gets tested instead of argued about. So before each of those reports the machine writes down what this one has to show for the bet to hold. Specific lines, not vague hopes. The note goes public, with its date on it, while nobody yet knows the answer.
Once the company reports, the same machine marks its own expectation and says whether the numbers support the bet, go against it, or land mixed, with the note it wrote beforehand still sitting right beside it. You can open both and check that the goalposts did not move in between. That is the whole reason for writing it down first.
Every bet ships with two lines I wrote before putting money behind it, shown on its page in the words I first used:
These are not vibes. They can actually happen or not happen: two quarters in a row of new orders shrinking, spending cuts turning into cancelled deals, a group of stocks trailing its own market for half a year. When the evidence sits against a bet for long enough, the machine raises a kill review flag on that page by itself, whether or not I like it. I can close a bet, but I cannot pretend the line said something else. A dead bet stays on the site with the cause of death attached.
This is the part worth stealing, even if you never read another word here. A bet you cannot lose is not a bet, it is a mood. Writing the losing condition down first is the cheapest way to find out which one you have.
One person, Theis (@theisbuilds on X). It is a personal project, not a firm, a fund or a newsletter business. These are bets I actually hold, which is exactly why I built something that stops me from quietly rewriting them later.
A bet you have written down properly. It has three parts: the idea itself, the companies that carry it, and the specific thing that would prove it wrong. That last part is the one people skip. A bet you cannot lose is not a bet, it is a mood, so every thesis here carries the line that kills it, written before the money moved.
No. I chose the ten bets and the companies. The AI's job is to judge the evidence each weekday, not to come up with ideas. It scores every signal from -2 to +2, names the one stock that best carries each bet today, and does it the same way whether the news is good or bad. It is the referee, not the player.
Please do not trust it. Check it. The receipts page is written by the pipeline itself, so every pick is priced from the last close before it was published, meaning nobody got to pick a better moment afterwards, and the calls that went badly sit next to the ones that went well. The dates and the maths are all there to be picked apart.
There is a 0 to 100 number behind the scenes and it is deliberately hidden. A single number is a black box that invites you to trust the output instead of reading the evidence. The site only shows things you can check: a chip saying whether the day supports the bet, goes against it or is too mixed to call, the daily score for each signal, the company numbers, and the reason given for the pick. Form your own view. That is the entire idea.
Each weekday the AI scores every signal a bet depends on, from -2 to +2, based on that day's headlines and price moves. The chip is the average. Strongly supports means clearly positive (1.2 or better), supports means mildly positive, mixed means the signals cancel out, and against or strongly against are the same thing in reverse. It grades the day's evidence for the idea, not the money, so a bet can be losing money while the evidence behind it reads green, and the other way around. On a short bet it flips, so evidence the names should fall counts as support.
Because I hold some of them as shorts, meaning I make money if the shares fall. Two of the ten are like that: the Musk story premium and software priced per employee. Pretending everything is a buy would be dishonest. On a short, the evidence flips, so the shares doing badly is what supports the bet, and the site shows it in red.
There is no single AI trade here. There are ten specific bets, each with a condition for being wrong that was written down first. Seven are bets that something rises, two that something falls, one on a whole style of company. If the AI trade really is cracking, it shows up as evidence turning against particular bets and their kill lines tripping, visibly and on the record, rather than as a feeling.
Every weekday, around 07:00 CET. The pipeline pulls fresh prices, company numbers and news, grades all ten bets again, and rebuilds the site on its own, with nobody in the loop. Weekends and holidays it rests, like the market.
The map on the front page showing how the ten bets connect. Some reinforce each other, meaning they ride the same force from different angles. Some are in tension, meaning the same force taken from opposite sides. $TSLA, for example, sits on one bet that it rises and one that it falls. The brain grows a new node and new wires every time a bet is added.
A long bet makes money when the shares rise. A short bet makes money when they fall. On a short the evidence flips, so the shares doing badly, or the business getting worse, is what supports the bet, and the pick is the name with the most to lose rather than a suggestion to buy anything.
Yes. The site says bets I actually hold, and that is the whole reason for building something that keeps me honest about them. That said, nothing here is investment advice or a recommendation. It is a public record of one investor's reasoning.
No, and that is the entire point. Every daily read and every pick is saved with its date and never edited afterwards. The tape shows exactly what the machine said on the day it said it. I can add a new call or close a bet, but nothing already published gets quietly rewritten. If a call aged badly it stays there, aging badly, in public.
When the evidence sits against a bet past the line I drew in advance, its page raises a kill review flag on its own. If I close the bet it does not vanish. It stays on the site with the line that killed it and the reason it tripped, so the failure is as visible as any win. Retiring an idea is fine. Hiding that it failed is not.
Prices and company numbers from Yahoo Finance, headlines from news searches aimed at each specific signal, and the daily grades from the AI (Grok, xAI). Every page shows its own numbers and links out to the headlines that fed the read, so you can trace any call back to where it came from.