AiDept Solutions is a product studio. It makes its own websites, apps, and AI tools, ships them to production, and keeps running them for the long haul. Seven are live today, alongside the trading systems it runs on its own capital.
live products designed, built, and run in-house every day
bots and trading systems running live on house capital
projects in development, built but not yet launched
projects in all, across products, systems, and development
Everything here was designed, built, launched, and is run day to day by AiDept Solutions. Each product offers something functional, useful, and interesting, and each appeals to a wide range of users.
A plain-English guide to California workers' compensation for people hurt on the job. An AI intake asks what happened and connects qualified cases to vetted attorneys. The attorney partners pay for those referrals.
The San Diego edition: local content, answers for injured workers in the county, and the same attorney pipeline.
In-house systems, run on the studio's own capital, built with the same discipline as the products. An edge has to be proven on data before money moves, and every system runs under loss caps and a kill switch that a human controls.
Kalshi runs daily markets on the high temperature in 16 US cities. This bot forecasts each one by blending several weather models, weighting each by how wrong it has been lately, and bets only when its forecast disagrees with the market by more than the fees. It trades the studio's own account, and the profit is the product.
An automated trader on Interactive Brokers, run on the studio's own account. A risk governor sits above every strategy: daily loss breakers, a cap on total open risk, no trading into earnings, and a paper-to-live gate it had to pass on the numbers and with a human sign-off before the first real order. A learning module studies the account's own trade history and flags the habits that cost money. Revenue is simply what it earns in the account.
Watches every trade on Polymarket as it happens and flags the ones worth a second look: a brand-new wallet placing a large bet, a trade too big for the market's liquidity, a repeat whale showing up again. Each flag is scored against what the market did next, so the record is honest. It runs in-house today, and a web app with free and paid tiers is being built for other Polymarket traders who want the same alerts.
Kalshi also lists NFL player props. This bot builds parlays from high-probability legs, an edge that was proven on settled markets with public data before it was allowed to pick a single leg. Projections, injury reports, Vegas lines, and usage data feed each week's picks. It trades the studio's own account under per-parlay and weekly loss caps.
Anyone selling an investment property through a 1031 exchange has 45 days to name the replacement and 180 to close, and the clock does not care how thin the market is. This bot works both sides of that clock. Months before a sale, it mines county assessor rolls, tax-delinquent lists, recorder filings, and estate records to surface owners showing stacked signs of motivation, such as an out-of-state owner with a 20-year hold, a loan maturing within the year, and a suspended LLC. Once the clock starts, it scrapes on-market listings, parses the broker emails and offering memos that land in the inbox, and screens every deal against the exchange math, with one or two fast-close backups kept ready. Every candidate is ranked by exchange fit times motivation, and a deadline engine escalates alerts at days 30, 40, and 44.
Works in progress at various stages of development, each with real code behind it. Some are waiting on a launch gate, some are still being built. Nothing here is live for the public yet.
A multiplayer game where you play a commercial real estate investor: walk the map, haggle with vendors, buy and improve buildings, and climb a live leaderboard ranked by portfolio value. For people who like business and tycoon games and want one built on real property mechanics. The server does all the money math and dice rolls, so nobody can cheat from their browser.
The longer play is behavior and decision mapping. Every bid, haggle, hold, and walk-away is a decision under uncertainty, and the game records them all. Over many sessions that becomes a profile of how a player actually decides, across risk, loss aversion, patience, and negotiation style, which is rarely how they say they decide. The game is the front door; that decision data is the long-term asset.
Lease help for small and mid-size businesses about to sign or renew a commercial lease, from someone who is not paid by the landlord and not paid on commission. Tenants pick a flat-fee package, book it, and pay online. A glossary and a library of articles answer the questions tenants actually search for, which is how they find the site.
Group celebration boards for birthdays, work anniversaries, and farewells: the online version of the card passed around the office. One person starts a board and shares a link, everyone adds a message, and the finished board lands in the guest of honor's inbox on the day. The twist is the AI layer. The guest of honor can be placed into face-swap scenes, and contributors can generate a custom image for their message instead of hunting for a GIF. Paid boards, with a free watermarked tier.
MonitorA monitor that watches for signs of military escalation between China and Taiwan before it reaches the headlines. It tracks public market data, physical-world signals like flight restrictions, and news flow, compares each to its own normal range, and raises a graded alert only when several unrelated signals move together. Built to position the studio's own portfolio ahead of an event, not for sale. It runs on public data only and, by default, never places an order without a human approving it.
Trading toolA desk tool for trading S&P 500 futures. It reads the live tape and order book from CME, works out where large buyers and sellers are actually defending price, and puts those levels on a chart with alerts when price reaches them. It is for in-house intraday trading and never sends an order; a human does that. Whether a level holds is measured by replaying it against history, never assumed.
Three disciplines, one studio. Every product goes through all three, and the studio stays with it after launch.
Every product starts as a problem worth solving and a page someone would actually use. The flows, the interface, and the content structure are designed before any code is written, with search and AI answer engines in mind from the first sketch.
Websites, web apps, and AI-powered tools, shipped to production on Next.js, Vercel, and Supabase. Agents and automation run on Claude, GPT, and open models. Tracking and analytics are wired before launch, never after.
Launch is the start. Everything that ships gets monitored, published, tuned, and iterated on: uptime, content, search and answer-engine upkeep, performance, and the numbers that say whether a product is earning its keep.
Four phases, from a question to a product that runs on its own. Each has a gate, so nothing ships before it has earned it.
Start with a real question: who needs this, what would they search for, and how does it pay for itself. If the answer is weak, the idea stops here.
Design and ship on a modern stack, structured for search from day one, with clean tracking in place before anything goes live.
Deploy to production, connect analytics and channels, and confirm real people can find it and use it before calling it live.
Monitoring, content, search upkeep, and tuning. The product keeps improving instead of going stale.
Partnerships, questions about a product, press, or a project you think fits what AiDept Solutions builds. Email is the fastest way in.
AJ@aideptsolutions.com
