Chessigma: Revenue, team and how it was built
Chessigma (AI chess coach on top of free game analysis, Belgium): US$16.7K verified revenue in one month, September 2026. Evidence grade A: disclosed by a company or a filing, or verified by a payment processor. Team: 2 co-founders, Mehdi Ben Haddou and Eliot (who also works as a software engineer at Amazon); no other staff named. Launched 27 March 2025; still running, revenue growing.Added 25 October 2026
- Outcome
- US$16.7K verified revenue in one month, September 2026Revenue growing
- Evidence
- A disclosed by a company or a filing, or verified by a payment processor
- Launch
- 27 March 2025
- Status
- Still running
- Team
- 2 co-founders, Mehdi Ben Haddou and Eliot (who also works as a software engineer at Amazon); no other staff named
- Outside funding
- None reported (no source mentions outside investment)
- Group
- Learning and study
- Main channel
- Creators and short videos
Chessigma is a chess training website with two layers. The free layer is game analysis by Stockfish, the strongest open-source chess engine. A chess engine is a program that searches the possible moves of a position and gives each position a score, for example +1.0 when White is about one pawn ahead. Since 2020 Stockfish scores positions with NNUE (efficiently updatable neural network), a small neural network trained on billions of positions, so even the free layer uses machine learning, but it is a standard engine that every chess site uses. The paid layer, Supercoach, is the new part: a personal report, daily drills made from the player's own mistakes, and a chat coach whose answers are written by a large language model, a text model of the kind behind ChatGPT, reached through the OpenRouter service.
Mehdi Ben Haddou launched the site on 27 March 2025 and runs it with a co-founder, Eliot, who also works as a software engineer at Amazon. The legal pages name the operator as Arche Labs Ltd, trading as Chessigma. Revenue is verified: TrustMRR, a site that reads revenue directly from a startup's payment provider, shows US$16,651 of Stripe payments in September 2026, US$17,454 in the 30 days to 25 October 2026 and US$73,209 in total. It also shows US$7,622 of monthly recurring revenue (MRR) from 593 active subscriptions. MRR counts only subscription revenue and spreads a yearly payment over twelve months, so it is lower than the cash that arrives in a month, which also includes yearly and one-time lifetime payments.
This page grades every number by its evidence. Monthly revenue grew from US$3,738 in May 2026 to US$16,651 in September 2026, but the growth slowed: September was 5.6% above August, and the last 30 days were 3.7% below the 30 days before. The page also checks the fair-play question, because the site offers a free best-move tool, and shows four failed startups that taught or hosted board games, or sold personalized learning, before language models could explain a player's own mistakes.
How much revenue does Chessigma make?
Each row shows the evidence grade and the source numbers from the list at the end of this page. The grades are explained in How Aiolum verifies.
| Fact | Value | Grade | How we know | Sources |
|---|---|---|---|---|
| Verified revenue, September 2026 | US$16,651 | A | Monthly revenue timeline of the TrustMRR profile, read from Stripe through an API key. Stripe is not the merchant of record. Payments through the Apple and Google app stores, if any, are not in this figure. | [1] |
| Verified revenue, last 30 days to 25 October 2026 | US$17,454 | A | TrustMRR shows a change of -3.7% against the 30 days before. Two days, 7 and 8 October 2026, brought US$4,661 of it. | [1] |
| Verified revenue, all time to 25 October 2026 | US$73,209 | A | US$73,095 of it in the last 6 months, that is from May 2026. | [1] |
| Verified revenue, January to April 2026 | US$18 to US$42 a month | A | In those months the only paid plan was a US$3-a-month supporter plan; the AI coaching plan was marked coming soon. | [1, 4] |
| Verified revenue by month, May to August 2026 | US$3,738; US$7,828; US$13,654; US$15,775 | A | May 2026 is the month the Supercoach plan went on sale. | [1, 5] |
| Growth from August to September 2026 | 5.6% | calc | Calculated: 16,651 divided by 15,775 equals 1.056. From July to September: 21.9%. | [1] |
| Monthly recurring revenue, 25 October 2026 | US$7,622 | A | TrustMRR metric snapshot, with +14.1% in the last 30 days. MRR counts subscriptions only, with yearly plans spread over 12 months. | [1] |
| Active subscriptions, 25 October 2026 | 593 | A | TrustMRR metric snapshot. | [1] |
| Annual run rate of the recurring revenue | US$91,464 | calc | Calculated: 7,622 times 12. | [1] |
| Visitors, last 30 days to 25 October 2026 | 913,959 | B | Read by TrustMRR from Google Analytics. Revenue per visitor: US$0.02 (TrustMRR); 17,454 divided by 913,959 equals US$0.019 (calculated). | [1] |
| Paying customers so far | 1,713 | C | Company pages: "1,713 happy customers so far" and "1,713 trained with Coach Froggy". Not a payment figure. | [2, 3] |
| Visitors in the first week after launch | more than 100,000 unique visitors | C | Founder blog post of 6 April 2025. | [6] |
| Visitors a day, May 2025 | about 7,000 unique visitors | C | Founder blog post of 15 May 2025 on how the analysis works. | [7] |
| Team | 2 co-founders | C | The site links Mehdi and Eliot as founders, the blog lists Eliot as co-founder and Amazon software engineer, and the pricing page says Mehdi or Eliot handle refunds. No other staff are named. | [2, 11, 14] |
| Outside funding | none reported | C | No source mentions an investment. Mehdi Ben Haddou's profile mentions an earlier job as founding engineer at a Y Combinator company; that is his former employer, not an investor in Chessigma. | [1, 19] |
| Time from launch to the headline month | 18 months | calc | Calculated: March 2025 to September 2026. Time from the paid launch in May 2026: 4 months. | [1, 6] |
How to read this table: a grade A value was read by TrustMRR from Stripe, the payment provider, through an API key. A grade B value was read by TrustMRR from Google Analytics. A grade C value comes from the company's own pages or a founder-written listing. A "calc" value was computed on this page from the rows above it.
| Plan | Per month, billed yearly (US$) | Per month, billed monthly (US$) | Limit |
|---|---|---|---|
| Free | 0 | 0 | Unlimited game review and puzzles; random daily plan, 5 opening lines, 2 bots, 1 Woodpecker puzzle set |
| Supercoach monthly | 14 | Personal daily plan, Chess DNA report, all trainers and bots unlimited, AI coach Froggy 200 messages a month, no ads | |
| Supercoach yearly | 8.08 | Same as monthly, AI coach 400 messages a month (the page adds "x2"); US$97 billed yearly | |
| Lifetime | 490 one-time | 490 one-time | Same as yearly, one payment, no renewals |
The monthly price of US$14 is shown next to a struck-through US$19. The yearly plan is US$97 billed once a year, which is US$8.08 a month and 42% less than 12 monthly payments (calculated). The page sells the plans as a founding class ("Founding Class #2"); in May 2026 the yearly plan cost US$140 and the lifetime plan US$340. All paid plans have a 7-day money-back guarantee. The phone app is free and does not unlock the paid plan; the paid features work on the website and the desktop app.
When did Chessigma launch, and how did it grow?
From a free game reviewer launched on Reddit to a paid AI coach with US$16,651 of revenue in one month. Each entry carries its grade.
- 27 March 2025CThe founders launch Chessigma as a free tool that gives unlimited game analysis reports with no sign-up. They post it on Reddit. [6]
- March and April 2025CA day after the Reddit post, a chess YouTuber posts a short video about the site. The founders write that it went viral and was reposted on TikTok, Instagram, Facebook and WhatsApp groups, and that the site had more than 100,000 unique visitors in the first week. [6]
- 15 May 2025CA blog post explains that the site compiles Stockfish to WebAssembly so the player's own computer does the analysis, at about 7,000 unique visitors a day, which keeps hosting cheap and the core report free. [7]
- 8 July 2025CA blog post describes the planned AI chess coach that reads a player's whole game history. The current version of that post says the coach is live. [8]
- February to April 2026AThe pricing page sells a US$3-a-month supporter plan (no ads, a Discord badge) and lists a Pro plan with "Advanced AI coaching" at US$9 a month as coming soon, with a waitlist. Verified revenue is US$21 to US$42 a month. [1, 4]
- May 2026ASupercoach goes on sale. On 28 May the pricing page says "Class #1 grabbed $9/mo and sold out", offers 200 more seats at US$14 a month with 174 left, and includes an AI coach named Tommy with up to 200 messages a month. Verified revenue for May: US$3,738. [1, 5]
- June to September 2026AVerified revenue rises each month: US$7,828, US$13,654, US$15,775 and US$16,651. [1]
- 15 August 2026CEliot publishes "Is Using Chess AI Cheating?", which says that engine help during a rated game is cheating and that engine analysis after the game is training. [13]
- 25 October 2026AThe plans cost US$14 a month, US$97 a year or US$490 once, and the coach is now called Froggy. TrustMRR shows US$7,622 of MRR from 593 active subscriptions and US$17,454 in the last 30 days. [1, 2]
Why did Chessigma work?
Four causes, in the order they mattered. Each one is a reusable lesson.
Free where the leader charges
Chess.com gives free users one full game review a day; Chessigma gives unlimited reviews with no account. That one difference brought the first wave of users from Reddit and a creator video, and the free tools (a best-move finder, rating calculators, opening guides, and pages translated into five more languages) keep bringing visitors.
The user's computer does the work
Stockfish runs in the browser as WebAssembly, a format that lets compiled code run inside a web page. The founder wrote in May 2025 that this keeps the hosting bill low, so a free product with hundreds of thousands of visitors does not need investors to pay for servers.
Sell the part an engine cannot do
An engine says which move was best but not why a player keeps losing. The paid plan turns many games of engine analysis into a report on the player's habits, daily drills built from those mistakes, and a language-model coach that answers questions with the player's own numbers. Revenue stayed under US$50 a month for four months of a supporter plan and passed US$3,700 in the month the coach went on sale, by TrustMRR's figures.
Founding classes and price steps
The coach was sold in numbered founding classes: the company's pricing page reported that the first class at US$9 a month sold out, and the second class at US$14 had a limited number of seats. A yearly plan and a lifetime plan bring cash up front, which is why monthly revenue is about twice the MRR.
How it was built
These details come from the company's own pages, its privacy policy and the founder-entered part of the TrustMRR profile. They are grade C. [1, 7, 9, 15, 20]
- Next.js for the web app, Supabase for the database, sign-in and storage, Vercel for hosting, and Stripe for website payments (TrustMRR tech stack and the privacy policy).
- Stockfish compiled to WebAssembly runs on the player's device. Cloudflare delivers the engine files, and Amazon Web Services runs server-side analysis of imported games.
- The coach and other chess processing run on Modal, a cloud service for running code on demand. OpenRouter, a service that forwards requests to many AI model providers, sends the coach questions to the models that write the answers. The pages do not name the model.
- Resend for account and product emails, PostHog for product analytics, Google AdSense ads for free users, and RevenueCat for subscription records of the phone apps.
- A desktop app for Mac, Windows and Linux that shows the live site in its own window, and lighter iPhone and Android apps with game review and the opening trainer only.
- A public MCP server, a connector that lets Claude, ChatGPT and other AI assistants call Chessigma's chess tools, with links back to the site.
Tools named in the sources
Each one opens its page in the Tools tab, with every story that uses it.
Who tried this before Chessigma, and why did they fail?
Four companies in the Loot Drop database taught chess, hosted a board game online, or sold personalized learning before language models could explain a player's own mistakes. The values below are from that database and are not independently verified.
Chesscademy
The promiseInteractive chess lessons and videos that make learning chess easy for beginners and intermediate players.
Why it diedFree chess content was everywhere, Chess.com offered lessons and a community in one place, and a freemium model without a paid feature people wanted did not bring enough revenue.
Kaya.gs
The promiseA modern online platform for the board game Go with new features for beginners and experienced players.
Why it diedTechnical debt and a small team could not deliver reliable real-time play, and voluntary donations did not cover server costs and development.
Knewton
The promiseAdaptive learning: software that analyzes each student's answers and recommends a personal learning path.
Why it diedThe technology was complex and costly to put into schools and publishers, free platforms such as Khan Academy grew faster, and newer machine-learning tools removed its technical lead.
Juni Learning
The promisePersonal one-to-one online coding and math lessons for children, with vetted instructors.
Why it diedEvery lesson needed a paid human instructor, customer acquisition was expensive, and after the pandemic families left faster than the costs could be covered.
The failed startups come from Loot Drop, a free database of dead startups, retold here in our own words. Loot Drop describes its entries as AI-assisted summaries of public sources, so these cards carry no evidence grade.
What changed
Chesscademy and Kaya.gs show the old problem of board-game sites: the basic service (lessons, a place to play) was available for free elsewhere, and the extras did not make people pay. Juni Learning shows the other side: personal teaching does make people pay, but a human teacher for every student cost more than the price could cover. Knewton tried to automate personal teaching with its own software and raised US$182.3 million for it.
Chessigma uses two things those companies did not have. Stockfish became free, very strong and able to run inside a web browser, so a two-person team gives away unlimited analysis at almost no cost. Language models, rented per answer through OpenRouter, turn that analysis into explanations in plain words about one player's games. The free part brings the visitors; the coach is the part they pay for, at US$14 a month instead of the price of a human chess teacher.
Can one person copy Chessigma now?
The pattern, yes. This market, with care.
Other chess sites also sell analysis and training, and Chess.com, the largest, limits free game reviews in order to sell its premium plan. Chessigma's free tier depends on public game data from Chess.com and Lichess, and the founders already changed their move labels after a trademark discussion with Chess.com. The pattern is general: give away the expensive-looking part that can run on the user's device, and charge for a personal layer that a language model adds on top.
Where it is still open: The same structure fits other games and skills where a free engine or solver exists but explains nothing: Go and shogi engines, poker solvers, bridge, and typing or music practice with automatic scoring.
- 1Find a feature that the market leader limits for free users and that you can run on the user's own device at almost no cost.
- 2Give it away without an account, and post it where the players already are (Reddit, chess creators).
- 3Build pages for every small search the players make: calculators, opening guides, a best-move finder, translated versions.
- 4Sell a personal layer on top of the free data: a report on the user's habits, daily drills from their mistakes, and a coach that answers with their own numbers.
- 5Open the paid plan in small, numbered classes at a low first price, then raise the price for the next class.
What could go wrong for Chessigma?
The risks are the mirror image of the causes above.
Growth has slowed
Monthly revenue grew about 3.7 times from May to July 2026, but only 5.6% from August to September, and the last 30 days to 25 October were 3.7% below the 30 days before. Two days in October brought more than a quarter of the last 30 days of revenue.
Cash revenue is about twice the recurring revenue
MRR was US$7,622 while the last 30 days brought US$17,454. Yearly plans at US$97 and lifetime plans at US$490 are paid up front, so part of each month's revenue does not repeat. The recurring part is the more stable measure.
The fair-play line
The free best-move tool returns Stockfish's move for any position. The company's pages say that using an engine during a rated game is cheating and list study uses only, and the paid plan sells training, not live help. Misuse by some visitors is still possible, as with any engine.
Claims to customers that do not agree
The pricing page says players gain 112 Elo rating points in their first month, and the Supercoach page says 91. Neither figure has a stated method.
Dependence on the big chess sites
The free tier imports games from Chess.com and Lichess, and the paid plan reads the same history. A change in their public data access would hit both layers.
Sources
The grade next to each source is the best grade that the source can support. "Read in full" means that the text of the page was downloaded and the numbers were checked in it.
- A TrustMRR markdown profile: Chessigma, revenue synced 25 October 2026Read the markdown profile in full on 25 October 2026. Revenue metrics are read from Stripe through an API key; the description, tech stack and founding date are founder-entered (grade C).
- C Chessigma pricing page, 25 October 2026Read in full. Source for the plans, prices, plan contents, the refund guarantee and the 112 Elo claim.
- C Chessigma Supercoach page, 25 October 2026Read in full. Source for what the paid plan does, the founding classes, the customer count and the 91 Elo claim.
- C Chessigma pricing page, Wayback Machine copy of 16 April 2026Read in full. Shows the US$3 supporter plan and the AI coaching plan marked coming soon.
- C Chessigma pricing page, Wayback Machine copy of 28 May 2026Read in full. Shows Supercoach on sale, the sold-out first class at US$9, the second class at US$14 and the AI coach with 200 messages a month.
- C Chessigma blog, 6 April 2025, by Mehdi: launch postRead in full. Source for the launch date, the Reddit post, the creator video and the first-week visitors.
- C Chessigma blog, 15 May 2025, by Mehdi: Free Chess Game Review and Analysis (how the analysis works)Read in full. Source for Stockfish in WebAssembly, the daily visitors in May 2025 and the move labels.
- C Chessigma blog, 8 July 2025, by Mehdi: The AI chess coach that actually knows your gameRead in full. The post has been updated since its date; used for the plan of the coach.
- C Chessigma privacy policy, read 25 October 2026Read in full. Names Arche Labs Ltd and lists the service providers, including OpenRouter for the coach answers.
- C Chessigma terms of service, read 25 October 2026Read in full. Names the operator, Arche Labs Ltd, trading as Chessigma, and the payment and refund terms.
- C Chessigma home page, 25 October 2026Read in full. Source for the free analysis, the AI coach offer, the founder links and the author name in the page data.
- C Chessigma Next Move tool pageRead in full for the fair-play check: the listed uses and the answer that engine use in a rated game is cheating.
- C Chessigma blog, 15 August 2026, by Eliot: Is Using Chess AI Cheating? Where the Line Actually IsRead in full for the fair-play check.
- C Chessigma blog author page: EliotRead in full. Source for Eliot's role as co-founder and Amazon software engineer.
- C Chessigma download pageRead in full. Source for the desktop and phone apps and which features each includes.
- C Stockfish home pageRead the section on NNUE evaluation.
- C Ship or Die profile of Mehdi Ben HaddouRead in full. Lists his other shipped products (Tweet.rip, PolyMRR, Homefast), whose home pages were read for the purpose check.
- C Loot Drop database export, 1,749 failed startupsLocal export. AI-written summaries, not independently verified. Source for the failed predecessors.
- C humansfix.ai developer profile of Mehdi Ben HaddouRead in full. Self-written profile; source for his earlier jobs.
- C Chessigma MCP server page for developersRead in full. Source for the connector for Claude, ChatGPT and other AI assistants.