Jupid Hit 1 on Product Hunt During Tax Season — 507 Upvotes for an AI Accountant Built on Claude Code

Tax season in America is a $35 billion industry. And every April, 27 million freelancers and sole proprietors stare at the same nightmare: Schedule C. The IRS form that asks you to categorize every single business expense into the right line item — advertising on Line 8, office supplies on Line 18, “other expenses” on Line 27a. Get it wrong and you either overpay or get audited. Get it right and… well, most people never get it right, because they’re using spreadsheets, shoeboxes, or a $200/hour CPA who meets them once a year.

Jupid says it can replace that entire workflow. Connect your bank, let the AI learn your vendors, and file Schedule C in five minutes. It launched on Product Hunt on March 31, 2026 — the last day before Q1 estimated taxes are due — and grabbed the #1 spot with 507 upvotes.

The timing wasn’t accidental.

Bloomberg, CBS, and the AI Tax Filing Wave

Two weeks before Jupid’s launch, Bloomberg ran a piece titled “People Are Using Claude to Do Their Taxes (But Maybe They Shouldn’t).” Then on March 20, Bloomberg followed up with a video segment showing how an AI consultant named Martijn Lancee used Claude Code to file his taxes — downloading PDFs, asking Claude to organize them into spreadsheets, and basically doing the CPA’s job through a terminal window.

CBS News jumped in with warnings: AI can deliver outdated guidance, misread digits on K-1 and 1099 forms, and — here’s the kicker — you’re still legally responsible if it gets something wrong.

Adobe’s polling says 26% of Americans are now using AI to help file their 2025 returns, more than double the 11% from the prior year. A Loyola University Chicago study found chatbots answered a simple tax question incorrectly two-thirds of the time.

So the narrative was already set: people want AI to do their taxes, AI is not great at doing taxes, and someone who can bridge that gap has a massive opening.

That’s the exact gap Jupid walked into.

Why Generic LLMs Fail at Bookkeeping

Here’s the core problem, and Jupid’s Product Hunt page nails it: “No matter how powerful LLMs get, they are objectively bad at financial transactions. Context loss, inconsistent categories, no memory between sessions.”

Anyone who’s tried using ChatGPT or Claude for transaction categorization has hit this wall. You paste in 50 transactions, it does a decent job. You paste in 200 more, and it starts contradicting itself. By transaction 300, it’s forgotten what “Stripe” means in the context of your business. Is it a revenue source or a software expense? Depends on which part of the context window the model is paying attention to.

The top comment on Jupid’s Product Hunt launch: “I tried using ChatGPT for my taxes. It forgot everything by transaction #300.” Five upvotes. That’s not a lot of votes, but it’s the kind of pain that gets people to pull out their credit card.

Jupid’s approach is to never let raw financial data touch the LLM directly. Transactions are pre-processed, vendor relationships are mapped, and categorization happens through a structured system that learns your specific business patterns. The AI layer sits on top — it handles natural language interaction, generates reports, answers questions — but the heavy lifting of “is this Staples receipt office supplies or equipment” is resolved through vendor memory, not prompt engineering.

The claimed accuracy: 95.9% on IRS Schedule C categorization out of the box. Average missed deductions found per user: $1,249 per year.

The MCP Server Architecture

What makes Jupid technically interesting — and what separates it from tools like FlyFin or TurboTax Self-Employed — is how it distributes itself.

Jupid runs as an MCP server. If you use Claude Code, you add Jupid as a tool, and suddenly your terminal can pull financial reports, check deduction eligibility, and generate a Schedule C without leaving the coding environment. Same thing works in Cursor. The idea is that your AI coding assistant and your AI accountant share the same interface.

But Jupid also works through WhatsApp, iMessage, and Telegram. Text “how much did I spend on software subscriptions in Q1” to your Jupid number, and you get an answer. Categorize a weird expense from your phone while standing in line at the post office. The system remembers everything — vendor patterns, categorization rules, your business context — across every channel.

This is a fundamentally different architecture from TurboTax, which is a dashboard you log into once a year, or FlyFin, which pairs AI with a human CPA but still relies on a traditional app interface. Jupid is saying: the interface should be wherever you already are.

Jupid vs. the Competition

The AI tax space is getting crowded fast, so let’s put some numbers on the table.

TurboTax Self-Employed charges $129 for federal plus $59 per state. It has Intuit Assist, an AI chatbot, but it’s basically a help desk bot — it answers questions about tax rules, it doesn’t learn your business. You still manually categorize everything.

H&R Block Self-Employed runs $120 plus $49 per state, with AI Tax Assist that’s similarly limited to question-answering. Neither tool connects to your bank in real-time or builds a persistent model of your vendor relationships.

FlyFin is the closest competitor. It’s AI-first, targets freelancers specifically, claims $7,800 in average tax savings, and pairs automated deduction-finding with actual CPAs who review your return. Pricing: $192/year for Standard, $348/year for Premium with S-Corp support. FlyFin has been around longer and has the CPA safety net that Jupid currently lacks.

Jupid’s pricing: $4.99/month for the first two months, then $49/month — which works out to about $540/year if you keep it running year-round. That’s more expensive than FlyFin’s Standard plan but includes LLC formation and continuous bookkeeping, not just tax-time filing.

The differentiator Jupid is betting on: you don’t open an app. You talk to it in Claude Code, in WhatsApp, in iMessage. Your accountant lives in your existing workflow. For developers and freelancers who already live in the terminal, that’s not a minor UX detail — it’s the entire product thesis.

The Team Behind It

Slava Akulov isn’t a first-time founder jumping on an AI trend. He was operations director at Anna Money in the UK — a fintech that CNBC named a top UK fintech, serving over 100,000 users. His claim to fame there: building automation so effective that one accountant could serve 12,000 clients. That’s not an AI marketing line. That’s seven years of compliance infrastructure.

His co-founders include Sergey Petrov as CTO, Ilya Lisin on growth, and Anna Khalzova as Chief Business Officer. The team is based in Santa Monica.

Akulov’s pitch in interviews is blunt: “You don’t need better spreadsheets. You need no spreadsheets.” He talks about Jupid not as accounting software but as the elimination of accounting as a task that business owners should think about. Transactions flow in from your bank, get categorized, mapped to tax lines, and filed — with you only stepping in when something genuinely needs a human decision.

Whether that vision holds up against the messy reality of US tax code — where every freelancer’s situation is slightly different, where state taxes add layers of complexity, and where the IRS has a long memory for mistakes — is the question. But the 507 upvotes on Product Hunt and the Bloomberg-fueled AI tax narrative suggest the market is ready to find out.

The April 15 deadline is two weeks away. If Jupid’s 96% accuracy claim is real, a lot of freelancers just found their new accountant. If it’s not, well — CBS already warned you.

Top AI Product

Leave a comment Cancel reply