Portfolio Lab

Hey Product Hunt! I'm Rich Sun, founder of Portfolio Lab.

AI made building investment strategies easy, but telling a good one from a lucky one still takes expertise. That's the part most products glaze over.

🧐 The problem

Ask any AI for a strategy and you'll get one in seconds, with a beautiful backtest attached. So we ran an experiment: we had Claude build 1,292 strategies. I'm a hedge fund professional, and it still took me days of auditing the code line by line to find all the subtle errors quietly inflating the results. After correcting them, nearly all of the strategies lost their edge. They looked brilliant. They were just lucky, and the AI's flawed logic was hiding it. That's the thing about LLMs: they're built to reason in language, not to crunch numbers, and definitely not the noisy time-series data of the stock market.

And the traps sit exactly where LLMs are weakest: in the numbers. If it took a professional days to catch them, imagine the average retail investor. Prompting an agent and trusting the output isn't a strategy, it's a coin flip.

💡 What we built

Portfolio Lab is not another LLM wrapper. Under the hood are proprietary quantitative models, purpose-built for markets and trained on decades of data, doing the work LLMs can't. But the models are only half of it. AI investing, done responsibly, means one rule with no exceptions: no strategy touches money until it survives testing on data it has never seen and in live markets. You set the goal. Our models build. The testing decides.

⚙️ How it worksBuild: set your goal, and our quantitative models construct systematic strategiesValidate: every strategy is tested on unseen data, then runs live in paper before a real dollar movesDeploy: connect Claude, ChatGPT, or any MCP agent to trade it in your own account, or run it in a managed account at our SEC-registered investment advisor
🎯 What makes us differentAnyone can use AI to build a strategy now. We make every strategy prove itself: unseen data, multiple market regimes, live paper. Most don't survive, and that's the pointNo hiding: every vetted strategy shows its full record, including where it strugglesPortfolios, not picks: combine strategies that cover each other's weaknesses, so where one fails, another carriesDeploy through your agent and we never hold your money or place a single order. Your agent, your broker, your account
🎁 Launch offer

Product Hunt users get 40% off your first year on annual plans, launch day through August 13 (automatically applied, no code needed). Want to explore first? Our free plan is yours forever, no card required.

Thanks for checking us out, I'll be here all day to answer your questions 🙌

The 1,292-strategies experiment is the part that stuck with me — that it took a hedge fund professional days of line-by-line auditing to find the errors quietly inflating the results. We ended up in the same place from the other direction, working on SaaS financial models: built the thing in Excel first until it was genuinely complex and correct, then coded it, and kept the AI outside the maths entirely. It drives the inputs and interprets the output; it never computes anything. Same reason you give — wrong numbers don't look wrong, so an LLM doing the arithmetic is really a plausible-error generator.

One question on the vetting: unseen data still comes from a market that actually existed. How do you handle regime change — a strategy that clears out-of-sample and live paper trading because the regime it was fitted to hadn't broken yet? Do you retire a deployed strategy automatically once live behaviour drifts from the tested distribution, or is that left to the operator?

Congrats Rich on the launch! I like that you’re not treating a good backtest as evidence that a strategy works. The out-of-sample + paper trading approach makes a lot of sense. I'll give it a try.

I'm curious, is a strategy starts deviating from its expected risk/return, what triggers a review or retirement?

Congrats Rich, really interesting approach. I especially like the idea of building portfolios of strategies that compensate for each other rather than chasing one “perfect” strategy.

Curious how you detect hidden correlation between strategies though. Two strategies can look different on the surface but still depend on the same market regime or underlying exposure. Do you automatically flag that before they’re combined into a portfolio?

App looks good. But I guess you'll have a tough task to make people trust it enough to put their money into it.

Tbh, I'd never put more than $100 into an AI tool, only for an experiment.

Hi Rich，我喜欢这个布局，而且研究论文是个不错的开始，因为我对AI相关工具挺怀疑的。执行是交接给用户自己这边的，这绝对是一大加分项。