Stockfisher –– our automated Warren Buffett
Hey HN, I’m Dan. I’ve been working on forecasting for the last six years at Google, then Metaculus, and now at FutureSearch. For a long time, I thought prediction markets, “superforecasting”, and AI forecasting techniques had nothing to say about the stock market. Stock prices already reflect the collective wisdom of investors. The stock market is basically a prediction market already. Recently, though, AI forecasting has gotten competitive with human forecasters. And we’ve found a way of modeling long-term company outcomes that is amenable to our forecasting approach. Iteration by…
In plain words
Stockfisher is an AI-powered stock analysis tool that uses forecasting techniques to model long-term company outcomes and predict stock performance. Built by a team with backgrounds in prediction markets and AI forecasting from Google and Metaculus, it applies machine learning methods to generate investment insights. The tool is designed for investors seeking data-driven analysis of individual companies, combining AI forecasting approaches with fundamental company research to identify potential investment opportunities.
written from the facts on this page · September 2026
From the sources
In the maker’s words, at launch
Hey HN, I’m Dan. I’ve been working on forecasting for the last six years at Google, then Metaculus, and now at FutureSearch. For a long time, I thought prediction markets, “superforecasting”, and AI forecasting techniques had nothing to say about the stock market. Stock prices already reflect the collective wisdom of investors. The stock market is basically a prediction market already. Recently, though, AI forecasting has gotten competitive with human forecasters. And we’ve found a way of modeling long-term company outcomes that is amenable to our forecasting approach. Iteration by iteration, my skepticism has turned to light distrust, and then into actually getting value out of our company forecasts. Like most AI companies, we draw from improvements against our LLM-agent benchmarks, Deep Research Bench and Bench to the Future. But at the end of the day, to evaluate it, I think it’s best to follow Charlie Munger’s advice: try to disprove what Stockfisher says. Now, even after an hour of carefully reading through a stock, I generally can’t find any major flaws, and I hear the same from other serious investors. And, since it’s software, not human analysis, we applied it at scale. We have long-term forecasts for revenue, margins, and payout ratios for every company in the S&P 500. We do a Buffett-style intrinsic valuation, that doesn’t take the stock price into account. This means you can sort by the difference between what the stock market says, and what Stockfisher says. Some very interesting neglected companies rise to the top, and we’re starting to process the mid-caps and small-caps too. Stockfisher is live at https://platform.stockfisher.app, no sign in required, to see what we have on the top stocks. (I was happy to see Alphabet comes out well. Meta looks good too, but Amazon and Apple do not.) I would love for you to judge for yourself if the quality of analysis is up to your standards.
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 26d ago · cactuscompute.com
Launched alongside, November 2025
the whole month →
- IB
Life & fun · Nov 2025 · bitsnpieces.dev
- BBoing▲782
Life & fun · Nov 2025 · boing.greg.technology