April 14, 2026
Vibes vs. portfolios—place your bets
Show HN: LangAlpha – what if Claude Code was built for Wall Street?
Open‑source “vibe investing” bot drops; fans cheer, skeptics say it prints losses
TLDR: LangAlpha launches an open-source AI assistant for ongoing stock research, pitching a smarter, memory-driven workflow. The thread explodes into a split-screen: skeptics warn it’ll torch savings, critics demand proof it works in the real world, and fans hail the bold, polished release as hacker catnip.
Show HN just got a Wall Street remix: LangAlpha promises “vibe investing” with a persistent AI research buddy that remembers your stock ideas, fetches data, and spits out morning notes with slick charts. The builders jump into the comments to say it’s fully open-source, works with any AI model, and comes with a full web app—basically “Claude Code, but for portfolios.”
The crowd? Absolutely split. The first hot take slams the brakes: “Then people would lose a lot of money.” That set the tone for the doomer vs. dreamer showdown. One camp demands receipts—“show us real cases, prove the data is right, not just vibes.” Another throws shade at the launch post itself, calling it “a low effort Claude response,” accusing the write-up of sounding AI-generated. Meanwhile, the hype squad rolls in with guitar riffs: “Never anger the gods of code,” complete with metal-horn ASCII art, cheering the audacity of shipping a polished, open-source finance agent.
Underneath the drama, the pitch is simple: instead of one-off questions to a chatbot, you keep a living research workspace that updates as the market moves. It’s a bold idea—and the comments are a cage match over whether that boldness leads to smarter decisions or expensive lessons.
Key Points
- •LangAlpha is an AI agent harness for investment research that maintains persistent workspaces for iterative, Bayesian-style analysis.
- •It deploys parallel subagents to gather market, news, and macro data and can produce morning notes with interactive visualizations.
- •Programmatic Tool Calling lets the agent write and run Python against MCP servers for complex analysis while reducing LLM token usage.
- •A finance research workbench offers a web UI with TradingView charting, live WebSocket market data, and subagent monitoring.
- •The system includes automations, a provider-agnostic LLM layer with failover, a 24-layer middleware stack, and security via pgcrypto and credential redaction.