How to analyze Korean stocks
PRISM-INSIGHT splits Korean stock analysis across 13 agents, generates reports, and executes strategies. The author separately reports simulation returns and real-money account performance.
news.ycombinator.comSuggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
Use [Korean market dataset] to build a simulation of a multi-agent stock-analysis workflow. Divide work across [technical analysis], [trading flows], [financials], [news] and [market conditions]. Generate reports and log simulated decisions with their reasoning in [dashboard]. Keep simulation results separate from any real-money records.
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source author
- prism_insight
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- The author reports 408% across 51 trades in the first simulation season, 11% realized returns in the second, and a separate $10,000 account up 9.35% since late September.
- Human involvement
- The author published trading reasoning and performance; the post does not describe human approval of each trade.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- Automatically detected surging Korean stocks twice daily.
- Generated reports using specialized collaborating analysis agents.
- Used GPT-5 for trading decisions and executed strategies.
The author built PRISM-INSIGHT, an open-source system in which 13 specialized agents collaborate on Korean stock analysis. It covers KOSPI and KOSDAQ stocks and automatically detects surging stocks twice daily. Agents divide analysis across areas including technical indicators, trading flows, financials, news, and market conditions, then generate reports and execute trading strategies.
MCP servers provide access to live market data, web search, and financial APIs. The author uses GPT-4.1 for analysis, GPT-5 for trading decisions, and Claude Sonnet 4.5 for the conversational bot. Reports and alerts are available through Telegram, while a dashboard exposes trades, decisions, performance, and agent reasoning. The implementation uses Python, SQLite for trade history, Playwright for PDF reports, and matplotlib for charts.
The author distinguishes simulation performance from a real-money account. They reported a first simulation season returning 408% across 51 trades, a second season with 11% realized returns against KOSPI's 16%, and a $10,000 account up 9.35% since late September. These are the author's figures; the post does not describe independent verification or human approval of individual trades. Who it's for: developers researching transparent multi-agent market-analysis systems.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Hacker News · prism_insight
Frequently asked questions
Which models handled analysis and trading decisions?
The author used GPT-4.1 for analysis, GPT-5 for trading decisions, and Claude Sonnet 4.5 for the conversational bot.
Were the reported returns all from real-money trading?
No. The author reported 408% for the first simulation season and 11% realized returns for the second. Separately, they reported a $10,000 real-money account up 9.35% since late September.
More details
Remaining steps
- Displayed trades, decisions and reasoning on a public dashboard.
About this account
The system automatically detected stocks, generated reports and executed strategies using specialized agents and market-data tools.
Platform: Hacker News. Attribution: https://news.ycombinator.com/item?id=45946056. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: prism_insight on Hacker News
Source published
Added to Agents In Use
Source checked
Tools and topics
PRISM-INSIGHT · MCP servers · GPT-4.1 · GPT-5 · Telegram · SQLite
Stock analysis · Multi-agent · Korean markets
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.