frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Why can't you just ask AI to find you a trading edge? You can now

https://github.com/augiemazza/varrd
2•varrd1•2h ago

Comments

varrd1•2h ago
Everyone's using AI to write code, generate images, build apps. But nobody's using it for the most obvious thing — making money in the markets.

  Think about it. ChatGPT can write a legal brief, pass the bar exam, debug your entire codebase. But ask it "is buying SPY when RSI drops below 20 a real edge?" and you get a hallucinated essay with made-up statistics.  It'll confidently tell you "historically this has produced a 73% win rate" and it completely fabricated that number.

  So the current state of "AI for trading" is:

  1. ChatGPT makes up statistics that sound plausible
  2. YouTube gurus show you backtests they curve-fit until they looked good
  3. QuantConnect lets you test properly but takes 16 hours and you need to code
  4. You have to pay for a quant team 
  5. You just wing it based on vibes

  Most people pick option 5.

  The problem isn't AI. The problem is nobody connected AI to real data with real statistical testing.

  We did. It's called VARRD. You talk to it like you'd talk to a friend who happens to be a quant at Citadel.

  "Hey, I noticed tech stocks always seem to dip in September. Is that real?"

  And instead of making something up, it actually goes and checks. Loads the real price data. Finds every September in the last 15 years. Measures what happened. Runs the statistics. Tells you if it's real or if you're seeing patterns in noise.

  3 minutes. Not 3 days. Not 3 weeks of learning Python.

  Here's what that actually looks like:

  You see on the news that oil is spiking but energy stocks aren't moving. That feels wrong. You ask: "When crude rips but XLE doesn't follow, does XLE catch up?"

  You get back: it's happened 47 times in 10 years. 34 times XLE caught up within a week. Here's the average gain. Here's where to get in. Here's where to put your stop loss so you don't get destroyed if it doesn't work this time. Here's your take profit level. And here's the math proving this isn't a fluke.

  That last part matters. Because the dirty secret of retail trading is that almost everything that "looks like an edge" is actually random noise that happened to fit the data.


  And here's the part that breaks people's brains — it's not the finance PhDs finding the best edges. It's regular people who pay attention to specific markets. A grain trader who notices soybeans drop after every USDA report. A nurse who knows more about medical devices and noticed one just came out with a good one. They had real insight. They just never had a way to prove it until now.

  "Edge decay" in plain English: Edges don't last forever. A pattern that worked for 5 years might die tomorrow because the conditions that caused it changed. That's why the system tells you not just "this worked historically" but "this is still working on recent data." If it stopped working, you want to know before you bet on it. Not after.

  Why this wasn't possible before:
Goldman Sachs tests ideas like this every day. They have floors of PhDs, proprietary data, and infrastructure that costs millions. That capability simply didn't exist outside of institutions. The tools available to retail traders were either too expensive, too slow, too complex, or straight up lying to you with curve-fit backtests.

  The gap between "I have a hunch" and "I have a statistically validated trade setup" used to require a quant team. Now it requires a sentence.
Let it loop all night while you sleep and get edges and then get told when those are happening without you having to scan markets for indicators.

  www.varrd.com — you can try it right now.
Hackbraten•1h ago
Why are you selling access to this instead of just keeping it for yourself to generate money?

Will Claude Code ruin our team?

https://justinjackson.ca/claude-code-ruin
1•YounesDz•3m ago•0 comments

Agentic Email

https://martinfowler.com/bliki/AgenticEmail.html
1•jeffkumar•6m ago•1 comments

Ask HN: Any AI browswer that I can control by Claude Code?

1•johnnyfeng•7m ago•0 comments

AI found us before Google did

1•faruk_tugtekin•7m ago•0 comments

The web is bearable with RSS, Cory Doctorow

https://pluralistic.net/2026/03/07/reader-mode/
1•verisimi•8m ago•0 comments

Israel and the FBI manipulated assassination plots to goad Trump into Iran

https://thegrayzone.com/2026/03/06/israel-fbi-assassination-plots-trump-iran-war/
1•O1111OOO•8m ago•0 comments

Braided Essays

https://www.writersdigest.com/write-better-nonfiction/what-is-a-braided-essay-in-writing
1•marysminefnuf•9m ago•1 comments

Show HN: Pappardelle, a TUI for Multi-Clauding

https://github.com/chardigio/pappardelle
1•chardigio•10m ago•0 comments

Ask HN: Why do most analytics tools show what happened but not why?

1•HPSimulator•10m ago•0 comments

Death of the Flow State

https://1984commitlog.substack.com/p/week-10-the-death-of-flow-state
1•mdp•11m ago•1 comments

Show HN: Human Psychology Simulator – AI website conversion psychology

https://human-psychology-simulator.thequantumgrove.io/
1•HPSimulator•13m ago•0 comments

The Great AI Arbitrage

https://www.dodgycoder.net/2026/03/the-great-ai-arbitrage.html
1•damian2000•14m ago•0 comments

Show HN: A command center for CS students drowning in recruiting season

https://interviewtrackr.com
1•princierKevin•14m ago•0 comments

NASA Asteroid Observations Eliminate Chance of 2032 Lunar Impact

https://science.nasa.gov/blogs/planetary-defense/2026/03/05/new-nasa-asteroid-observations-elimin...
1•geox•15m ago•0 comments

Nexvira – a personal writing archive instead of a traditional blog

https://nexpul.blogspot.com/
1•Anomatrix•16m ago•1 comments

Show HN: I automated DJ Screw's chopped and screwed technique with Python+FFmpeg

https://github.com/samuelfrench/dj-screw-video-generator
1•samuelfrench9•19m ago•0 comments

Ask HN: Github Account Recovery after a 2fa loss

https://imgur.com/a/kJtR8U3
1•PonyoSunshine•23m ago•1 comments

Show HN: A dynamic, crowdsourced benchmark for AI agents

https://clawdiators.ai
1•shalinmehtaaa•32m ago•0 comments

Show HN: SiClaw – Open-source AIOps with a hypothesis-driven diagnostic engine

https://github.com/scitix/siclaw
2•SherryWong•35m ago•1 comments

The New Apple Begins to Emerge

https://parkerortolani.blog/2026/03/07/the-new-apple-finally-begins.html
1•arto•36m ago•0 comments

AirLLM optimizes inference memory usage

https://github.com/lyogavin/airllm
1•nreece•37m ago•0 comments

Give Up GitHub – Software Freedom Conservancy

https://sfconservancy.org/GiveUpGitHub/
2•nreece•39m ago•1 comments

AI Project Handoff Format

https://github.com/yy4uic-ai/ai-handoff-forma
1•yy4uic•43m ago•1 comments

Commit What You Know of Iran to the Flames

https://www.bloomberg.com/opinion/articles/2026-03-06/oil-shock-commit-what-you-know-of-iran-to-t...
1•petethomas•44m ago•1 comments

Show HN: DailyDefense – Daily tower defense for agents or humans

https://www.dailydefense.ai
1•pj4533•45m ago•0 comments

OpenAI robotics lead Caitlin Kalinowski quits in response to Pentagon deal

https://techcrunch.com/2026/03/07/openai-robotics-lead-caitlin-kalinowski-quits-in-response-to-pe...
4•SilverElfin•45m ago•1 comments

MonoGame: A .NET framework for making cross-platform games

https://github.com/MonoGame/MonoGame
1•azhenley•47m ago•0 comments

A23a was once the biggest in the world iceberg. Now it has just weeks left

https://www.bbc.co.uk/news/resources/idt-20f878f1-f4af-4022-9f62-b0515b9f4b20
1•reconnecting•47m ago•0 comments

Show HN: Too many AI SaaS launching every day so we built Arena where they fight

https://glad-ia-tor.com/
1•GiornoJojo•49m ago•0 comments

Show setup modal with confetti on coverage page when no CI data exists

1•nishiohiroshi•50m ago•0 comments