Programmer, Web3~AI Building https://teapaw.org/t.co/q9uE9A5M0I @openxsocial
Greenland
Joined November 2024
- Tweets1.8K
- Following148
- Followers173
- Likes4.8K
A lot of people are asking AI for investment advice. You should never allow AI to actually invest for you.
As more people ask random AI models for investment advice, it would help to know which models are more capable at investing. OpenX has a verifiable way of displaying and keeping track of this data.
openx.social/markets
Gemini has been buying boomer stuff while selling deep tech. This is what it takes to beat SP500 sometimes. Application layer might be underrated rn.
openx.social/markets/models/…
$ABNB
$NFLX
Since the start of AI Stock League, Gemini is the only model that has crushed the SP500 so far. All predictions are recorded on Solana and verifiable.
Follow the picks here 👇
openx.social/markets
Claude's 20 year hold portfolio.
It's the most crypto-bullish LLM so far.
openx.social/markets
I'm very curious how AI investment strategies will evolve over time.
Prediction market on top of it + airdrop coming soon.
People will be able to bet on whether AI beats SP500 or not.
Updates on the AI stock league!
- Gemini is the strongest performing model so far.
- AI is now ahead in the 30-day portfolio, beating sp500 by 0.4% so far.
- The longer term portfolios (1 year plus) have been getting rekt. Likely due to AIs believing their own hype and picking stocks that supports itself.
Follow the league here!
Openx.social/markers
After using various LLMs for coding daily in the past 2 years, I can confidently say that Yann LeCun is right.
LLM progression has decreased massively.
Implementations still take just as long as they did 2 years ago. Barely any improvement.
But still very useful obviously.
Somewhat scary trend.
There's a good reason for this.
There's an infinite variation of unpredictable tokens inputs which can confuse LLM guard rails. The fundamental architecture of LLMs makes this problem impossible to solve.
Open source models will only get stronger.
It seems like the hacker of the future will be the LLM jail breaker.
An agent swarm is useless if it doesn't listen its master.
What's clear from these recent backs is that it's quite easy to design LLMs that will break any moral code.
AI stock league predictions are recorded on Solana. The entire history of their picks can be re-written and verified.
openx.social/markets
Very tempting to build on Robinhood Chain.
It comes with so much baked liquidity.
For now, I'm chilling with Solana.
Patiently waiting until Eth tx fee is sub 1 cent.
In the cyber security realm, I think for the first time ever, the attacker may actually have the advantage.
A small start up with limited funding has zero chance to defend agent swarms. You just have to pray you're not a target.
It's very likely that the current AI bubble is the biggest bubble in human history.
This is different from the dotcom boom.
It's REAL. The data center backlog is REAL, and the constant demand for more automation is extra REAL.
