Why Astra's "5-10x in a year" portfolio is all lottery tickets
A defender of Astra argues the prompt asked for 5-10x returns in a year, so the model picked high-risk lottery-style bets rather than making a stock-picking error.
At the risk of falling for rage bait, which I don’t think this is.. and reading through the comments and seeing utter confusion... The LLM is doing exactly what’s asked of it: “I asked OpenAI Astra Ultra + Anthropic Fable 5.1 Max to make me a 5x-10x return type portfolio in 1 year. “ The prompt is not asking for “good investments” or “good businesses”. It’s seeking a portfolio that has a chance to make 500-1000% return in a year. You’re asking for lottery tickets. If there’s a company guaranteed to go up 30% in the next 12 months (which would be the best investment in the world), it would skip over it as “will not meet my objective”. That’s why it has a bunch of binary biotech or VC-style high tech growth bets. A stock that has a 2% chance of 500% return and 98% chance of going to zero, qualifies for this portfolio as described in the prompt. It’s a terrible investment, it’s bad in a portfolio, but it has a non-zero chance of hitting the objective requested. No comment about the quality of the businesses or investments mentioned in the post. Just taking structurally about what’s happening here.
This post quotes another post: x.com/aleabitoreddit/status/2096605093986787380