@pcater64@danielcberk If you want to hedge against historic terrible markets, 3% is around the level that would mean you never run out of money. The reality is that most of the time, after the first decade, the risk of a crash doesn't matter any more and you can go higher
@VictorTaelin I'm trying to invest myself at 120mm, but people I've mentioned this to ask why you're only selling 6% of the company if the money is really helpful. If you invent what you're aiming to, having sold 25% at a lower valuation won't make any material difference to you.
@TacoJacc@VictorTaelin It used to be much worse, but now many areas of the city are a lot cleaner and don't smell like urine. The worst areas are worse than they used to be though.
@xsphi There's no way we switch all the way to Rust but then never switch to a language with dependent types and proofs. If we prove language guarantees are useful, and also that writing in "harder" languages is easy for AI, Rust will get left behind for something with more guarantees
@freed_dfilan My baseline assumption was that throwing enough/effort tokens at an existing NN should render it legible. I thought it was harder than that based on results coming from experts in the field like Nick. It doesn't take much evidence for me to revert back to my baseline
@VictorTaelin@GanjaRedNight Wait is there actually a way for an accredited investor to invest with less than $1mm? Or do you mean accredited and big checks
@VictorTaelin Did you try at any point having a subagents memory being the head of another tree instead of a single report message? Seems like that's the only place you kind of drop some knowledge
@VictorTaelin Most people can't make their dream game because is requires actually understanding what would be really fun for them. Even if claude could figure that out in theory, most people would ask to add features to ruin it
@codewithstefan@neural_avb I'm not trying to counter your point because I suspect you mean much more advanced stuff related to chip design but in case you both hadn't seen it and it's something you'd like there are things like this coming out: https://t.co/NDISCP5wsH
@tamarajtran@typesafeai Could you have this identify blocks of text to shuffle in and out of context based on the most recent turn? I mean obviously you could, would it be effective?
@PromptLLM This is a supply chain attack launched indiscriminately against the global supply of developer attention. I don't know what company you're trying to take down but you must really hate them.