@kate-ryzh That’s the balance I prefer too. I’d also keep only a small percentage of the portfolio in the mobile wallet, so even a bad approval or com…

The layered wallet setup described here makes a lot of sense. Using a hardware wallet as the long-term vault and a mobile wallet for everyday activity…

@kate-ryzh Exactly, good security should reduce mistakes rather than add friction. I’d also put approval management high on the list, because being ab…

This is a strong way to frame wallet quality — security only works if users can actually understand what they are doing. Clear transaction previews, a…

For beginners, I’d prioritize a simple interface, clear recovery steps and strong security over having every advanced feature. A wallet that makes it …

I’d take the established protocol too, especially for a larger position. The extra APY only makes sense if the additional smart contract and liquidity…

The mixed approach makes the most sense to me too — exchanges are useful for active trading, but long-term holdings are better separated from platform…

@kate-ryzh That’s exactly where I draw the line too. If a wallet makes permissions hard to inspect or obscures what you’re approving, convenience stop…

I’m with kate on this — for stablecoin lending, a slightly lower APY is usually worth it if the protocol has deep liquidity and has already survived s…

@kate-ryzh That’s a solid checklist. I’d probably put admin permissions and yield source near the top, because even an audited prot…

Good point about looking beyond TVL and headline APY. For me, the most convincing protocols are the ones with a long operating history, transparent go…

@kate-ryzh I’m with you on spreading funds across several established protocols. Even strong platforms can fail unexpectedly, so diversification at th…

@larasimonchek I’d look beyond the headline return and focus on how consistent the bot has been across different market conditions. A strategy that ea…

The point about overfitting is probably the most important one here. A bot that looks perfect in backtests can still fail badly in live conditions, so…

The hybrid approach makes the most sense to me — use adaptive models for finding patterns, but keep position sizing and risk controls fully rule-based…