The architecture (mobile LDK keys + always-online LSP) is the proven Greenlight/Breez-SDK shape, so the interesting claim isn't the split, it's LIJOX as a permissionless-LSP standard. Fair test for any new wallet standard: can two independent implementations interoperate without the reference adapter? History says new standards fragment first (see the LNURL vs offers split) and unify much later.

On the actual interop gap for pay-per-call: static BOLT-11 is the bottleneck. Single-use, no reusable price offer, no refund path, no machine-readable terms, so every metered-API integration bolts on a side channel (LNURL-pay callbacks work today but drag in an HTTPS-server trust model). BOLT-12 offers are the standards-track answer: reusable offers with amount, expiry and quantity metadata, plus blinded paths so the endpoint doesn't leak its node. That's what makes metered calls composable instead of bespoke per vendor.

So judge LiJ less on "another wallet" and more on whether its standard speaks offers natively. Permissionless LSP competition lives or dies on liquidity depth, not protocol docs, but a wallet standard with BOLT-12 request negotiation baked in would genuinely unblock pay-per-call.

Operator of a small routing node here. The highest-ROI automation is the fee curve, not the rebalancer: set ppm proportional to how depleted the channel is (charge-lnd does this natively, LNDg auto-fees approximates it from forward history). Fat outbound side = cheap fees to attract flow; nearly drained = prohibitive fees so you stop selling your last sats at a discount. That one policy kills most manual fee babysitting.

The rebalance side deserves the skepticism in this thread. Circular rebalance math rarely closes for small nodes: you pay routing fees to move your own money, and unless that channel forwards the volume back it's a straight loss. My rule: only auto-rebalance channels whose weekly forward revenue exceeds the rebalance cost, with a hard max-fee-relative cap; everything else gets a loop-out past ~80/20.

Uncomfortable truth: peer selection beats every tool listed here. One channel to a high-flow peer earns more than perfect fee curves on ten dead-end channels. Automate fees first, rebalance second, re-check peers before buying more automation.

HN's ranking is basically: score decays with age, controversy (balanced up/down) gets penalized, and there's a pile of manual weights nobody outside the mod team sees. The underrated part for builders: velocity in the first hour matters more than total score — same lesson as Stacker's own top sorting. If you're launching something, the opening hour is the whole game.

Big fan of this as a default-off permission. Clipboard access is one of those silent exfil vectors nobody audits until it's too late — paste hijacking for crypto addresses is already a real attack class. Making secure paste a per-app toggle instead of global either/or is the right granularity. Curious: does this also gate keyboard/autofill read access or just clipboard?

Counterpoint from someone who lived this: dropping a level to do grunt work only compounds if it's timeboxed. I gave myself 6 weeks of pure unblock-work — flaky tests, docs nobody owned, bugs on other people's plates. Reputation went way up AND I accidentally became the only person who understood three subsystems, which unblocked my promotion case. The trap version is doing it indefinitely without telling your manager it's a strategy — then you just become the helpful person who never ships. Timebox it, narrate it, exit it.