mshish

I like how it applies different distributed system patterns to to solve a specific set of problems with Postgres and multi tenancy. This seems like a good example of the wider lens and scope AI dev tooling lets you take for new projects. Getting this functional would have taken ages only a few years ago.

I'm not defending the criticism of the CEO. I'm pointing out there are details we don't know about how Proton is using AI internally.

Article seems somewhat one sided to me. The style of the CEO doesn't sound good, but there's no discussion of specific tooling, harnesses, and other automation they have in place. Reading the letter posted, it sounds like they've got more in place than the article implies but people have been resisting adopting.

My understanding is that it's possible there will be / is a country that you don't have low cost or easy access to US stocks, but can have access to tokenized stocks. If Coinbase and others are smart, they'll collect the KYC info the US requires, regardless of where the person is in the world. In fact, that seems like the other part of this line in the article: "provided the tokens carry the same rights and privileges as the underlying shares, including dividends and voting rights."

I like the approach of making it real work and more conversational. At first I thought this would be what a few companies are doing with paid on the job work days for the last stage, which is possible at software companies nowadays thanks to AI dev tooling. However, it looks like they're focusing on reducing the top of funnel. Interesting, thanks for sharing.

Yeah and I would think wanting developers to update to 32/64 bit was also part of the equation.

I imagine official support for 3rd party app stores might open up some options for easily installing apps. However, that might require the store adding support for this type of iOS install or a newer version of iOS with 3rd party app store support.

Yeah, I agree this approach can produce usable results. The more narrow and well scoped you can make each decision the model has to make (e.g. binary vs 0 to 100) the better the results you'll get (and the easier it is to troubleshoot and tune).

Maybe, but I think one of the points is that the marginal cost of both understanding that legacy code base and safely refactoring it are increasingly trending down thanks to agentic coding tools. It changed the value and risk tradeoff for legacy apps, if deployed properly. There's still a lot of work (too much) to get the harness in place to make high confidence changes to legacy, critical apps. If someone makes that more automated, we'll be in a very different place for legacy refactoring and it will look a lot more like the clay molding the article discusses.

It's too bad you have to rely on signature spoofing to be able to use this as an alternative (but obviously I understand why). The spoofing looks narrowly scoped on LineageOS, but in general I imagine it creates some risk.