I was curious about this since it kind of makes sense, but I offer a few reasons why I think this isn't the case:
- In the 10% noise case at least, the second descent eventually finds a minima that's better than the original local minima which suggests to me the model really is finding a better fit rather than just reducing itself to a similar smaller model
- If it were the case, I think we might also expect the error for larger models to converge to the performance of smaller models? But instead they converge lower and better
- I checked the logged gradient histograms I had for a the runs. While I'm still learning how to interpret the results, I didn't see signs of vanishing gradients where dead neurons later in the model prevented earlier layers from learning. Gradients do get smaller over time but that seems expected and we don't have big waves of neurons dying which is what I'd expect to have the larger network converge on the size of the smaller one.
I built a personal finance app (https://tender.run) in the style of mailbox (swiping, keyboard shortcut, inbox-based workflow for reviewing transactions).
It's built on the automerge CRDT and sqlite running in the browser, which has been really fun to work with. I'd like to keep going, though honestly I've struggled with the marketing side (growth has been slow) and it's a pretty competitive space.
Originally this was a privacy angle - the data is primarily stored on your device, with backend storage that’s treated like backup and sync only. I had plans for e2ee that built on this but it didn’t turn out to be a big differentiating factor.
Working in local-first turns out to be really nice for making the app feel super snappy. The responsiveness you see in the demo is the performance you can expect in day to day usage.
Sort of relatedly, I’ve been fighting Safari bug for years that feels like it has to be related to dates in js.
Safari’s saved credit card support works pretty well, but when I switch between US and Asia timezones the expiration date will shift by a month in one direction. My best guess is that they store the expiration as midnight local time on the 1st of the month in a js Date which can shift when you travel outside of that timezone.
I’ve been building a personal finance app that runs fully in the browser (using the automerge crdt and sqlite) for over a year now at https://tender.run.
Recently I’ve been taking more of being able to flexibly run sql against this data, and this past week I’ve been working with d3 to make fancy sankey graphs to show income/expense flows. Quick preview here: https://demo.tender.run/reports/sankey
It's sad to hear that the new app encrypts everything.
A long time ago I worked on hacking airplay support for sonos speakers and it wouldn't have been possible without inspecting a lot of plaintext wireshark traffic.
As much as I've have cited, loved, and recommended sourcegraph (even going so far as to help run the open source version at a previous co), I never paid a cent for the product.
I'm curious about the line of thinking in leaving open source behind, but it seems somewhat unsurprising in that lens.
why would consumers who can afford to use credit cards with rewards switch to fednow payments? are there equivalents for chargebacks, fraud protection, and rewards percentage?
I just tried to do this with chase, capital one, and citi. The first two only let you pay up to 10% more than your balance (balance, not credit limit) and citi only 7.5% - nowhere near 10x your credit limit.
I run a personal finance app that's built as a PWA (https://tender.run, my email is in my profile).
I would say running as a PWA has been a mixed bag. There are quite a few missing features across web platforms (for starters, background sync, full featured push notifications, haptics) that make it hard to be competitive with native apps. Every WWDC, we watch in hopes that Safari gets more PWA features, but it's a thin drip. There's a reason so many apps repackage into electron/cordova/whatever shells.
As other folks in thread have pointed out, it's probably worth thinking through why a product works particularly as a PWA vs alternative distribution methods.
It doesn't seem like it will i.e. _delete_ your username/password, but the behavior is actually automatic without user prompt.
See 1:10-ish[0] for the demo - I found this at least somewhat surprising, though I submit this as a passkey advocate (for a while, my side business only-supported passkeys, but we backed away from that).
We haven’t written much about our server-side sqlite. I’d say our write workloads are easily within what sqlite can do for now. At one point we looked into doing per-customer databases for scaling write perf (since sharding by customer can be a lot cheaper in sqlite than other dbs), but it’s not necessary yet.
We run sqlite at https://tender.run, both in our backend (as a primary data store, backed by litefs) and on the client (as a index and cache).
In general works really great, esp. in terms of speed. There are some limitations worth knowing though - for instance, figuring out making write throughput reasonable with the combination of wal mode, busy timeout, and begin immediate.
Another that I've had to work around recently is the lack of materialized views - you can make and query from a view, but it's just a select statement that gets shoved into queries that use it. To get the perf of a materialized view, you'd have to do something manual and custom. For the most part, this isn't an issue simply due to how fast sqlite is, but in the browser it's a bit more significant because of how much slower the wasm-compiled sqlite is.
I was curious about this since it kind of makes sense, but I offer a few reasons why I think this isn't the case:
- In the 10% noise case at least, the second descent eventually finds a minima that's better than the original local minima which suggests to me the model really is finding a better fit rather than just reducing itself to a similar smaller model
- If it were the case, I think we might also expect the error for larger models to converge to the performance of smaller models? But instead they converge lower and better
- I checked the logged gradient histograms I had for a the runs. While I'm still learning how to interpret the results, I didn't see signs of vanishing gradients where dead neurons later in the model prevented earlier layers from learning. Gradients do get smaller over time but that seems expected and we don't have big waves of neurons dying which is what I'd expect to have the larger network converge on the size of the smaller one.