The AI part of this is a red herring. This is above all a big devops failure.
Three takeaways:
1. TEST YOUR BACKUPS. If you have not confirmed that you can restore, then you don’t have backup. If the backups are in the same place as your prod DB, you also don’t have backup.
2. Don’t use Railway. They are not serious.
3. Don’t rely on this guy. The entire postmortem takes no accountability and instead includes a “confession” from Cursor agent. He is also not serious.
4. See #1.
Running a single bad command will happen sometimes, whether by human or machine. If that’s all it takes to perma delete your service then what you have is a hackathon project, not a business.
> If my account tries to send someone $3 million, I'd prefer that it's intermediated by a confused bank employee staring at a screen
This is a nice lens for looking at when stablecoins make sense.
If you're an American using your Chase account to buy coffee at Starbucks, the permissioned, heuristically fraud-checked, slow-settling tradfi system is well optimized for you.
If you are an importer buying $3m worth of bulk coffee from Kenya, you would much rather have an instant 1:1 USD transfer on beautifully efficient machine consensus.
In many countries in the world, the banking system is extractive and unreliable. The "confused employee" is not there to help you. The two weeks of money in transit is no benefit, just a source of additional counterparty risk, cost, and delay.
An immutable and transparent ledger is not for everything but it is a useful primitive.
Many skeptics assume that stablecoins are just about regulatory arbitrage.
That's part of it, but:
1. Progress often depends on evolving obsolete regulation.
Uber works much better than taxis (once upon a time, people could "call a dispatcher" an hour in advance, wait on hold, etc) and yet in the early years they had to work around taxi regs.
2. Blockchains are a fundamentally more robust way to run a ledger.
If any of you have ever written software touching tradfi custody you'll know about "reconciliation"--start of every business day, you get a dump of files in your FTP server in various proprietary formats. You parse the transactions and they don't add up. The Recon team hand-corrects and recategorizes edge cases so that the balance deltas match transaction totals and everything ties out.
This type of absurd duct tape is ubiquitous, and it's a major reason why trad rails have multi-day settlement times and even longer for international. Inflates team size and cost required to run a product. SWIFT is a messaging system -- bankers use it to essentially text each other about wires to figure out issue resolution. Some lower-level trad payments regulations are written assuming that this level of manual oversight is required to prevent ledgering errors and ensure sound accounting.
Stablecoins run on transparent, precise ledgers with machine consensus. This doesn't solve everything, but there are large categories of issues that can occur in trad payments that do not exist onchain.
3. Control is liability.
Some important regulations actually encourage blockchain-based payments. For example, money transmitter law places significant requirements on custodial money transmitters (you take money from Alice, with a promise to give it to Bob) that do not apply to noncustodial channels (you give Alice a mechanism to send directly to Bob).
> The problem with all that, is the fact it remains possible to create a protocol with N big institutions [...] This maintains many benefits of the blockchain and lacks many issues (fast, simple, near zero cost)
That's more or less exactly what this is. Stripe is launching an EVM L1.
The Ethereum Virtual Machine part gives it a mature tech stack with experienced developers and auditors. Plus, well-tested smart contracts that have already processed billions of dollars on other chains can be deployed on Tempo.
The "Stripe L1" part will ensure that it's fast, simple, near zero cost.
> Maybe deleting the native apps and using the webapps will raise the barrier to entry high enough that I'll only use them when there's actually something I want to do on them.
That's worked pretty well for me. Also check out Kill News Feed and Kill Tweet Stream to avoid distraction if you need to use eg Facebook events or Twitter DMs.
> Social media weakens this skill because it’s engineered to be addictive.
Absolutely. App companies like FB and Twitter are on a constant quest to eat up as much of your time as possible, in pursuit of metrics like "daily active users" (they want you there daily) and "average session time" (they want you there as long as possible).
Ultimately, they're maximizing ad revenue. An hour of your time spent feed scrolling is worth about 10 cents to Facebook.
