wirelesspotat

I'm pretty sure the article is joking

> If the vulnerability were critical, someone would have merged it by now.

> GitHub Copilot can automatically suggest fixes for security vulnerabilities. Instead of updating to a patched version, let AI generate a workaround in your own code.

Agree with other commenters that the title is a bit confusing and should be renamed to something like "Is your Postgres workload read heavy or write heavy?"

But title aside, I found this post very useful for better understanding PG reads and writes (under the hood) and how to actually measure your workload

Curious if the tuning actions any different if you're using a non-vanilla storage engine like AWS Aurora or GCP AlloyDB or Neon?

Absolutely agree

I don't think Anthropic's $200/mo unlimited plan is evidence that free plans are a ticking time bomb

It's evidence that gen AI products is incompatible with unlimited plans. To be sustainable, it requires usage limits or usage based pricing

Sequence | London / NYC (Hybrid) or EMEA (Remote) | Full-time | sequencehq.com

Sequence is reinventing accounts receivable for modern businesses. We're one of the first a16z backed companies in Europe, building an AI first revenue automation platform that enables sales and finance teams to streamline the quoting, billing, invoicing and revenue workflow.

Our tech stack includes Kotlin, TypeScript, React, Postgres, and GCP. We're looking for engineers who enjoy being hands-on, shipping code, and working closely with customers to solve complex financial problems.

Open roles:

* Senior Product Engineer (Backend) | London / NYC (Hybrid) or EMEA (Remote) | £75k-110k + equity

* Product Manager | London / NYC (Hybrid)

Apply: https://www.sequencehq.com/about

I'm building an app for splitting bills with friends, mainly for holidays and roommates: https://quicksplit.io

It does one thing and it does it well. I built it with a friend and we released the iOS app last year. Currently building the Android app

Sequence | London (Hybrid), EMEA (Remote), NYC/SF (Onsite) | Full-time | sequencehq.com

Sequence is reinventing accounts receivable for modern businesses. We're building a flexible toolkit that helps B2B finance teams scale their revenue collection infrastructure with a world-class billing engine.

Backed by top investors including a16z and Salesforce Ventures, we're a team with decades of experience building category-defining marketplace, fintech, and enterprise software.

Our tech stack includes Kotlin, TypeScript, React, Postgres, and GCP. We're looking for engineers who enjoy being hands-on, shipping code, and working closely with customers to solve complex financial problems.

Open roles:

* Senior Product Engineer (Backend) | London (Hybrid) / EMEA (Remote) | £75k-110k + equity

* Senior Product Engineer (Frontend) | London (Hybrid) / EMEA (Remote) | £75k-110k + equity

* Product Manager | London / NYC / SF (Hybrid)

* Founding US BDR | NYC / SF (Onsite) | $100k OTE + equity

Apply: https://sequencehq.com/careers

Big +1 for Bunny.net - I moved my current company to Bunny and it's been excellent. Super fast (for our PoPs at least), reasonable pricing, love the image optimizer & edge rules (especially for solving header issues when embedding documents), has a Terraform provider, and I was able to set most of it up in a day. Was a night and day difference from GCP's Cloud CDN

I'm in a similar boat. Maybe Lobsters hasn't made headway across the Atlantic yet?

Good to see Skip on the home page! We were evaluating Skip just a couple weeks ago for a side project.

The issue we ran into is that we've already built a native iOS app with SwiftUI + a bit of UIKit. Integrating Skip with an existing app seemed like a significant task

Does that hold true in your experiences? Do you have any examples of small- or medium-sized existing apps that have migrated to Skip?

Does anyone have experience using dokku-postgres?

The GitHub readme is well documented but hard to know how that translates into the dev exp, like with scaling or upgrades and if its features are comparable to managed Postgres providers (I'd assume no but happy to be proven wrong!)

[0] https://github.com/dokku/dokku-postgres

That's a fair point - it can vary a lot between groups and it can certainly relate to life stage / situation. Students and roommates are a large portion of the market for bill splitting apps. In my experience those groups tend to care more about splitting things accurately / fairly, perhaps as they have a more limited income

Thanks!

Yep the split is the same but the conversions back to your base currency that can be a bit of a faff, especially if we're on holiday and doing these calculations multiple times per day

For example if I go to a restaurant with a large group and each item on the bill is in yen, we need to sum each person's total in yen and then convert each of their totals into GBP so we can each transfer what we owe (easy enough)

But depending on the exchange rates and rounding, those calculations can sometimes be off by a penny or two. Some people don't mind, others do

Yep! We have a bunch of ideas for features that we can start building now that we've launched the MVP.

That includes categories, activity timeline, receipt scanning, sign in with Apple, and a web app... to name a few :)

How far was the server from the Supabase instance?

I have a toy project with a server on Fly.io and the DB on Supabase within the same city

When sending API requests to the server, my end user latency was around 25ms-100ms (depending on the endpoint and how many DB calls it was doing)

Now with Supabase on Fly, that API latency is down to 17ms-70ms. But Supabase on Fly is still in alpha so it's not relevant for production yet

Like another commenter mentioned, managed container platforms like Cloud Run and Fargate don't have the same cold start limitations as serverless functions like Lambda.

Serverless functions spin up resources and run isolated code on-demand for every request as it's received. If you have no traffic, it scales down to zero and cold starts can be a concern. And paying for compute per-request like this can get very expensive at scale.

While Fargate is advertised as "serverless", I'd think of it as a fully container platform, it's closer to Heroku than it is serverless functions. You point it to an image, tell it how many instances you want to run, maybe set up some autoscaling rules (like to add more instances when CPU reaches X%), and then it'll run those instances 24/7. From what you've described, it sounds like a good fit.

Even on low-volume projects I've noticed reliability issues with Fly.io too. For example today we have requests timing out at the gateway (before hitting our services) while the status page indicates their current outage is only related to delayed app logs [0][1]. Haven't gone deep enough with the other new-ish PaaS providers (like Render and Railway) to get a feel for their limitations.

Unless you're certain you need K8s and have the in-house expertise to support it, I'd suggest going for Fargate or Cloud Run. Are there any requirements that you're concerned they won't meet?

- [0] https://status.flyio.net/incidents/q4tvqhthjmsy - [1] https://twitter.com/davydog187/status/1774874993618255977

> On 22nd February 2024 at 08:13:57 UTC the ID of the message threads table grew larger than the maximum size of the "int" data type. Writes to this table continued to work fine however there were four tables that had columns with references to the message_thread ID that were still using the int data type. Writes to these tables started failing immediately as the message_thread ID referenced was too large for that data type.

Not GP but I have the same set up and have only had a couple times where I received unexpected emails to an inbox. One was from GitHub profile scraping, can't remember what the other one was.

One of the surprisingly useful areas is with deliveries. Emails from delivery companies (especially from abroad) often don't specify what it is they're delivering or who it's from. But with this set up I can easily check that with the email address they're sending to, to know who it's from

Have you tried studying using the SAT Prep Black Book? It explains really well how to study for the test and why the test is designed the way it is

Update from 13 mins ago:

> If you are one of the impacted users who encountered data loss, you will have received an email from us with detailed information and next steps. If you have not received an email from us, you should not have experienced any data loss. We’ll be focusing on restoring lost data over the upcoming days and will post an update when this work is complete, followed by a detailed post-mortem.

From the README, pg_rman provides point in time recovery (PITR) and maintains a backup catalog:

> The goal of the pg_rman project is to provide a method for online backup and PITR that is as easy as pg_dump. Also, it maintains a backup catalog per database cluster. Users can maintain old backups including archive logs with one command.

GP did agree that there are some cases where this doesn't apply:

> Not all businesses are the same, and I'm sure there are some where you need to run auth in-house (like if you're selling on-prem software), but I'm mindful that most companies simply cannot run auth securely.

Reminds me that there's some interesting audio techniques in the new Avatar game, like ray traced audio:

https://youtu.be/LRI_qgVSwMY (skip to 13:16 for the audio section)

I think the preceding sentence has some important context.

They're specifically talking about businesses with high employee turnover, like call centers:

> Amazon said its thin client targets businesses with high employee turnover, such as call and payment processing centers, where AWS customers say it can be hard to recover expensive, work-issued devices.

A really bright LED light (190W) to help with winter and spending so much time indoors working.

Oddly, I don't see anything about pricing for Workers AI on the Workers pricing page[0] but their Workers AI blog post from Sept 2023[1] says the pricing is per 1k "neurons":

> Users will be able to choose from two ways to run Workers AI:

> Regular Twitch Neurons (RTN) - running wherever there's capacity at $0.01 / 1k neurons

> Fast Twitch Neurons (FTN) - running at nearest user location at $0.125 / 1k neurons

> Neurons are a way to measure AI output that always scales down to zero (if you get no usage, you will be charged for 0 neurons).

Here's the key detail:

> To give you a sense of what you can accomplish with a thousand neurons, you can: generate 130 LLM responses, 830 image classifications, or 1,250 embeddings.

[0] - https://developers.cloudflare.com/workers/platform/pricing

[1] - https://blog.cloudflare.com/workers-ai/

Why is Yudkowsky's HPMOR a "warcrime"?

Exciting to see how consistently Fresh is improving.

We're trying out Deno and Fresh on a recent project and it's been excellent so far.

Only downside is the limited plugin ecosystem (would love to see some Zod schema validation and OpenAPI plugins for API routes, for example)