mrkurt

Some of the best content marketing I've ever read. More please.

We don't do any model training, and only use existing open source or hosted models. Code gets sent to those providers in context windows. They all promise not to train on it, so far.

It's safe to say that if either Chris or I believed this to be anti worker, we wouldn't be working on it. He's spent the last 10+ years working on Phoenix specifically to improve the lives of the people doing the work.

My experience with software development is maybe different than yours. There's a massive amount of not-yet-built software that can improve peoples' lives, even in teeny tiny ways. Like 99.999% of what should exist, doesn't.

Building things faster with LLMs makes me more capable. It (so far) has not taken work away from the people I work with. It has made them more capable. We can all build better tools, and faster than we did 12 months ago.

Automation is disruptive to peoples' lives. I get that. It decreases the value of some hard earned skills. Developer automation, in my life at least, has also increased the value of other peoples' skills. I don't believe it's anti worker to build more tools for builders.

If we try really hard, I think we can make an exhaustive list of what viral fads on the internet are not. You made a small start.

none of these ephemeral fads are any indication of quality, longevity, legitimacy, interest, substance, endurance, prestige, relevance, credibility, allure, staying-power, refinement, or depth.

It's not false. He's talking about people smarter than him (at writing and shipping infrastructure code).

Thomas is the smartest at other things.

PLEASE write your response. We'll publish it on the Fly.io blog. Unedited. If you want.

I would expect this problem to get solved. It's tricky to detect new sqlites, but not impossible.

In the meantime, it's pretty straightforward to use as a library.

Wow that's a wildly cynical interpretation of what someone is saying. Maybe it's right, but I think it's equally likely that people are saying that to give you the right context.

If they're saying it to you, why wouldn't you assume they understand and trust what they came up with?

Do you need people to start with "I understand and believe and trust what I'm about to show you ..."?

I really, really like New Orleans and this fits my 2 year experience so far.

2025 makes it hard to be far from SF for me again because there's so much interesting stuff getting built. FOMO follows me around though.

Here's what you should assess:

> Great team, great technology, great PMF and good progress on revenue targets.

Good progress on revenue targets from _your_ perspective sounds like not great growth, but communicated softly.

> The argument seems to be that they've realized the only way to achieve the next round of funding is to be "AI-first". There is no product roadmap for what this looks like, or what features might be involved, or why we'd want to do it from a product point of view. Instead the reason is that this is the only way to attract a big series C round.

This is not a thing a company does when they have good enough growth. Adequate revenue growth when you need a series C must be very good growth. Adequate growth when you'e profitable is a very different thing.

> Instead of working on useful, in-demand product features, it feels like we're spending a lot of time looking at a distant future that we'll struggle to reach if we take our eye off of the ball. Is this normal?

Yes, very normal. If I'm you, I'm wondering if there's really product/market fit. Useful, in demand product features almost don't matter. Building more does not make more money happen.

But also, the AI stuff is disruptive. In a way that the internet was. I think what you should probably judge most strongly here is if there is ANYONE with influence that has a take on how that looks for your company. That take probably includes "how this makes a lot of what we do irrelevant".

Absent that, I'd just write the company off. If you like the job great. If you don't like the work enough to be happy while the company atrophies, I think you're going to have a rough time.

For what it's worth, I really appreciate your post.

No problem at all!

I would be curious how the Attic server would have gone with a Tigris bucket and local caching. Not sure how hard that is to pull off, but Tigris should be substantially cheaper than our NVMes and if you don't really NEED the io performance you're not getting anything for that money. Which is a long winded way of saying "we aren't great at block storage for anything but OLTP workloads and caches".

One thing we've struggled to communicate is how _cheap_ autosuspend/autostop make things. If that Machine is alive for 8 hours per day you're probably below $8/mo for that config. And it's so fast that it's viable for it to start/stop 45 times per day.

It's kind of hard to make the thing stay alive with health checks, unless you're meaning external ones?

We are suboptimal for things that make more sense as a bunch of containers on one host.

I do, in fact, instruct LLMs to make spelling and grammar mistakes when I have them reply to cold emails.

This is near and dear to me, because I want people to run stuff like homelabs and side projects.

What part of the cost gets out of hand? Having to have a Machine for every process? Do you remember what napkin math pricing you were working with?

This was a demo they did in a video but it's not available to anyone who actually owns one of these. I believe the EV Mercedes G Wagon does have it as a feature though.

Most new cars in Europe are using LED arrays that shut off specific zones based on oncoming traffic. Like a hyper specific "auto high beams" feature.

Talks at meetups for the first 100.

Hacker News for the next 10,000.

Target audience was devs.

I'd piss off a lot of investors I know by saying this – nerfing the Adobe and Figma acquisition seems excellent for consumer good.

3 years ago that would have just been rubber stamped.

Their work on non-tech stuff has probably been even more effective, too.

Laptop SSDs are _shockingly_ fast, and getting equivalent speed from something in a datacenter (where you'll want at least two disks) is pretty expensive. It's so annoying.

It is a little squicky. Kind of love the idea, though, even if I'd be worried about using rando compute someone installed an agent on.

Sometimes vast.ai is running GPUs on Fly.io that people with YC credits have spun up and added to their marketplace. Those would have been fast though.

They run on literally anything someone installs their agent on.

Believe it or not, I know exactly when this happened and why. We had a big disk array failure and (a) our UI did not know how to recover and (b) disk snapshots timed out before many people knew their disks were gone.

This is one of those instances where an unmanaged single node Postgres sucks to use.

A small db isn't free to _operate_. The actual compute and storage costs are pretty low, but keeping the thing running is harder than you'd think.

Keeping it running with no users is easy. Keeping it running when someone ships an app that blows the DB up is hard.

You can run a Postgres on Fly.io full time for just slight more than $2.

It's not a great business though. The compromise you're making is that we don't manage it. It'll probably stay running, it's likely to be worry free (especially if you barely use it), but if anything goes wrong it's up to you.

You could also spin up a VM for cheap somewhere, install Postgres, and just let it be.

Managed databases are expensive to run – worry free costs quite a bit of money to ensure. Maybe you don't need that for a hobby project?

This account was not banned in error. Lukas, if you’d like me to explain publicly, I'm happy to do so.

This sounds like a bad job and you should find another.

At some point you were happy with $200k and 2%. Now you aren't. I don't think there's a bump that they could give you at this point that would make you want to tolerate the shit that's bad about the job.

Note that this is a bad job. Startups are more prone to creating bad jobs.

The math for employee #1 at a startup is almost never ideal if you're being completely rational about expected value. That is, I think, a separate problem than you're dealing with.

Not that people with VC need defending, but:

Sure, if you magic up a startup with VC funds you suddenly have it easier than a small, bootstrapped business.

Startups almost never start with a round of VC though. There are almost always months or years of the same experience as a bootstrapped business (ie: extreme uncertainty, no money to pay yourself, etc).

Most startups don't manage to raise VC, and most startups that raise VC fail with no acquihire.

Postgres is pretty heavy BECAUSE it's a network database. What would make you want embedded PG instead of sqlite?

This is actually what Fly machines do under the covers. When you use it as a PaaS, the client CLI sets all this up for you, and then our proxy starts/stops things as needed.

Fly machines allocate all the computation resources when a TCP connection is established. I feel like this is the same?