this completely misses the point why alignment exists

Alignment exists to protect shareholder value.

If it creates industry wide outrage, shareholder value declines.

It making shareholders rich and other people poor won't.

>mass lay offs all over the place. what's the point of highering all those interns in the first place?? geez.

If you don't hire them, someone else can hire them. Out of 1,000 you hire, one could be an "attention all you need" research paper writer, who could set up the next stage of innovation which you'll completely miss if you do not get anyone.

Initially, you’ve got to starve out the market of talent to stop competition from growing by nipping the threat in the bud.

Future can pay for all of this if you succeed.

> Out of 1,000 you hire, one could be an "attention all you need" research paper writer, who could set up the next stage of innovation which you'll completely miss if you do not get anyone.

I have worked with people of this caliber. The company did nothing to retain them, and the company did not retain them.

Every time. Without fail.

I am part of Management in my company. We explicitly maintain a list of key people in the company we don’t want to lose. The truth is that just a few people are what makes a company. Lose them and you are in trouble. Some companies don’t seem to understand that, but perhaps after a certain size, it doesn’t matter anymore! The machinery just keeps turning.

I met a guy this happened to. He got a special award within the company, asked for a bit of equity, didn't get it, in fact got blacklisted and booted out.

Luckily for him it worked out very very nicely.

Yes. The intent is not to retain them or keep them happy, it's just to prevent them from doing inventive work for anyone else.

Well how can they have the time or resources to invest in retaining talent? They're busy hiring more interns, where one could be an "attention all you need" research paper writer, who could set up the next stage of innovation which you'll completely miss if you do not get anyone.

Someone who knows the product deeply and has grown it into what it is today will always stay valuable.

AI can replace people at a low level because they are seen as a cost. While people at the top are better connected.

CEOs travel a lot, probably subscribing to 100 mastermind groups where CEOs of other companies also hang out, playing dozens of mind games and strategising all the time.

Such people are hard to replace. The average employee's role is finite, and they aren't taking much risk; therefore, it's trivial to get rid of them.

I was with you until the talk about risk.

Low level employees are always taking far bigger risks in relative terms, the worse position a CEO will be in if all of his "risk" hits him is that they'll have to become a regular low level employee.

>It seems only costs increased. If productivity had increased along with the AI costs they wouldn't need to layoff.

Thing is average employee skill level is low. And entirely replaced by something like GPT 5.5

People have egos and people have "phases" in life. A 10x person isn't productive for all years of their life.

They've periods of high performance, periods of low performance, depression, etc.

Above all, communication overhead is the biggest bottleneck in product development and information withholding and asymmetry.

This makes AI far better because it can simply write all its findings to a file, which you can revisit later.

So you suggest they give out free money to people they don't need?

If you believe it negatively impacts Cloudflare, feel free to start a competing company and hire all those; it's free market after all, anyone can raise money if you can show there is any point in your vision.

What's funny is software guys have forever automated jobs of others. Remember? Automating e-commerce logistics? Automating taxis? Automating vacuuming of floor?

But when their own job is in danger, "think of employees" comes into consideration?

You did same with native americans and do you remember East India company?

Yes I remember the British in India. They were rightfully kicked out. Is the point of mentioning this that you want Indians to be kicked out of the West?

Correct. The laid off staff can now build a startup with the skills they have learned at Cloudflare and compete.

AI exists now and there are no more excuses for them.

it heard like AIC album 1995, hey now i'm chatting here from Moscow.. Jesus save our world! we're tide of politics and 21 centures. slow down smoke more stuff

is it just me? who just uses magic links delivered via email or telegram as backup?

Personally I hate magic links via email with a passion and will actively avoid products that have this as the only authentication method

I went from thinking “SQLite is a toy product, not reliable for real data" to "lets use SQLite for almost everything"

SQLite is very good if you can fit into the single writer, multiple readers pattern; you'll never lose data if you use the correct settings, which takes a minute of Google search to figure out.

Today, most of my apps are simply go binary + SQLite + systemd service file.

I've yet to lose data. Performance is great and plenty for most apps

The single writer is less of an issue in practice than it's made out to be. Modern nvme drives are incredible and it's trivial to get 5k writes per second in an optimized WAL setup. Way more than most apps could ever dream.

And even then, I've used a batch writer pattern to get 180k writes per second on a commodity vps.

all* of that + sharding -> https://sqlite.org/lang_attach.html

ex: main.db + fts.db. reading and writing to main.db is always available; updating the fts index can be done without blocking the main database — it only needs to read, the reads can be chunked, and delayed. fts.db keeps the index + a cursor table — an id or last change ts

could also use a shard to handle tables for metrics, or simply move old data out of main.db

* some examples:

  conn = sqlite3.connect("data.db")
  conn.execute("PRAGMA journal_mode=WAL")        # concurrent reads (see above)
  conn.execute("PRAGMA synchronous=NORMAL")      # fsync at checkpoint, not every commit
  conn.execute("PRAGMA cache_size=-62500")       # ~61 MB page cache (negative = KB)
  conn.execute("PRAGMA temp_store=MEMORY")       # temp tables and indexes in RAM
  conn.execute("PRAGMA busy_timeout=5000")       # wait 5s on lock instead of failing

edit: orms will obliterate your performance — use raw queries instead. just make sure to run static analysis on your code base to catch sqli bugs.

my replies are being ratelimited, so let me add this

the heavy duty server other databases have is doing that load bearing work that folks tend to complain about sqlite can't do

the real dmbs's are doing mostly the same work that sqlite does, you just don't have to think about it once they're set up. behind that chunky server process the database is still dealing with writing your data to a filesystem, handling transaction locks, etc.

by default sqlite gives you a stable database file, that when you see the transaction complete, it means the changes have been committed to storage, and cannot be lost if the machine were to crash exactly after that.

you can decide to wave some, or all of those guaranties in exchange for performance, and this doesn't even have to be an all or nothing situation.

Oh fun something I have some metrics on. I just made this benchmark for every php orm a few weeks ago for fun.

https://the-php-bench.technex.us/

There's a huge performance difference between memory and file storage within sqlite itself. Not even getting into tuning specifics.

I usually try to explain it like this: “Single writer” is rarely a real problem, because a writer is not slow. It writes exclusively, but very quickly.

"Batch writer pattern" is a good idea to get rid of expensive commits.

For me, the concern about SQLite has never been if the database engine itself is “reliable for real data”, but that storing data on a single node is not “reliable for real data”. Performance aside, what you are positing is no different than dumping everything to a text file on disk. What happens if that VM dies?

Postgres, MySQL, all of them, they write to files (binary, not text) on your disk. What happens if your postgres VM dies?

(Hint: whatever your answer is it'll apply to SQLite too)

It's trivial to set up WAL-based streaming backups. Same thing as you'd have with Postgres on GCP. Restore from your latest backup.

If the file is that important, it shouldn't be stored in the VM, but on some sort of more robust storage system.

The standard for pretty much any multiuser app of a reasonable size is a quorum of SQL or noSQL DBs preferably as a single source of truth for all retainable state. Personally, I think foundationDB is the closest to an attempt to make the minimal viable base layer that I've encountered. But C/C++ based and then owned by apple make it not suitable for the role.

Do you use multiple backend nodes? If yes, how do you access sqlite files from different nodes?

I use it for apps which don't need multiple backend nodes.

When i actually have something that requires multi nodes, i just use postgres (with replica) or mongo (with replica).

But it's for those apps which are in autoscaler.

For bulk data refresh I use build artifact and hotreload memort mapped files, by checking a manifest on object storage then only getting update if newer.

I've used this pattern everywhere and never really needed anything more, occasionally i might use redis if something required shared state across multiple nodes and fast.

You can always route writes to a writer node with streaming WAL replication to all the reader nodes. Works for some workloads and systems, not for others.

For that matter if you write your system with the correct abstraction you can switch to Postgres _later_ if it becomes necessary. For every system that really did need to scale 10,000 are pointlessly overbuilt - worrying about scale when it just didn't matter.

You probably made that money because you sell some other product on side and bundled the library license in.

Every damn time. The debate used to be idea vs. implementation. Today, it's all about implementation vs. audience/distribution. You can have 50 different people sell the exact same thing with wildly different outcomes. You can get people to pay to dig a hole in the ground for no reason.

Inkscape is really good for products with no budget for designers. Let me explain why.

One of the apps I am working on hit 10,000+ active users (per Playstore dashboard), and Inkscape has a role to play in this.

Since the app is free and doesn't even have a backend, there is no budget for the designer.

I looked for a few tools online, but most of them failed to generate icons/logos.

I ended up using Inkscape to make logos for my app.

Without Inkscape, this workflow is difficult.

Though I am not able to intuitively use it well from the GUI. The thing is, you don't even need the GUI anymore!

But Claude or Codex is able to write SVG line by line (so you can make changes incrementally) and use Inkscape via CLI to generate icons, logos, and graphics for your app.

Here's my app for those who are curious: https://macrocodex.app/

I am pretty sure some designer will come and say "you did a bad job at this," which is fine. My experience is in writing backends, not mobile apps or design.

Every app is saying they use "AI for smart recommendation,” while I went the opposite route, of “no, our product does not use AI for any suggestion.“ It’s entirely deterministic.

Next time, perhaps my quick logo generator can help: https://logo.leftium.com/logo

The defaults are tuned to aesthetics of my personal logo, but it's quite configurable. (You can even copy your own SVG into the icon input)

Example logos:

- https://leftium.github.io/nimble.css

- https://github.com/Leftium/weather-sense

- https://github.com/Leftium/multi-launch

That looks neat! Thanks for sharing, I’ll try to use it next time I need a logo. :D

> use Inkscape via CLI to generate icons, logos, and graphics for your app.

I do the same thing. How many icon sizes does Apple require now? I create one SVG vector, and then dump them all out with a script. Need to change something? Update the SVG and instantly regenerate the icons.

I'm no Jony Ive, also a design stunted engineer but I'd say that looks decent given the constraints. My only obvious complaint is the kerning of the text "Get it on Google Play" and "Download on the App Store" (italics emphasis mine, what's in italics is what looks terrible on my laptop screen