mvkel

This should be a wake up call. A product's value is not a function of its code elegance. Nobody who matters notices the code, or cares. This is hugely inspiring to the most lazy+clever engineers, because it frees up so many thinking calories. Instead of trying to perfectly choreograph every bit of architecture to optimize for 1M concurrent users, you can spend 0.1 of the time and get things out the door, where you learn if spending even a minute of your time was worth it. Even better, when you realize tech debt is something that never needs to be paid down, you can focus all your energy on evolving your thinking patterns, not being bogged down in refactoring things that you've spiritually moved on from. An engineer's time is so precious; it needs to be spent thinking, not coding.

Related, I have a personal project that has several modules that run in a sequential pipeline. As each module is activated, an animated portrait avatar of a Starcraft II character appears, so at a glance, I can see which module is running at any one time. If I were hand-coding the app, this little 'feature' never would have existed, because it would have felt like a waste of dev calories. Now it's there because it's fun. Coding with AI will mark the return of whimsy in software.

> The Hatchet TUI is not well supported on mobile devices.

Isn't everyone else remoting into a Claude instance on their phones?

"Built to Sell" is a better book, starting with the premise that your company needs to be positioned to be bought, not sold. If you are reaching out to someone to buy you, you are effectively accepting an 80% price cut.

This was one of the biggest hurdles I had to overcome when I was a wee lad combing through "Professional PHP Programming." All of the examples it gave were foo/bar, and I couldn't make the intellectual leap to understand what the real world use cases would be.

It wasn't until I tried building something (mad libs) that things "clicked"

Seems short-sighted to commit to not running ads exactly like OpenAI plans to. You win in the court of public opinion for a few months, then look like a hypocrite when you're inevitably forced to run ads because your investors demand it. Sort of like how safetyism was convenient marketing until it became clear that it was revenue repellant and they quietly walked it away.

> We are certainly closer now to being able to prototype and go to market faster with a product

Absolutely. But this begs the question that businesses want to also sign up to maintain whatever product they've built, on top of their core business.

"Service" is the word that people seem to be forgetting in SaaS. If you roll you own, all you have is software.

It's this generation's "build vs buy." I imagine it will play out the same way, like a revolving door. Customers churn because they can "build it themselves," then a year later when they're sick of maintaining a mess of code for some internal system instead of delivering value to their own customers, they come back. A blip.

Why anyone wouldn't want to be using the SOTA model at all times baffles me.

Going dumb/cheap just ends up costing more, in the short and long term.

So here we have OpenSSL, coded by humans, universally adopted by the Internet, universally deemed to be terrible code.

More evidence that "coding elegance" is irrelevant to a product's success, which bodes well for AI generated code.

This seems like a cheap trick to hook someone into a blog post (ironically, Gwern seems to disregard this almost universally).

If I were reading a book and each chapter started with such a "hook," it'd start to feel like a LinkedIn post.

Chapter 1: I didn't know what it felt like to be alive until I was dead...

Chapter 2: Death was nothing compared to what came next: judgment.

Chapter 3: I thought I knew what judgment was until...

More engineers aren't the solution. Less engineers are.

https://en.wikipedia.org/wiki/Brooks%27s_law

It's wild how the layers of abstraction just keep expanding

> tests, documentation, CI, issue trackers

Exactly. In many engineering camps, it's not unreasonable to say that almost all of this has no benefit to the end-user, even indirectly.

https://agilemanifesto.org/

Just my experience in vibe coding apps from gpt-3.5 onwards (mostly NextJS or Node). In gpt-3.5, I had to really hand-hold it, getting it to write one function at a time, then a separate task to glue the functions together.

Now, it can build almost all of an app from a single prompt, but will start to rewrite utility functions, or modules, forgetting that they already exist. Some of this is still solvable with clever prompting, but if you're just attacking it without thinking, ~30,000 LOC seems to be the app 'size' that it will start to exhibit those behaviors.

There's only one difficulty setting

Style = syntax, taste, architecture choices, etc. Things you would see on a 15-year-old Java app.

i.e. not a greenfield project.

> Other people are just less picky than I am

I think this is part of it.

When coding style has been established among a team, or within an app, there are a lot of extra hoops to jump through, just to get it to look The Right Way, with no detectable benefit to the user.

If you put those choices aside and simply say: does it accomplish the goal per the spec (and is safe and scalable[0]), then you can get away with a lot more without the end user ever having a clue.

Sure, there's the argument for maintainability, and vibe coded monoliths tend to collapse in on themselves at ~30,000 LOC. But it used to be 2,000 LOC just a couple of years ago. Temporary problem.

[0]insisting that something be scalable isn't even necessary imo

Related: does anyone know of a way to delete the original videos on trimmed ones? Apparently all edits to videos are non-destructive, so the 10-minute video that I trimmed to 2 seconds still takes up 8gb.

Incredible. I have such a hacked-together system to get my iCloud Photos backed up to an external drive while not filling my main laptop drive. This would be even better.

Lemay's appointment was widely celebrated, but he'd been at apple since 1999 and never got the gig. My guess is that there are valid reasons for that that may not be design-related.

Seems very clear now that we are going to see touch screen MacBooks. Which is a very silly idea. But explains why the UI "snaps" like an iPad, and everything is designed for touch.

Demanding every intimate personal detail of a human whose paycheck you happen to underwrite feels a little ... inhumane.

I think the difference (at least with macOS) is the fundamental things that -were- working don't suddenly break*

*macOS26 excepted

Twitter started as an internal service for existing Odeo employees, so an unfair comparison if you're talking about cold, starting a social network. Twitter didn't start cold.

These are great ideas!

I am saying for it to succeed as multi-user, it needs to be useful first for a single user.

This app (like any consumer social app) needs to first solve the cold start problem: make it useful for a single user, layer the social on top.

Instagram had photo filters; Strava had activity stats. What could this have?

One thing I think it's still always valid to complain about is when your time is stolen. It's precious, and you can never get it back. If the unpaid experience slows you down unnecessarily, I think it's okay to be frustrated.

The truly infuriating experiences are like cable tv, where the paid experience is terrible, too.

Google does know how to do the paid version well. YouTube Premium is a great example. Massive music library and no ads, ever. It's astounding how much better the experience is.

How's their spam filtering?