A few weeks ago I had Fable build https://curry-combat-2.com: Super Curry Combat II Turbo: Legend Edition, a Street Fighter clone where every fighter is a Time Curry character.
It, was to me, surprisingly good for an AI slop project. Though Fable not suggesting the Doctor from Red October or Homolka from Congo in its initial roster is the biggest indication to me that we're far from general AI.
To show you your hyperbole: Allowing monkeys on typewriters.
LLMs are neither IDEs nor random.
I am very sceptical about iterative AI deployment too. People pretend the success threshold is vibing somethging that gets widely used, but its more than that. These one-shot solutions are not project maintenance. Answer yourself this one, could LLMs do what the linux kernel cummunity did over the same time span? This would be a good measure of success and if so, a strong argument to allow generated contributions.
I'm a manager at a large consumer website. My team and I have built a harness that uses headless Claude's (running Opus) to do ticket work, respond to and fix PR comments, and fix CI test failures. Our only interaction with code is writing specs in Jira tickets (which we primarily do via local Claudes) and adding PR comments to GitHub PRs.
The speed we can move at is astounding. We're going to finish our backlog next quarter. We're conservatively planning on launching 3x as many features next quarter.
Claude is far from perfect: it's made us reassess our coding standards since code is primarily for Claude now, not for humans. So much of what we did was to make code easier for the next dev, and that just doesn't matter anymore.
When is your website going to be complete? Are you sure those features are what the users need? What happens to the team after everything is done? What happens to the site after the team is gone?
This was always going to fail because the requirements were insane: solving SQL, XML and JSON generation should have never been part of the spec. There are hundreds, if not thousands of libraries for the those cases.
The developers should have never tried to boil the ocean: simple Kotlin or Scala style interpolation with overloading `+` for concatenation is the 80% use case here. Yes it doesn’t do everything. But it makes life a lot more ergonomic. And they rejected this solution out of hand! So instead of something useful but not perfect we get nothing.
also, there isn’t even a barrel let alone any sort of rifling? So how could we conceivably measure laptops in terms of barrel length if there’s no diameter or length?
Apple first-gen products are often half-assed. But they iterate and improve on a regular cadence. The current Apple Vision Pro is a dev kit, it will get better.
$68 billion is a lot, but it also only three Big Digs ($22 billion). If Chernobyl killed the Soviet economy it was because it was in already terrible shape.
It, was to me, surprisingly good for an AI slop project. Though Fable not suggesting the Doctor from Red October or Homolka from Congo in its initial roster is the biggest indication to me that we're far from general AI.