The underlying problem is that it's difficult to make a principled decision. So, it's ripe for SEO mining and such fuzzy takes happen.
The meta point of the article, I agree with though. The line is moving and AI makes having multiple platforms with code specific to them faster. Getting them right is still difficult.
Ghostty has crashed nightly for me with ~10 terminals open across a few windows. So, I haven't been able to run it nor would I want to
embed it inside anything I daily drive.
I have opened right now about a dozen Ghostty windows and about 20 tabs in each window, i.e. more than 100 shell instances.
I have started in as many of them as I could, before becoming too bored, a "ls -lR" on a file system with many millions of files.
I could not see any problem, much less any crash. I have been using Ghostty for a few months, very intensively, all day long, and I have not seen any crash or other suspicious behavior.
If you have seen a crash, perhaps there was either some specific version of Ghosstty that had a bug, or, more likely, some weird interaction with some other software that you have, and which might be buggy, e.g. the GPU driver. (I am using an NVIDIA GPU.)
Just commenting to say I've had the same experience. Been using Ghostty since (I was aware of) its release and it has been buttery smooth ever since. I occasionally see people talking about crashing or other issues they've had with Ghostty but I've not seen anything of the sort.
> Building a team to operate based on your own personal preferences is selfish leadership... or even dictatorship.
Is there really anything wrong with that? Most managers manage their teams the way they're used to. Founders build their startups the way they're used to—based on their own experience and mistakes. Founders and managers don't adapt to the team's needs. Instead, they look for a team that will adapt to them.
With AI it feels writing software that is open is less attractive. It's hard to trust OSS made recently b/c you can tell if someone knows what they're doing and even spent any time on quality. Also, often times people don't reach for software others make (unless it's boring and old stuff, in which case this advice doesn't apply.)
IMO types are the main lever you can use other than procedural abstraction. I feel that Haskell gives you both in a way that marries them for maximum constraint-building. Constraints that prevent illogical or illegal programs are the bread and butter of reliable software.
I've been poking at running LLMs in the browser. It feels like we're definitely close (<1 year) to seeing real use cases there.
Ubiquity and coverage of devices is what will take longest. Largely dependent on how well we can shrink models with similar performance and how much we can accelerate mobile devices. This feels like it's but further (<3 years?)
Reading only the abstract: LLMs prefer output of their own generation over humans or even other models.
This is a very good reason to avoid using model-generated data to train future models. We'd be deepening this bias by continuing to do that, essentially forcing society to reshape their output using LLMs to increase engagement. This feels like a form of enshittification that doesn't just touch one product but all of society.
The meta point of the article, I agree with though. The line is moving and AI makes having multiple platforms with code specific to them faster. Getting them right is still difficult.
reply