As I’m reading this, I’m thinking about how in 1980. It was imagined that everyone needed to learn how to program in BASIC or COBOL, and that the way computers would become ubiquitous would be that everybody would be writing program programs for them. That turned out to be a quaint and optimistic idea.
It seems like the pitch today is that every company that has a software-like need will be able to use AI to manifest that software into existence, or more generally, to manifest some kind of custom solution into existence. I don’t buy it. Coding the software has never been the true bottleneck, anyone who’s done a hackathon project knows that part can be done quickly. It’s the specifying and the maintenance that is the hard part.
To me, the only way this will actually bear the fruit it’s promising is if they can deliver essentially AGI in a box. A company will pay to rent some units of compute that they can speak to like a person and describe the needs, and it will do anything - solve any problem - a remote worker could do. IF this is delivered, indeed it does invalidate virtually all business models overnight, as whoever hits AGI will price this rental X%[1] below what it would cost to hire humans for similar work, breaking capitalism entirely.
[1] X = 80% below on day 1 as they’ll be so flush with VC cash, and they’d plan to raise the price later. Of course, society will collapse before then because of said breaking of capitalism itself.
Why are "tools" for local IO interesting and not just the only way to do it? I can't really imagine a server architecture that gets to read your local files and present them without a fat client of some kind.
What is the naive implementation you're comparing against? Ssh access to the client machine?
It's early days and we don't fully understand LLM behavior to the extent that we can assume questions like this about agent design are resolved. For instance, is an agent smarter with Claude Code's tools or `exec_command` like Codex? And does that remain true for each subsequent model release?
It’s a distinction that IMHO likely doesn’t make much difference, at least for the mostly automated/non-interactive coding agent use case. What matters more is how well the post-training on synthetic harness traces works.
Seems pretty promising. The language looks clean which I can't say for some of the alternatives.
The thing that I find so challenging about these types of systems is scaling up the richness of the playback.
Very quickly I find I need to integrate animations, lip sync, vfx, timed event triggers... For that you really need some kind of timeline. Delays don't cut it. So then these clean text driven systems are at best an early step in a large process or abandoned for a more integrated solution.
But I really do long for the ability to import simple narrative scripts like this even in a full production system.
One of these days I'll try to build the high production value system in a way that keeps both the full, in editor, narrative graph and the simple narrative script files alive and synced.
Nice. Something short of the full lift that might be useful is to add a reflection API along with the playback API. That way implementing import/export of .lor files to an engine format becomes a lot more appealing. Something to think about, anyway.
I do wish someone attempted a ground up AR OS where experiences are shared to nearby users by default. AR versions of airdrop are pervasive, that sort of thing.
For example, think of the opposite of Apple's People Awareness feature. Instead of an immersive experience fading away when a person comes near, the AR user's experience fades in as you approach.
I think it would be pretty magical, honestly. One of the wow moments the public never got to (because of adoption rates) is a shared AR experience. Really compelling stuff.
and god help you if those loops are pairing People and Products.
though now that I write that out... it would be really nice if you could optionally type iteration vars so they couldn't be used on other collections / as plain integers. I haven't seen any languages that do that though, aside from it being difficult to do by accident in proof-oriented languages.
You usually don't need an index that can't be used elsewhere. If you don't then you can abstract it away entirely and use an iterator or foreach features.
Depends on the language. Doing that is a huge pain in Go (until fairly recently, and it's still quite abnormal or closure-heavy), so the vast majority of code there does manual index-pairing instead of e.g. a zip iterator when going through two paired arrays.
Political parties, social networks, religions. these are all engineered systems. All of them including AI involve people. For starts nobody is going to do the massive amount of work to train a useless AI that is skeptical and cynical. Imaginination, Agreeability (which causes hallucinations) is a feature, not a bug. In humans and in LLMs.
For the same reason the things listed above are popular may be the reason why the most popular LLM ends up not being the best. People don't tend to buy good things, they very commonly buy the most shiny ones. An LLM that says "you're right" sure seems a lot more shiny than one that says "Mr. Jayd16, what you've just said is one of the most insanely idiotic things I have ever heard... Everyone in this room is now dumber for having listened to it. I award you no points, and may God have mercy on your soul"
If they fail, doesn't software and the giant companies that make it go back to owning the world?
reply