fizx

Eh, I've been mountain biking seriously for 15 years, and there's kids in their teens running circles around me because their free time, risk tolerance, and brain development just means they win.

At some point 10+ years of adult development will get you pretty good at something, but being competitive beyond say a high-school varsity level with a late start is very difficult.

I think Sam is an exec at the company.

Presumably it took more than one attempt, or there's some humanpower in the budget.

Is this sort of re-inventing Graph RAG from another angle, or does it feel novel?

What policy would you use?

Firecracker doesn't reclaim RAM well, so they put this limit in to make sure you don't suffer from the eventual memory bloat problems.

A2A solves the problem of independently developed agents talking to each other. The bigger problem is "how do i trust your agent works?", which would be solved by having a believable system of Agent Evals, so I could test that your Agent actually does what it says on the box. Otherwise an Agent Directory and automatic dispatch of agents is close to useless.

The ratio of proofreading to grandiosity is impressive.

Performance enhancements are what allow you to train a bigger model.

Plenty of everyone doing it wrong, but the most WTF of all the WTFs is the backup storage.

Put your backups in S3 *versioned* storage on a different AWS account from your primary, and set some reasonable JSON lifecycle rule:

     "NoncurrentVersionExpiration": {
        "NoncurrentDays": 30,
        "NewerNoncurrentVersions": 3
     }

That way when someone screws up and your AWS account gets owned, or your databases get deleted by an agent, it doesn't have enough access to delete your backups, and by default, even if you have backups that you want to intentionally delete, you have 30 days to change your mind.

Sure, but you then can't access the computronium, because communication is so bad.

Also, I wonder if good old-fashioned computing is interesting at all to a civilization that's had access to advanced AI and quantum computing for a while.

Like we haven't really figured out how to get an ML model to run on a quantum computer, or how to build a quantum-native computer (i.e. surface of a black hole, or some other way that doesn't rely on our current sense of quantum error correction), but I don't know of any physical laws that preclude it.

I'd bet if aliens invaded our galaxy, they'd go for the super black holes in the center, or some other resource beyond our use and understanding, not this random water planet on the edge.

I mean you can use sqlite as an index and then rebuild all of Lucene on top of it. It's non-trivial to build search quality on top of actual search libraries too.

O'Reilly's "Relevant Search" isn't the worst here, but you'll be porting/writing a bit yourself.

I enjoy Japanese joinery, but for some reason the housing market doesn't.

hey, they have 9 8's

Does anyone know how much the cost-per-token is trending down year-over-year for models of similar quality? Seems like whether this idea works really depends on that curve.

What a fever dream!

Add reinforcement learning to figure out which skills are actually useful, and you're really cooking.

This isn't about income.

- The hard part of a compiler is the parser. Find an existing parser, then emit using that language's web templates.

- Your DB query is better in an inverted index.

- Consider your data locality carefully.

The goal of post-modern society is to create wealth without additional energy costs.

If I write software that schedules health care more efficiently, I've created value and therefore wealth. If I make a video game that's more fun than the previous generation of games, likewise.

I tend to look at consciousness as a spectrum. And when we reduce it to a binary (is it conscious?), we're actually asking whether it meets some minimum threshold of whatever the smallest creature you have empathy for is.

So yeah, Claude Code is more conscious than raw GPT. And both probably less than my dog.

Quickjs will do this.

Definitely, iff you have to provide the context with every task. If agent memory worked better and across your whole team, then providing context might be much easier

I read the article, and I'm struggling to see what ideas it brings beyond CodeAct (tool use is python) or the "task" tool in Claude code (spinning off sub-agents to preserve context).

Cursor when not in "MAX" mode does its own silent context pruning in the background.

I don't see how this very contrived example pipelines:

    client.getAll({userIds}).map((user) => user.updatedAt == new Date().toLocaleString() ? client.photosFor(user.id) : {})

or without the conditional,

    client.getAll({userIds}).map((user) => client.photos({userId: user.id, since: new Date(user.updatedAt).toLocaleString()})

Like it has to call toLocaleString on the server, no?

Also, your function needs to be very careful on closures. Date.toLocaleString and many other js functions will be different on client and server, which will also cause silent corruption.

You can swap bikes in the middle of the race if you have a mechanical issue. There was one famous time where someone climbed impossibly fast, had a mechanical at the top of the mountain, then finished the race on a different bike, leading us to forever wonder.

Don't be confused! That's not how they made the decision; it's how they're selling it.

Annoying, yes. Tractable, absolutely!

The "time dialation" is real. I mostly manage these days, yet my fun projects progress faster than they ever have, because I can prompt in the 2 minutes between meetings, and come back to significant progress.