pitched

Is this lazy version of RLM where we use subagents and get them to output markdown documents? With one root agent coordinating, it can create a reasonable prompt and provide paths to the markdown documents for the next agent. This keeps context really low because each subagent is filtering the context needed for the next by bouncing it through that document.

Proper RLM looks like you’re allowing agents to directly modify their own context though, like closing browser tabs they don’t need anymore. I haven’t seen anyone actually doing this though.

These things are still very far from human-level intelligence. Maybe a touch beyond a golden retriever in processing power. Ratcheting the intelligence up a notch from there is expensive. It’s a lot cheaper to simplify the problem it’s solving instead.

An approach I like to help solving this is antagonistic or review agents. The first agent decides that eye glows turn NPCs into enemies, the second agent is fully dedicated to deciding if that is valid. If the review fails, it leaves notes and the original agent tries again.

I think it’s more likely we’ll raise prices until demand lowers enough to match supply.

There isn’t much need to extract isotopes if you have an actual working reactor, is there? Just use that as-is to cause whatever damage.

In the Innovation Adoption Curve, we are absolutely beyond the Early Adopters phase and possibly the Early Majority. The growth rates necessarily have to start trending down because there’s no one left to sell to.

TFA isn’t about consumer usage, it is about training the next generation models. An interesting thought I heard recently is that a SOTA model has about the same parameters as the number of synapses in a Golden Retriever’s brain that are not dedicated to biological processes like breathing. Most of that should be wrapped in double quotes, don’t take it literally! But that number is about 100x lower than the same synapses in a human brain.

If the next order of magnitude costs 40B, I wonder if it’s even possible to get to the one after.

This is the exact kind of cause that triggers the social media algorithms. Part nostalgia, part underdog battle, and no problematic elements.

Sony removing the disc drive from PS6 will save them a lot of money if Xbox also removes it. I wonder if they’re back channel colluding on that right now?

It sounds like your opinion is that status quo is ok when: there is a spectator mode, it’s on a mobile device, there is multiplayer.

I still don’t quite understand in which cases the status quo is not ok though.

Sort of like how Diary Queen aren’t allowed to call their desserts “Ice Cream” because there isn’t enough dairy, we need to force retailers to start calling them game-licenses? This feels like a whole lot of a big fuss over something like that. I really feel like there must be something deeper going on.

Back in the old days, we would be pooling resources like that to host a server for the clan. And I wonder if the reason those games are less popular has more to do with marketers being good at their job than a change in the base gameplay.

There are local clubs available but everyone wants to play on the nice FIFA field, which is the one that has to be rented.

For single-player games, those are definitely media. For multiplayer games, that feels a lot closer to a sport.

We rent time at a football field. We buy tickets to watch a single match. There are parallels here to not owning video games. I don’t really understand why one is so heinous.

Let’s say there’s a new rule implemented by the NBA that no one likes (similar to a fear of live service games changing). How is that resolved there and why can’t that solution work for video games?

I think a big thing we’re currently missing here is something like a community field or park. Why are there no open-source, community-run Diablo projects for example? If no one cares enough to do that, maybe this isn’t so big of an issue.

That upfront convenience here leads to a long tail of job security when it inevitably goes spaghetti. Win-win!

Or squirgle jail on a transparent background instead of a grey one? Why would cursor collision with an image stop working because a new input method?

The only reason to run MacOS over Linux is the hardware. Arm MacBooks are unreasonably good but don’t support Linux (Asahi is still a bit WIP).

They sell hardware, not software, so the state of things makes sense. It is so disappointing though.

No restricted for me, either on or off VPN. This might actually be an account issue on your side? Maybe a work account?

Oh wow, I really mixed that up! Thanks! That really changes this lol

The difference now is that the learning is optional (more often but not always) to getting the task done.

When gaining mastery is not a requirement to doing novice-level work, many fewer people will get there. It takes more dedication than it did before.

Math is a language to explain systems. Teaching someone that force varies linearly to mass is a helpful first pass. It isn’t exactly linear but is not exponential at all.

Gaining expertise is always the hard part and our new LLM overlords are making that much harder. So the simple “pure” functions as a teaching aid have never been more important.

End users have never cared about how the sausage is made though.

There are a set of these “false beliefs” people can hold that are clearly untrue but are beneficial to them in other ways. This one about “hard work leads to success” is a very powerful one. Lean the hell into it if you ever come across an opportunity to!

From what I gather, this is Anthropic-only because of disagreement there. OpenAI is not subject to apply the policy and the US has no power to limit foreign companies like Deepseek and GLM.

Anyone willing to accept an efficiency loss is already using an interpreted language and doesn’t have this class of problems.

China is benefiting a lot from releasing the models open-source and those benefits immediately end if they start doing closed-source releases. It would be very short-sighted

Without going that far, send a few people a couple hours North instead and serve international customers from Canadian data centers. As far as I understand, it is only blocked in the US, right?

I’m on your side in that I would never take a contract to actually do this, but…

If we swapped out the IAM backend for something extremely simple like just private keys (one per allowed service or JWT-style list all services in the key), then we could have something that looks/feels pretty similar. With a 2$ token spend.

Not at all the same but it would look/feel pretty close.

you might be able to make a version of some small subset of aws services that runs works ok

lol, you’ve got that goblet of koolaid with me! Equal parts horrifying and interesting that it might not be impossible

The good ones all seem to be pointing in the direction of Django. Which, on its own, says a lot about how likely people will care about vibe-coded anything, whether pooled or not.

lol but at least in comes in a nice cup then