Inufu

In that case the CLI only needs to import the plugin that defined that sub command, not all plugins?

Author here.

The standard way to do this is Reinforcement Learning: we do not teach the model how to do the task, we let it discover the _how_ for itself and only grade it based on how well it did, then reinforce the attempts where it did well. This way the model can learn wildly superhuman performance, e.g. it's what we used to train AlphaGo and AlphaZero.

Well, growth has been on this exponential already for 5+ years (for the METR eval), and we are at the point where models are very close to matching human expert capabilities in many domains - only one or two more years of growth would put us well beyond that point.

Personally I think we'll see way more growth than that, but to see profound impacts on our economy you only need to believe the much more conservative assumption of a little extra growth along the same trend.

Author here.

The argument is not that it will keep growing exponentially forever (obviously that is physically impossible), rather that:

- given a sustained history of growth along a very predictable trajectory, the highest likelihood short term scenario is continued growth along the same trajectory. Sample a random point on an s-curve and look slightly to the right, what’s the most common direction the curve continues?

- exponential progress is very hard to visualize and see, it may appear to hardly make any progress while far away from human capabilities, then move from just below to far above human very quickly

Yeah exactly!

It’s likely that it will slow down at some point, but the highest likelihood scenario for the near future is that scaling will continue.

> You don’t have to randomize the first part of your object keys to ensure they get spread around and avoid hotspots.

Sorry, this is absolutely still the case if you want to scale throughput beyond the few thousand IOPS a single shard can serve. S3 will automatically reshard your key space, but if your keys are sequential (eg leading timestamp) all your writes will still hit the same shard.

Source: direct conversations with AWS teams.

It means you can for example no longer do things like get multiple disjoint references into the same buffer for parallel reads/writes of independent chunks.

Or well you can, using unsafe, Arc and Mutex - but at that point the safety guarantees aren’t much better than what I get in well designed C++.

Don’t get me wrong, I still much prefer Rust, but I wish async and references worked together better.

Source: I recently wrote a high-throughput RPC library in Rust (saturating > 100 Gbit NICs)

Requiring ownership transfer gives up on one of the main selling points of Rust, being able to verify reference lifetime and safety at compile time. If we have to give up on references then a lot of Rusts complexity no longer buys us anything.

You can use Claude Code with your own API key if you want to use more tokens than included in the Pro / Max plans.

AlphaZero did not run game logic on TPUs (neither chess nor other games), implementing it in C++ is more than fast enough and much simpler.

TPUs were used for neural network inference and training, but game logic as well as MCTS was on the CPU using C++.

JAX is awesome though, I use it for all my neural network stuff!

This is not a fake imitation, it's steaming results live as they are being generated by the model.

The 12 bit part refers to the colours being specified using three hex letters.

Yes, parameters are usually stored as float32, activations as bfloat16 or float16.

On many natural language tasks there can be significant overlap, making it difficult to judge performance. That's why I like more complex code generation tasks such the dataset we used for AlphaCode.