angusturner

In 2017 I worked tirelessly with my colleagues to implement and replicate the first transformer paper.

Yesterday I left Opus 4.8 to go do some architecture research, with GPU access.

It replicated and trained a credible baseline. It implemented some ideas I'd been thinking about, and wrote custom CUDA kernels for them. It read and summarised dozens of related papers.

It has since run dozens of experiments, with minimal supervision. When a model is unstable it kills it, documents why, fires off a new configuration.

The realisation that frontier labs are doing this at scale with unlimited GPU and token budgets.

It actually scares me a bit. The realisation that the next big breakthroughs will only have light human involvement.

The prospect of recursive self improvement feels more to real to me all of sudden

For small to medium projects, an LLM can write functional (if not well crafted) Rust.

Considering how easy this is now, why choose a heavier, slower and less typesafe language?

why read that, vs an actually well-written compiler though?

I thought exceptions tended to be made when its highly relevant to the technical topic at hand and also non controversial.

Outside a few weird online bubbles and pockets of the US, hardly anyone disputes the claim you are objecting to.

I feel this. I've had a few tasks now where in honest retrospect I find myself asking "did that really speed me up". Its a bit demoralising cause not only do you waste time, you have a worse mental model of the resulting code and feel less sense of ownership over the result.

Brainstorming, ideation and small, well defined tasks where I can quickly vet the solution : these feel like the sweet spot for current frontier model capabilities.

(Unless you are pumping out some sloppy React SPA that you don't care about anything except get it working as fast as possible - fine, get Claude code to one shot it)

I think most SWEs do have a good idea where I work.

They know that its a significant, but not revolutionary improvement.

If you supervise and manage your agents closely on well scoped (small) tasks they are pretty handy.

If you need a prototype and don't care about code quality or maintenance, they are great.

Anyone claiming 2x, 5x, 10x etc is absolutely kidding themselves for any non-trivial software.

Interesting... Maybe I need to investigate PayPal as an option here. Best case would be my bank eventually adds tap to pay natively

I recently made the shift to graphene from iOS and am mostly enjoying it.

The user profiles was slow to set up and not having shared filesystem between the user profiles creates friction. But I love that I can effectively sandbox my work apps, sandbox the Zuck apps etc, with different VPN profiles for each user.

Getting a burner google account (for gplay services) is a PITA if you are determined to get a clean slate from Googles tracking. Gplay is the only safe way to get certain apps at the moment, and make certain apps pass the device integrity checks.

I suspect one of the biggest barriers to mass adoption will be the fact that tap to pay doesn't work. IIUC apple/google pay are generally considered a privacy and security improvement over physical cards, since you don't give every merchant your actual card number.

Overall love the project and really nice to see such high quality open source software.

Yeah, worlds slowest and most in-efficient write-only database. And as soon as you need to interact with goods or services in the real world, then you still need trust anyway.

All these people harping on about: "Bro I just need to move my money without trusting anyone!, I just need a trust-less way to send currency bro!"

Trust is a good thing! Banks and financial middlemen aren't the devil. Look at how many TPS the visa network can do thanks to trust.

If it weren't for some minimum of social/institutional trust the whole of society would collapse anyway and your digital coins would finally converge to their true value (zero - or actually negative once you add in the externalities).

Fuck google for this. Awful decision. Guaranteed to be abused when Google or government despots decide that certain apps (or developers) aren't aligned with their interests.

Feeling very frustrated with the way the internet is going lately. This plus OSA + chat control. And compounded by the imperative for AI companies to keep hoovering up any and all data they can get their hands on, wiring it into "agentic" workflows and such.

You mean the guy that bans people from twitter for disagreeing with him? And has made a chatbot that spouts right-wing conspiracies in the name of being "anti-woke"?

As an Aussie, I was feeling somewhat consoled about the state of the US by the fact that the EU and UK still seem to have their heads screwed on.

OSA and chat control have made me seriously rethink that…

Has everyone lost their mind?

I wish people would stop parroting the view that LLMs are lossy compression.

There is kind of a vague sense in which this metaphor holds, but there is a much more interesting and rigorous fact about LLMs which is that they are also _lossless_ compression algorithms.

There are at least two senses in which this is true:

1. You can use an LLM to losslessly compress any piece of text at a cost that approaches the log-likelihood of that text under the model, using arithmetic coding. A sender and receiver both need a copy of the LLM weights.

2. You can use an LLM plus SGD (I.e the training code) as an lossless compression algorithm, where the communication cost is area under the training curve (and the model weights don’t count towards description length!) see: Jack Rae “compression for AGI”

There is an excellent talk by Jack Rae called “compression for AGI”, where he shows (what I believe to be) a little known connection between transformers and compression;

In one view, you can view LLMs as SOTA lossless compression algorithms, where the number of weights don’t count towards the description length. Sounds crazy but it’s true.

By the definition you have provided though, someone that has access to stable, safe or functional housing but then chooses to not to use it (eg opting to camp instead), is not homeless.

Edit: the word “lack” really is the key word. This implies no choice, right?

Agree the media is having a field day with this and a lot of people will draw bad conclusions about it being sentient etc.

But I think the thing that needs to be communicated effectively is that these these “agentic” systems could cause serious havoc if people give them too much control.

If an LLM decides to blackmail an engineer in service of some goal or preference that has arisen from its training data or instructions, and actually has the ability to follow through (bc people are stupid enough to cede control to these systems), that’s really bad news.

Saying “it’s just doing autocomplete!” totally misses the point.

One under appreciated / misunderstood aspect of these models is they use more compute than an equivalent sized autoregressive model.

It’s just that for N tokens, autoregressive model has to make N sequential steps.

Where diffusion does K x N, with the N being done in parallel. And for K << N.

This makes me wonder how well they will scale to many users, since batching requests would presumably saturate the accelerators much faster?

Although I guess it depends on the exact usage patterns.

Anyway, very cool demo nonetheless.

You assume that for small steps (I.e taking some noisy code and slightly denoising) you can make an independence assumption. (All tokens conditionally independent, given the current state).

Once you chain many steps you get a very flexible distribution that can model all the interdependencies.

A stats person could probably provide more nuance, although two interesting connection I’ve seen: There is some sense in which diffusion generalises autoregression, because you don’t have to pick an ordering when you factor the dependency graph.

(Or put otherwise, for some definitions of diffusion you can show autoregression to be a special case).

Hm suppose for argument sake that feeding a batch of data through some moderately large FF architectures takes on the order of 100ms (I realise this depends on a lot parameters - but this seems reasonable for many tasks / networks).

Now suppose instead you have an CTM that allocates 10ms on the standard FF axes, and then multiplies it out by 10 internal “ticks” / recurrent steps?

The exact numbers are contrived, but my point is : couldn’t we conceivably search over that second arch just as easily?

It just boils down to whether the inductive bias of building in some explicit time axis is actually worthwhile, right ?

I’m really glad to see people converging on this view because I feel a bit insane for not understanding all the hype.

Like, yeah, we need a standard way to connect LLMs with tools etc, but MCP in its current state is not a solution.

I think this is article is too generous about the use of stdio - I have found this extremely buggy so far, especially in the python sdk.

Also if you want to wrap any existing code that logs or prints to stdout then it causes heaps of ugly messages and warnings as it interferes with the comms between client and server.

I just want a way to integrate tools with Claude Desktop that doesn’t make a tonne of convoluted and weird design choices.

Yeah this is what I don’t really get about the stable coin hype.

In Australia for example, we are almost cashless now and most bank transfers are instant and cheap - no blockchain required.

Digital currency is a good idea - unclear what value the blockchain part actually provides?

I would actually say that scams and financial crimes seem to be the only concrete use case of the technology at the moment.

(Unless you count speculation/gambling).

The thing is though, once you regulate crypto then what’s the point? You are left with a highly inefficient/expensive and immutable database.

Bitcoin solves (or attempts to solve) for exchange in absence of trust and regulation. But this is a stupid thing to solve for, because without trust and regulation you can’t even have a functioning society.

The thing this really hits home for me is how Apple is totally asleep at the wheel.

Today I asked Siri “call the last person that texted me”, to try and respond to someone while driving.

Am I surprised it couldn’t do it? Not really at this point, but it is disappointing that there’s such a wide gulf between Siri and even the least capable LLMs.

I have been developing with MCP for a few weeks now, making some small python and javascript servers to integrate with Claude Desktop.

I am yet to see a use case that wouldn't be better served with an HTTP API. I understand the need to standardize some conventions around this, but at the heart of it, all "tool" use boils down to: 1. an API endpoint to expose capabilities / report the API schema 2. other endpoints ("tools") to expose functionality

Want state? ("resources") - put a database or some random in-memory data structure behind an API endpoint. Want "prompts"? This is just a special case of a tool.

Fundamentally (like most everyone else experimenting with this tech), I need an API that returns some text and maybe images. So why did I just lose two days trying to debug the Python MCP SDK, and the fact that its stdio transport can't send more than a few KB without crashing the server?

If only there was a stateless way to communicate data between a client and a server, that could easily recover from and handle errors...

I mean, it seems possible that with greater absolute wealth people might tolerate greater relative inequality but like … why run the experiment?

Anyway, that’s the path we are on - Piketty doesn’t hazard a guess about if/when society collapse, IIRC the point is just that things will only continue to get more unequal unless there’s a major crisis or some political intervention (e.g wealth or inheritance taxes - imagine trying to get the US to agree to that lol).

Also you can’t ignore relative inequality because wealth is strongly coupled to political power/influence.

Look at the head of DOGE - he spent $300m to get himself a top gov role, in charge of regulating his own business and restructuring US gov spending to suit his own whims.

In terms of percentages / orders of magnitude it’d be like your avg citizen spending $100 for a top gov role.

I personally think we should resist the temptation to separate out values from the discussion. Like, even if it was the case that some extremely high-level of inequality turned out to be "optimal" in some GDP-maximizing way, the question still ought to hinge on some notion of fairness or justice (in my view).

(Although, I guess if your metric is "citizen satisfaction" maybe that's not as terrible).

In any case, there are lots of case studies showing what happens at the extreme inequality end of the spectrum. As I mentioned in another comment, my favorites come from Piketty "Capital in the 21st Century" and Acemoglu & Robinson "Why Nations Fail".

In the latter, the case study on the rise and fall of Venice is particularly fascinating - huge economic growth due to inclusive economic institutions that promoted social mobility, followed by a downturn once the aristocracy moved to entrench their own interests at society's expense. This seems to be the central thesis of the book, although I'm only a few chapters in.

The parallels with modern US politics are pretty hard to ignore though.

When I was at uni I read Thomas Piketty's "Capital in the 21st Century", and parts of his newer "Capital and Ideology" (although I never quite got through that one).

The big takeaway for me was that wealth inequality never improves without some major catastrophe (war, revolution, plague etc). The proposed model is really intuitive and compelling (tldr; return on capital has historically always been higher than real growth, which guarantees indefinite concentration of wealth until there's a crisis).

Last year's Nobel Economics Prize winners, Acemoglu and Robinson, tell a similar story in "Why Nations Fail", which I am working through at the moment. Although in their case, they seem be suggesting a more causal link between erosion of political and economic institutions and the collapse of empires.

I wish these ideas were more broadly accessible and understood. The real risk of total societal collapse should transcend any partisan fighting about ideal tax rates, government inflation/unemployment targets etc. Everyone has a common interest in there not being a violent upheaval (arguably the rich most of all).

Last I checked the numbers, current wealth inequality seems about as bad as it was before the great depression. And its not enough to just say "well, absolute wealth is more important". As others have pointed out, its not stable to have such huge relative wealth disparities. And that's before you even consider corruption.

I have no idea what you are basing this claim on, although it seems in stark contrast to the evidence provided by other commenters in this thread.

AI is not a replacement for engineering expertise and deep understanding of the actual user problems that you are trying to solve.

Any 19yo at DOGE can have Cursor pump out shitty web portals without regard for the problem context or actual end user.

But that’s not what engineering is about and it’s depressing to see people celebrate and advocate for that kind of mediocrity.