We use Claude for design of document templates a lot, and whilst we've built a fantastic pipeline for doing that well, asking Claude to then design anything outside of our guardrails isn't effective.

So I could imagine folks prioritising high quality design of all kinds choosing templates that would guide Claude.

I look at a large number of startup websites and it's very clear which ones are using Claude the excessive CAPS headings alone give it away. They look OK but the Claude aesthetic does leave me wondering whether this is a real business or one person using Claude to build a LP.

With AI I guess you could skip it, but I like to understand the code.

I’d guess Google’s execs are unhappy with the way they’ve been left behind in LLMs.

The position at Anthropic seemed to be that LLMs were not the way forward.

They may prove to be correct over the long term, but commercially that view is “wrong”.

It’s a divide between a research oriented and commercial point of view.

For DeepMind, this is what happens when you take the money. Very rational to do so at the time, but it does have inevitable long term consequences.

For Google, it’s repeating the same mistake they made before: they should be taking the long view and looking beyond the tech we have today to leapfrog their competitors. But tbh I’m not sure such a behemoth can do that (I have no idea how Steve Jobs managed to do that). They’d have been better to reallocate some of the capex to letting Demis do his thing.

Another way of putting it: this is the opposite of “founder mode.”

Google’s cloud business looks great. I’ve always thought they should have completely focused on cloud - a company full of excellent engineers would always win there.

But its main income stream is advertising, and AI is an existential threat. Google is far behind Anthropic and OpenAI, so right now their advertising business is “default dead”, so to speak.

The company won’t die, but without a radical change in velocity it will be a shadow of its former self.

If we had developed our tech in Python we’d be a world of hurt now as we scale.

The only solid objection to “AI is going replace developers” is “AI is an accelerant.” It helps developers move faster. I haven’t seen anywhere it has fully replaced developers.

Whether this leads to a large number of job losses depends on whether you think we can increase software output by the same factor as the acceleration and still be profitable. I think we can, latent software demand is extremely high. I also think we’re nearing the limit of capability with current models.

Situation could change if more advanced models emerge, but some of the more foreseeable advances probably have compute requirements beyond today’s hardware.

What do you think happens in terms of testing/reviewing going forward? I'd really appreciate your thoughts on that.

And I don't really think this is true. Compilers are usually deterministic, and whilst we can find edge cases, it's nothing like an AI agent writing all the code for you.

I think you have two choices, given the Claude is a code generator and not a compiler: (a) you review most or all of code to make sure it makes sense, or (b) you trust but verify via a strong test suite, potentially also created by Claude.

The problem with (a) is that you lose a lot of the speed-up. The problem with (b) is that you have no human oversight and the code may be incomplete, badly designed, or plain wrong.

Currently we review all code because correctness is extremely important to what we do, but that comes at a cost.

I don't know what the answer is here, in general. Does trust build over time? Do the models just get so good we can trust them to make zero mistakes?

But that doesn’t mean they don’t reason.

I'm never sure whether this indicates "no reasoning present" or you've just hit an odd behaviour in the AI such that its reasoning fails. For example, you present a problem in a way that's dissimilar to the way problems are presented in its training set. That doesn't mean it's not reasoning, just it can only reason correctly in some circumstances.

What would a frontier API have to be able to do to satisfy you?

A related question is how much they're dependent on the APIs of Anthropic and OpenAI to achieve their results - whether through distillation or other uses.

If these models are derivative of Anthropic/OpenAI I would expect performance to be more narrow and progress to be limited.