The thing that stands out on that animated graph is that the generated code far outpaces the other metrics. In the current agent driven development hypepocalypse that seems about right - but I would expect it to lag rather than lead.

*edit* - seems inline with what the author is saying :)

> The data says: machines are improving at a constant rate. Humans are freaking out about it at an accelerating rate that accelerates its own acceleration.

> I don't personally think there is magic in building a Docker container. Call me old-fashioned.

I still vividly remember setting up gcc in a docker container to cross compile custom firmware for my cannon camera and thinking about the amount of pain my local system would have been in if I had to do all the toolchain work in my host OS. Don't know if it felt like magic, but it sure didn't hurt like the alternative!

For sure. Docker is rad (sorry Docker!)... all I'm saying is that I am not proud of the fact that I can do it and I don't think it moves the awesome needle - but it's still hard to get right and a pain in the ass. It's just an example of something I appreciate that I can automate now.

Oh, the joy that awaits you when you come back home to discover how the gardener interpreted "please trim the hedge by the gate a little".

> Using anything other than the frontier models is actively harmful

If that is true, why should one invest in learning now rather than waiting for 8 months to learn whatever is the frontier model then?

So that you can be using the current frontier model for the next 8 months instead of twiddling your thumbs waiting for the next one to come out?

I think you (and others) might be misunderstanding his statement a bit. He's not saying that using an old model is harmful in the sense that it outputs bad code -- he's saying it's harmful because some of the lessons you learn will be out of date and not apply to the latest models.

So yes, if you use current frontier models, you'll need to recalibrate and unlearn a few things when the next generation comes out. But in the meantime, you will have gotten 8 months (or however long it takes) of value out of the current generation.

You also don't have to throw away everything you've learnt in those 8 months, there's some things that you'll subtly pickup that you can carry over into the next generation as well.

Also a lot of what you learn is how to work around limitations of today's models and agent frameworks. That will all change, and I imagine things like skills and subagents will just be an internal detail that you don't need to know about.

snarky answer: so you can be that 'AI guy' at your office that everyone avoids in the snackroom

Because you might want to use LLMs now. If not, it's definitely better to not chase the hype - ignore the whole shebang.

But if you do want to use LLMs for coding now, not using the best models just doesn't make sense.

It's not like you need to take a course. The frontier models are the best, just using them and their harnesses and figuring out what works for your use case is the 'investing in learning'.

There's not that much learning involved. Modern SOTA models are much more intelligent than what they used to be not long ago. It's quite scary/amazing.

How could it be actively harmful if it wasn't harmful last month when it was the frontier model?

His phone number and email address are in his resume, so you could try to contact him

Please don't call people randomly. Unless you're offering him a job which is what the resume is for...

When the linkedin is down, medium is left unattended, the personal domain is not working, we can reasonably guess he doesn't (or is unable to) care about the project or online presence anymore.

You don't have to call them.

You can use WhatsApp, Signal, or SMS. Drop them a message, see if you get a reply.

I think that's silly. Do we really live in an age where we feel it's better to simply not communicate with people in the slightest?

Give them a call, you're not harassing them. If they choose not to answer or call back a voice mail number, then you can presume they don't want to be contacted.

> Give them a call, you're not harassing them.

Before posting this idea online... Maybe, possibly, but personally I still think it's a bad idea.

After posting this on HN - no! If you think it's a good idea, so will other people reading this. (And others have before you) After the post reaches the front page - absolutely no - there's a bunch of socially awkward people already thinking about calling the author and they really should NOT DO THAT.

The author owes us absolutely nothing and if they want to disappear, that's their right. Calling them is demanding their time in a not trivial to ignore way. Just write an email that can be deleted async.

You are right: it is silly, but also, given the amount of robo-calls in the US, cold calling someone you don't know is a good way to be put on auto-spam.

If you really want to reach out, his email seems to be the way he prefers to be reached, so that's what I'd recommend.

PS: He did some commits to his personal website about 1.5 years ago: https://github.com/Hopding/Hopding.github.io/commits/master

> I think that's silly. Do we really live in an age where we feel it's better to simply not communicate with people in the slightest?

I agree it’s silly. But it’s also the prevailing view that I’ve seen.

I still answer calls, even if 95% of them these days are either phishing attempts or vendors trying to sell me stuff. But my friends will text me first and say “can I call you” even if I say they can just call.

I think the post you're responding to would agree, but is trying to make the argument that it isn't worth the cost:

> spent insane amounts of money, labour, and opportunity costs of human progress on this

That said, I would 100% approve of certain people pouring all their energy into AI to rather focus on teaching squirrels chess!

I reckon a lot of re-writes tend to take "them 5x as long as they had hoped" and "isn't even any easier than the old code" exactly because writing the code wasn't the problem in the first place.

It's business logic, edge cases and other small necessary details that accumulated over time which make the code 'messy'. Once you've integrated all those in the new system, it likely looks equally messy. And discovering and implementing all those extra requirements is probably what took you the longest.

Not to say this applies to all re-writes or that AI tools can't help the process

Access is one of my concerns with coding agents - on the one hand I think they make coding much more accessible to people who aren't developers - on the other hand this access is managed by commercial entities and can be suspended for any reason.

I can also imagine a dysfunctional future where a developers spend half their time convincing their AI agents that the software they're writing is actually aligned with the model's set of values

Feels weird sharing a linkedIn post, so here is the text:

> Today, Y Combinator is announcing that YC-funded startups can choose to receive their funding ($500k) in stablecoins. > > We believe stablecoins like USDC are setting the stage for a new fintech renaissance and broader global access to financial services. Sending money should be as easy as sending a text message. Stablecoins make that possible: cheap, fast, and global, using currencies people already trust. > > Some of the fastest-growing YC startups in recent years like Aspora and DolarApp use stablecoins to power faster, cheaper financial services across India and Latin America. Plus, with the passage of the GENIUS Act and growing adoption by financial institutions, we’re bullish. > > Whether crypto-focused or not, we expect many YC startups to use crypto in some way, from payments to banking to capital raising. > > If you’re building onchain, apply for our Spring ‘26 batch by Feb 9: https://lnkd.in/dHEJ9Fc > > And for more on how we’re thinking about building onchain and the types of startups we’re excited to fund, check out our Request for Startups: https://lnkd.in/gSaJ_Xst

And how do we know you won't go and rent out your account to some AI once it's been white-listed (≖_≖ )