I presume post training is significantly easier than the distillation/training the top Chinese labs are doing.
I wonder if, similar to the American labs, they'll become stingy with their weights once they start getting immediately undercut by a wave of slightly better derived models.
I have this same cactus same thing, but later on I realized even when I do nothing, and even my tinnitus is there I realized it doesn’t bother me anyway. I miss that deep winter / snow silence now and then but kind of learned to live with it I think. I realize this is very different experience for everyone.
I noticed the more I do nothing (and think nothing) the more my tinnitus lessens. I think mine has something to do with permanent neck muscle tension, the result of everyday rush and stress.
I think you're equating doing nothing with doing nothing productive. You're usually still ruminating or trying to distract yourself from the source of the pressure.
I wonder how long it'll be before AI labs find a revenue stream that doesn't involve renting out their models, and stop letting us ride on their coattails.
They'd tease us with solutions to hard problems, with code that is orders of intelligence higher than any human or public model can grok.
That'd turn all the whining to begging real quick.
I'm guessing the next generation of US frontier models will be heavily anti-distillation at the cost of user experience (significant rate limiting, more flat out refusals, hidden thinking, etc).
And as long as they maintain a significant advantage in capability, we will continue to kiss the ring.
I'm assuming finding vulnerabilities in open source projects is the hard part and what you need the frontier models for. Writing an exploit given a vulnerability can probably be delegated to less scrupulous models.
I won't try to speak for anyone other than myself, but my multiplier is definitely over 1.5x, probably higher than 5x.
I choose to sit on my hands in my freed up time so upper management does not catch on to and exploit this fact. Eventually they will though via overzealous coworkers.
It’s easy to produce a high volume of code, sure, but it is not equally easy to test, verify, and integrate it. And with a high volume of code, there is a high volume of shit to review & test & integrate. For companies that give a shit about not vibe coding their way into a disaster (because they have lucrative enterprise contracts that depend on reliability & security), that’s the real blocker. (Plus, these types of projects are big, not trivial, and things are harder to integrate & properly test because of that.)
Not to mention, if a team wants to keep a semblance of understanding of what they own & ship… it can be exhausting to have a huge volume of new code coming into the system.
It’s definitely a productivity unlock. For sure. But there are a lot of knock-on effects we’re still figuring out that counteract how much extra “value” we’re shipping
In my case, the volume of code is roughly the same. I'm not using the efficiency towards pumping out more code, just using it to be AFK more.
I spend enough time iterating and refining to the point I'm comfortable taking ownership of the outputted code. Perhaps hypocritically, I do mald when people upload code for review that they clearly haven't taken the effort to read through critically.
People with a lower multiplier are either in the minority of developers solving genuinely hard/novel problems or, more likely, they've just not figured out how to tap into AI's potential.
Granted, to your point, a decent chunk of the HN crowd belongs to the former and can't relate to us paycheck stealers.
I always hear people say this, but it’s not clear to me what exactly is so difficult about using AI that otherwise-competent developers “can’t figure it out”
I wonder if, similar to the American labs, they'll become stingy with their weights once they start getting immediately undercut by a wave of slightly better derived models.
reply