alfalfasprout

Which makes it a very reasonable take. AI as a research and ideation tool is invaluable (provided due diligence is done on sources, etc.)

AI generated writing is (usually) hot garbage.

It's trivial to hit that kind of quota if you're trying to execute on major projects. Especially as you start having dozens or hundreds of subagents investigating, prototyping, and working on different things.

Yeah, and synthetic data is pretty darn good nowadays especially with access to frontier models.

The moat for OAI and anthropic seems to be very quickly shrinking. Chinese labs are now using RSI-like approaches and even without resorting to heavy distillation they're catching up in a couple of months vs. what would have been 6-12 months a year prior.

And as these models get better the pace of training is quickly speeding up too.

This doesn't bode particularly well for anthropic/OAI after they go public.

assuming demand remains elevated and growing, maybe. But spend on AI is pretty stratospheric right now... companies are already starting to clamp down on spend. This makes you really wonder if there will be sufficient demand at current commodity prices for eg; OSS models to justify all these data centers.

there are many things to blame POTUS for. This isn't one of them... most major fabs were already outside the US to begin with and the US chipmaking industry (well, really Intel) was really struggling even years ago.

How is this not a frontier model? It's bleeding edge in its own niche. It's not a frontier LLM; however, applicable to many of the things people use LLMs for.

very HN take. Reality is 3.8 flash is good at what it does (being fast + cheap)... but that market is becoming very saturated and google is not competing at all at the frontier where major spend lies.

Yeah but politicians get to claim this is a win. Let's be honest, that's the real reason the prosecution went for a deal this small.

Probably Finnegan's Wake

btw it's colombia not columbia.

If anything, AI tends to vastly overengineer things (in a bad way) which makes bloat inevitable. But I'm not sure layers of management are to blame. If anything, good management incentivize the right things and focus a team on work that matters. Bad managers only try to empire build and incentivize the wrong things in their engineers.

I've seen orgs that try to flatten too much end up with ICs basically left in the ether without meaningful direction which causes them to 'build to survive'.

Neither of them are "almost there"...

The weights are open, the training data obviously isn't. But a key thing you can do with open models is fine tune them. So while they may not be SOTA at everything, they can become SOTA at your particular business use case.

There's a massive difference between your kids vibe coding something and an engineer using AI to implement something. If you're unable to discern the difference, that's something to reflect on :)