HNHacker News
TopNewBestAskShowJobs

alfalfasprout

4,549 karma · joined November 11, 2015

submissionscomments
alfalfasprout··on xAI is looking more like a datacentre REIT than a frontier lab
It's not just that there's a circular deal it's that they're prevalent. And worse, with frontier labs IPOing seeking astronomical valuations that means a lot of the public is now exposed too (even if they don't all get fast-tracked into eg; the SP500).

The problem is the valuations assume astronomical growth... that is likely impossible for all of them to simultaneously achieve. Which means something's got to give.

alfalfasprout··on xAI is looking more like a datacentre REIT than a frontier lab
> the big providers are charging full freight for inference

They're not and it's not clear why you seem to believe that. The immense capex for buildouts, training costs, etc. are not rolled into inference costs. Moreover, companies are already rapidly starting to re-evaluate token spend.

alfalfasprout··on MiMo-v2.5-Pro-UltraSpeed: 1T model with 1000 tokens per second
Generally, I agree because what happens is the messaging around AI is doing more, faster. Not using AI to deliver at a higher quality level, etc. But I think it boils down to incentives and discipline. So given the incentives we have today at most workplaces faster AI will just be used to produce more slop.
alfalfasprout··on When AI Builds Itself: Our progress toward recursive self-improvement
Absolutely! Yes. This rhetoric of inevitability only benefits these AI companies.
alfalfasprout··on Uber's $1,500/month AI limit is a useful signal for AI tool pricing
Right. Which means tokens are actually being priced well under cost once you factor in all this datacenter/GPU capex. Also worth noting the datacenters are not purely for training. They're for inference too.
alfalfasprout··on Claude Opus 4.8
I'm actually currently studying this :)

Honestly... not that dramatically. Each release is much more marginal. And quoted official benchmarks doesn't translate very well into the real world.

4.7 regressed hard in some ways. But a compounding factor too is that the claude code harness seems to nerf the model after a few months. Probably to reduce token use.

So far 4.8 seems less verbose but we'll see in practice what it translates into meaningfully.

alfalfasprout··on Ferrari Luce
The interior is also, frankly, very meh.
alfalfasprout··on Redis array: short story of a long development process
Having tried something similar, the perceived speedup does not, in the steady state, last.

To get a quality, lasting, result you're ultimately having to carefully study everything otherwise you end up quickly accumulating cognitive debt and the speedup soon shrinks as you're constantly having to revisit the initial approaches.

alfalfasprout··on Ti-84 Evo
Heck, you could do a decent amount with the CAS back in the TI-89.
alfalfasprout··on Claude Code refuses requests or charges extra if your commits mention "OpenClaw"
on claude using bedrock it simply refuses to acknoweldge the existence of OpenClaw (Opus 4.7)
alfalfasprout··on AI Resistance: some recent anti-AI stuff that’s worth discussing
Fascinating, because I've seen the exact opposite across the PNW.
alfalfasprout··on AI-assisted cognition endangers human development?
Yep. And this is why as hard as AI companies are pushing that these tools can be a replacement for expertise, it's ironic that the experts are the ones that often get the highest ROI because they know how to converse about the relevant subjects with a high degree of precision (and know what to look for, what to challenge, etc.).
alfalfasprout··on Claude Code Routines
Yep. Trust is easy to lose, hard to earn. A nondeterministic black box that is likely buggy, will almost certainly change, and has a likelihood of getting enshittified is not a very good value proposition to build on top of or invest in.

Increasingly, we're also seeing the moat shrink somewhat. Frontier models are converging in performance (and I bet even Mythos will get matched) and harnesses are improving too across the board (OpenCode and Codex for example).

I get why they're trying to do that (a perception of a moat bloats the IPO price) but I have little faith there's any real moat at all (especially as competitors are still flush with cash).

alfalfasprout··on The Future of Everything Is Lies, I Guess: Safety
Well, part of the problem too is there's zero accountability. Who decides what it means to be aligned and how does that evolve over time?

No matter what, common people are quickly losing agency in that discussion.

alfalfasprout··on Molotov cocktail is hurled at home of Sam Altman
It's quite prevalent in tech too-- however, folks tend to be quiet because the "use AI for everything or else" hammer is being used across the industry.
alfalfasprout··on Good Taste the Only Real Moat Left
> Strangely I find traditional software engineers, especially experienced ones, are generally the worst at achieving success. They often treat working with an agent too much like software engineering and end up building bad software rather than useful solutions to the core problem.

This feels a bit like a strawman. How do you assess it to be bad software without being an engineer yourself? What constitutes successful for you?

If anything, AI tools have revealed that a lot of people have hubris about building software. With non-engineers believing they're creating successful work without realizing it's a facade of a solution that's a ticking time bomb.

alfalfasprout··on Anthropic expands partnership with Google and Broadcom for next-gen compute
100% agreed. Sadly, lots of people out there with the "trust me bro, just need more compute". Hopefully we don't consume all the planet's resources trying.
alfalfasprout··on F-15E jet shot down over Iran
Well, A-10s are well suited for strafing runs, etc. Presumably they'd be sent in if the area they're entering is presumed safe. That clearly didn't pan out.

The reality is avoiding a ground operation was probably the wrong move at this point (ignoring the spicier broader debate of if the whole Iran campaign was the right call or not)

It's really hard to truly guarantee surface to air capabilities are gone when you're relying purely on sat images + aerial surveillance (and obviously this carries risk). Iran has fairly portable SAM systems that are public knowledge.

alfalfasprout··on We haven't seen the worst of what gambling and prediction markets will do
Acting within your morals is not incompatible with serving the company's interests. Especially if it means your team is very much still competent while maintaining a culture that is healthy. That leads to better delivery.

Avoiding working in deeply unethical areas also shields the company from legal or PR liability.

alfalfasprout··on Arm AGI CPU
the whole AI space is rife with much worse example of what could be considered securities fraud tbh
alfalfasprout··on Two pilots dead after plane and ground vehicle collide at LaGuardia
Yeah but TCAS works inside each airplane. ATC (and ground operations) require coordinating across multiple types of aircraft, at airports across the world, with high precision AND humans in the loop (there are A LOT of edge cases).

This is a REALLY hard problem that the US cannot solve alone. It would require extensive global coordination.

Not insurmountable, but this is not something you can easily roll out piecemeal. If even a single aircraft lacks the compatible equipment you're back to the existing system.

alfalfasprout··on GitHub appears to be struggling with measly three nines availability
It's not though... it's sitting a hair above 90%
alfalfasprout··on A rogue AI led to a serious security incident at Meta
When organizational incentives penalize NOT using AI and firing the bottom x% regularly then are you really surprised LLM outputs aren't being scrutinized?
alfalfasprout··on Tech employment now significantly worse than the 2008 or 2020 recessions
Yep. I've seen this personally. The ROI of a great engineer is bigger than ever.
alfalfasprout··on [dead]
This reads like a bad linkedin post...
alfalfasprout··on Show HN: I built a sub-500ms latency voice agent from scratch
Also consider using Cerebras' inference APIs. They released a voice demo a while back and the latency of their model inference is insane.
alfalfasprout··on The path to ubiquitous AI (17k tokens/sec)
Cerebras is a totally different product though. They can (theoretically) run any frontier model provided it gets compiled a certain way. Like a wafer scale TPU.

This is using hardwired weights with on-die SRAM used for K/V for example. It's WAY more power efficient and faster. The tradeoff being it's hardwired.

Still, most frontier models are "good enough" where an obscenely fast version would be a major seller.

alfalfasprout··on Fff.nvim – Typo-resistant code search
I've been using this for a few months now and it's amazing. Crazy fast and the vast majority of the time I get the result I want in one shot.
alfalfasprout··on Fff.nvim – Typo-resistant code search
They have prebuilt binaries now for fff
alfalfasprout··on The only moat left is money?
The thing is, the barrier isn't near zero. The time to reach an MVP has just decreased. But you still very much need expertise, strategy, etc. to deliver something worthwhile. The bar has just increased.
← PreviousPage 2 of 34Next →