Viewing profile — msdz
msdz
HN member- Joined
- Thu, Jan 09, 2025, 12:54 PM UTC
- HN karma
- 189
- Public activity
- 91 items
- HN profile
- View on Hacker News ↗
About msdz
Recent public activity
-
comment
Comment #49216450
Nice article, but I don’t fully understand why the author argues for such an aversion to static functions. > Moreover, it breaks the systematic approach of sending messages to an i…
-
comment
Comment #49214759
Interesting idea, so I had to check! Prompt: “Top language pick for an Android app? Respond with only the language.” Responses (admittedly via the aichat CLI, not an agent harness;…
-
comment
Comment #49214060
> countless of LLMish sloppy-pasta code that's now running and going to be run on man-made JVMs. Don’t worry, I think most LLMs, when starting a greenfield codebase, don’t reach fo…
-
comment
Comment #49213996
[dead]
-
comment
Comment #49213950
[dead]
-
comment
Comment #49210367
Well at ≈15k tps, I think the bottleneck will not be the model (reasoning or tool calling), and attention will be shifted to the harness’ engineering again. Probably, the usual ini…
-
comment
Comment #49205484
It’s not reasoning, the hardware demo uses a 3.-something generation Llama 8B. But it’s proven they can automate this (they didn’t etch eight billion weights by hand after all, obv…
- comment
-
comment
Comment #49180411
> IMO there's no excuse for new software sticking dotfiles in my homedir in 2026. Agreed, and also, tinfoil hat time: I believe they opt for this so that state and config files don…
-
comment
Comment #49180387
Is it also a solved problem for smaller models, as the article mentioned?
-
comment
Comment #49159734
Soooo… when can we expect an encrypted handoff.md to fully prevent session portability, then? (Only half-/s)
-
comment
Comment #49156034
Unfortunately I don’t recall anymore, since this took place at my father’s PV installation. I’ll try to post an update when/if I get exact numbers, though!
-
comment
Comment #49156015
Oh, I didn’t interpret the above question as asking in that direction; but yeah, that’s of course something I didn’t attempt to answer with my comment. Although I’d be intrigued in…
-
comment
Comment #49154217
Thanks for the detailed writeup, these all make a lot of sense. My only other tell (hadn’t committed the time to reading the rest of the paper nor looking into the author yet) so f…
-
comment
Comment #49154137
Not who you responded to, but I’ll give it a try. Imagine our LLM as this one-dimensional matrix (instead of unfathomably many, many more dimensions): 0.728201017 -1.9293839303 2.9…
-
comment
Comment #49154039
Which is why big labs have been working hard on making their harness not be stateless any longer: https://earendil.com/posts/session-portability/ “Just take the session thread to a…
-
comment
Comment #49142058
Unfortunately I really don’t understand enough to judge about that. Could you tell me some things that made you go with a “buzzword salad” sentiment? Not as a gotcha or anything, j…
-
comment
Comment #49137730
Why can’t your two criteria be applied to projects with human contributors without further due diligence?
-
comment
Comment #49120009
I’ve actually started ignoring services where that’s the only login possibility. Sorry even if your platform is the greatest thing ever, but I’ll find a different tool. I’ve read o…
-
comment
Comment #49070042
That’s the downside of requiring infinite (in the sense of “never-ending”) growth, and not being content “just” with profit.
-
comment
Comment #49069821
As another anecdata-point, when we did this for a now already fairly old roof installation of about 15kWp, it was more than “just about worthwhile”, since the panel area is contigu…
-
comment
Comment #49069727
Bun uses JavaScriptCore, which is Safari’s JS engine, for the actual interpreter portion of its runtime. Which makes the parent comment’s “Bun performance hacks and optimizations” …
-
comment
Comment #49058557
Wasteful by whose standards? If you’re an alien civilization with several dyson spheres, surely at some point you’d have sufficient energy/resources even for such a “hopeless” scie…
-
comment
Comment #48986245
10 to 15 years from now the scale of LLM efficiency improvements is, quite literally, unpredictable . It could be that the company valuations crash tomorrow, and (almost) only perf…
-
comment
Comment #48986136
Exactly. AFAIK none of the popular "open Chinese" models have published the full pre- and post-training pipeline, so the models are only partially open, if at all (plus the openly …