HackerTrans
TopNewTrendsCommentsPastAskShowJobs

galsapir

80 karmajoined vor 6 Monaten
sparse thoughts on things i find online

Submissions

Giving a domain a hill to climb: benchmarking as data activation

sparsethought.com
12 points·by galsapir·vor 9 Tagen·7 comments

A bitter lesson for medicine, or a benchmark problem?

sparsethought.com
2 points·by galsapir·vor 28 Tagen·0 comments

Can LLMs Beat Classical Hyperparameter Optimization Algorithms?

arxiv.org
120 points·by galsapir·letzten Monat·20 comments

Gemma 4 E4B as a primary local LLM (replaced Qwen)

digg.com
2 points·by galsapir·letzten Monat·0 comments

PEEK: Give Your Agent an Orientation Cache (MIT CSAIL, Khattab group)

zhuohangu.github.io
3 points·by galsapir·vor 2 Monaten·0 comments

Hyperagents (Meta Research)

arxiv.org
2 points·by galsapir·vor 2 Monaten·0 comments

The Unreasonable Effectiveness of HTML

claude.com
3 points·by galsapir·vor 2 Monaten·2 comments

[untitled]

1 points·by galsapir·vor 2 Monaten·0 comments

The Comparator in Clinical AI

sparsethought.com
2 points·by galsapir·vor 2 Monaten·1 comments

Borges' cartographers and the tacit skill of reading LM output

galsapir.github.io
40 points·by galsapir·vor 3 Monaten·10 comments

[untitled]

1 points·by galsapir·vor 4 Monaten·0 comments

Best read of 2026 so far was written in 1880

galsapir.github.io
1 points·by galsapir·vor 4 Monaten·1 comments

Anthropic launched community ambassador program

claude.com
1 points·by galsapir·vor 4 Monaten·0 comments

LLMs as nudging research towards luke-warm middle

nature.com
1 points·by galsapir·vor 5 Monaten·0 comments

[untitled]

1 points·by galsapir·vor 5 Monaten·0 comments

[untitled]

1 points·by galsapir·vor 5 Monaten·0 comments

[untitled]

1 points·by galsapir·vor 5 Monaten·0 comments

[untitled]

1 points·by galsapir·vor 5 Monaten·0 comments

[untitled]

1 points·by galsapir·vor 5 Monaten·0 comments

How do you evaluate a foundation model before you know what it's for?

galsapir.github.io
1 points·by galsapir·vor 6 Monaten·1 comments

comments

galsapir
·vor 6 Tagen·discuss
thanks for reading it properly and engaging with the argument!

writing is hard, expressing ideas cleanly is harder! working on it.
galsapir
·vor 9 Tagen·discuss
curious where the disagreement lands: the claim i'm least sure of myself is that measurement alone already counts as activation (nothing in the weights changes, so it's a looser sense of the word than usual) the part i'd defend harder is the eval -> reward one: once a benchmark becomes the thing you train against, its flaws stop being measurement error and start being incentives. if you're pushing back somewhere in there, i'd genuinely like to hear it
galsapir
·vor 27 Tagen·discuss
really interesting that its basically almost 80% claude opus..
galsapir
·vor 27 Tagen·discuss
yeah its really counterintuitive i think; i.e, getting the right framework and structure for this to work probably isn't trivial, models really hate playing well together. i wonder how their version would fair in real world use.
galsapir
·letzten Monat·discuss
i feel like i've had exactly the same thought in the past :-0 might even have written about it. feel your pain
galsapir
·letzten Monat·discuss
sometimes I also feel it tries to optimise for "per line coverage" over more "real, complex use cases" type tests
galsapir
·vor 2 Monaten·discuss
hey that's pretty cool. I think I still prefer "distill HN" cleanliness though. What made you create this.
galsapir
·vor 2 Monaten·discuss
axon discharge is brilliant. adopting.
galsapir
·vor 2 Monaten·discuss
oh sorry! didn't catch the one Thanks, I'll comment there
galsapir
·vor 2 Monaten·discuss
[dead]
galsapir
·vor 2 Monaten·discuss
From the link: "Shot from 90 perspectives, 88 focus stacked images each. Nikon Z8, full frame, f/7.1, exposure 1/160, ISO 100, Laowa 180mm macro lens, with LED light and bluescreen." Insane!
galsapir
·vor 2 Monaten·discuss
I think the question he tried to raise was "is this needed? Aren't today's / tomorrow's models well-enough equipped to deal with just OPEN API?" (idk, just if I understand the question)
galsapir
·vor 2 Monaten·discuss
got me at "Most often scientists believe they understand more than they do, making their belief an illusion." but why is it still bothering me? 1. feels unfalsifiable in spirit 2. somewhat restates "all models are wrong, but some are useful" less cleanly 3. doesn't really offer like, what can we do as science people? tomorrow morning perspective
galsapir
·vor 2 Monaten·discuss
[dead]
galsapir
·vor 2 Monaten·discuss
author here. the part i'd actually like discussion on is the buried finding: physicians+GPT-4 didn't outperform GPT-4 alone on the management cases, and on the landmark cases the model alone beat the model+physician. the paper reports it and moves on. that's the 2026 question, and it's the one a Science-level platform could have been used to ask
galsapir
·vor 3 Monaten·discuss
Hey thanks! I do wonder that. I think that even if specifically for code smell the things would be subtler, for other forms of AI driven averageness (especially in areas where we can't RLVR the models to perfection) it might still be present. But yeah I wonder how those thoughts will age (and how we'll update our priors accordingly).
galsapir
·vor 3 Monaten·discuss
yeah I was really thinking about what the best "umbrella term" would be here. Since "LLM" is too widely used in a really specific context and "AI systems" felt niche I ended up with "LMs". Idk, up for debate..
galsapir
·vor 3 Monaten·discuss
haha that's a style choice (takes more work to get lowercase text these days). But yeah legit ;-)
galsapir
·vor 3 Monaten·discuss
Thanks! I'll check it out.
galsapir
·vor 3 Monaten·discuss
[dead]