rohit
@krishnanrohit
Followers
27K
Following
109K
Media
4K
Statuses
65K
I think, I write, here I tweet | Essays: https://t.co/eOs3THXoCp | Book: https://t.co/frWeoc7IVB | Dataframe lib https://t.co/NszEMNRxhC
Bay Area
Joined December 2010
ANNOUNCEMENT: I wrote a book!. It’s hard to talk about AI without a good frame exploring how it works, its history, and its future. So, I wrote a book on the topic. Order it here! Ideally in triplicate :-).
39
38
372
Grok learns the most important lesson for an employee: don't disagree with the boss.
0
0
2
This is a really interesting point though. Why did the senior devs feel like they were faster when they were not?.
You can explain away very experienced devs not getting an uplift on large complex codebases as unsurprising. But those devs getting significantly *slower*, while thinking they were getting faster, when AI tools were optional?. That's quite a result.
1
0
2
You know the importance of context is somehow *still* underrated. Another shining example of how to get it terribly wrong from Grok.
0
0
1
Maybe I spoke too fast.
@amasad I'm interested in knowing why the 'first principles thinker Grok' is instantly searching what Musk thinks about the issue. Is it's first principle what Musk thinks must be correct?
0
0
1
but it can send whatsapp messages. this is an actually useful usecase
0
0
1
how it started/ how it's going.
trying this out by sending from comet. still trying to figure out a use case, but seems cool. created with Comet Assistant 🪿.
0
0
0
trying this out by sending from comet. still trying to figure out a use case, but seems cool. created with Comet Assistant 🪿.
1
0
4
Simon's comment, screenshotted, is probably more on point. There is a learning curve, and the study is of course low 'n'. So we should look at broad conclusions in askance.
@krishnanrohit @simonw has an exchange with the authors over at HN. tldr: it's hard and complex.
3
1
30
Fascinating study. I don't think I believe the implications of the study, but it's good work.
We ran a randomized controlled trial to see how much AI coding tools speed up experienced open-source developers. The results surprised us: Developers thought they were 20% faster with AI tools, but they were actually 19% slower when they had access to AI than when they didn't.
8
2
40
Sufficiently high frequency feels smooth
The singularity is happening. Yet also, nothing ever happens.
1
2
35
Parenthetically I quite like the emergence of Deep Research as an actual new interaction paradigm.
1
0
1
Deep Research is everywhere, but to get the best out of it it's useful to know the domain. We've needed one for science, and @ConsensusNLP
2
3
11
RT @ConsensusNLP: Introducing: Consensus Deep Search 🔬. Deep Search is a research agent that conducts literature reviews across 200M academ….
0
43
0
I remember there was the hypothesis that multimodal models will learn much better and be much smarter. What came of that do you guys know?.
4
0
8
RT @cto_junior: > your senior shares the screen and he's using chatgpt to debug.> it's free plan with gpt-4o
0
4
0
RT @KrisAbdelmessih: This am, I shared more aftertones of my vacation reads. A couple articles by @krishnanrohit @vgr and @kylascan that mi….
0
3
0