cgearhart

6 ms·

31.
by cgearhart 10mo ago
So…great for prototyping (where velocity rules) but somewhere between mixed to negative for critical projects. Seems like this just puts some mildly quantitative numbers behind the consensus & trends I see emerging.
32.
by cgearhart 10mo ago
“Better than me” != “good” I know approximately nothing about approximately everything. Claude seems pretty good at those things. But in every case I’ve used Claude Code for something I do know about it’s been unsatisfactory as a solo ope
33.
by cgearhart 10mo ago
I love the way you phrased this.
34.
by cgearhart 10mo ago
Best of luck! I hope it’s as amazing for you as it was for me.
35.
by cgearhart 10mo ago
I had been working in civil service for the US Navy for about 10 years in operations research & systems engineering. It was very hard to break out of that role to any private industry—especially for the ML roles I wanted, which I thin
36.
by cgearhart 10mo ago
Nah. There was some of that as the tools available to unis improved alongside Udacity, but it was a very intentional choice. The business with GT made $X/year while the consumer & enterprise businesses brought in $20X/year. It
37.
by cgearhart 10mo ago
In 2013-2016 Udacity was very actively collaborating with GT and had in-house content production. The projects were designed by highly experienced instructors in direct partnerships with real companies to make them realistic and relevant, a
38.
by cgearhart 10mo ago
I was also in the first cohort. Graduated in 2016. Transformed my career. 100% would recommend.
39.
by cgearhart 10mo ago
I was the head of enterprise curriculum in 2018 and an OMSCS grad in 2016. This was a weird time to work for Udacity and the company went thru a major shakeup in 2019. The “breakup” with GT happened before the focus on enterprise and the en
40.
by cgearhart 11mo ago
When I say something like this it usually means “I don’t want to dictate your job to you. You’re here because you’re smart, ambitious, and capable. We’ve talked at length in team settings and 1:1 about our goals. What do you think are the
41.
by cgearhart 11mo ago
So DSA means a lightweight indexing model evaluated over the entire context window + a top-k attention evaluation. There’s no soft max in the indexing model, so it can run blazingly fast in parallel. I’m surprised that a fixed size k doesn’
42.
by cgearhart 11mo ago
It’s mind boggling that image generators can solve physics and chem problems like this—but I will note that there are a few slight mistakes in both. (An extra i term in LHS, a few of the chemical names look wrong, etc.) Unbelievable that
43.
by cgearhart 11mo ago
Many years ago it was Hershey chocolate fudge jars that we’d put in the cupboard when they were empty. Then shrinkflation led to smaller jars and it lost some of the charm. I still have a few of the old ones.
44.
by cgearhart 11mo ago
I’m in the USA. I can drop into the local Target and pick up a set of 6 pint glasses for around $10. Meanwhile a set of 6 Picardie 500ml glasses is $48 at Williams-Sonoma. (A bit cheaper online.) Worth every penny so far.
45.
by cgearhart 11mo ago
I have bought a lot of glassware over a couple decades. My usual trick is to look at restaurant supply shops, since they usually have a wide variety to select trade offs in style, price, quality, etc. Somehow I’d never come across Duralex u
46.
by cgearhart 1y ago
BDFL left a long time ago. It’s not opinionated anymore. The language went from being small enough to fit in that guy’s head to a language controlled by committee that’s trying to please everyone.
47.
by cgearhart 1y ago
I’ve been trying uv lately to replace my normal workflow of selecting a python with pyenv for the shell, then making a venv, then installing a bunch of default packages (pandas, Jupyter, etc). So far the only benefit is that I can use just
48.
by cgearhart 1y ago
I think there’s often a misalignment of incentives when annual perf reviews are judged on feature work delivered not quality. Engineers who spend any time polishing or finding and fixing bugs wind up rated mid, while folks who quickly crank
49.
by cgearhart 1y ago
I keep talking to people who say stuff like “Claude wrote it all for me in a day”, but when I look at the code (or try it myself) it’s just so much useless code. I recently asked for Python code to parse some data into a Pandas dataframe
50.
by cgearhart 1y ago
This is a fair point and clarification. :-)
51.
by cgearhart 1y ago
The title is a succinct snippet that spoke directly to researchers at the time. The transformer architecture is somewhat obvious (especially in retrospect), but it was still very surprising because no one was really going this direction. Th
52.
by cgearhart 1y ago
I mean, they’ve released a framework to access the on device model https://developer.apple.com/documentation/FoundationModels
53.
by cgearhart 1y ago
I find this damn impressive, but I’m disappointed that while there’s some version of the final proofs released, I haven’t seen the reasoning traces. Although, I’ll admit the only reason I want to see them is because I hope they shed some li
54.
by cgearhart 1y ago
The current gen VLA architectures include some tricks (like compressed action tokenization and diffusion decoding) to reach action frequencies between 50-200hz. I think they’re _more_ efficient this way than regular LLMs trying to do everyt
55.
by cgearhart 1y ago
Threads like this are why I love HN. Thanks for teaching me new things. :-)
56.
by cgearhart 1y ago
I think there was some debate on this, actually. I did a lot of research on the subject in the late 2010s and it seems like there were those who felt like limiting the branching factor was the goal, while others felt like fast eval to guide
57.
by cgearhart 1y ago
In old-fashioned AI, it was generally believed that the best way to spend resources was to exactly evaluate as much of the search tree as possible. To that end, you should use lightweight heuristics to guide the search in promising directio
58.
by cgearhart 1y ago
It seems like there’s been a lot of progress here, but it also seems like there’s an elephant in the room that RNNs will _always_ have worse memory than self-attention since the latter always has complete access to the full context. We pay
59.
by cgearhart 2y ago
Why is there an expectation that “nearby” tokens are relevant to increase the information in the similarities? That seems like it would hold true within individual words, but the whole point of attention was to solve long range dependencies
60.
by cgearhart 2y ago
I don’t think there’s consensus. Some papers have shown that just giving the model more tokens improves the results, ie chain of thought allows more computation to happen and that is enough to improve results. Others have argued that the sm

More ›