tibbar

16 ms·

121.
Cancer(jacobbrazeal.wordpress.com)
2 points by tibbar 11mo ago|0 comments
122.
by tibbar 11mo ago
Sometimes people use "waste" in the sense of "spent selfishly/frivolously." I enjoy evenings spent with prestige television and a tub of ice cream but, past a certain point, it's definitely a waste of time, esp
123.
by tibbar 11mo ago
I think you need to pair empathy with its counterpart, "willingness to be disliked." Empathy is great for building relationships but, taken alone, can make you a slave to doing whatever people want. Which stops working as soon as
124.
by tibbar 11mo ago
Well, you also have to allow it to train on your data. Although this is not explicitly about your Google drive data, and probably requires you to submit a prompt yourself, the barriers here are way to weak/fuzzy for me consider grantin
125.
by tibbar 11mo ago
Oh, certainly. And if you have any resources besides attention - money, or social capital, for instance - people will happily take those too. It's not that they are wrong to ask, but the need is truly infinite, and no one will create g
126.
by tibbar 11mo ago
Important context is that the author was a social outcast as a child. I also had this experience, and I can tell you, I was just desperate to figure out how to get people to like me. It wasn't that I wanted to manipulate them, I just r
127.
by tibbar 11mo ago
This resonates. I spent many years going hard at `5: Connection is about projecting love and acceptance`. I focused on providing unconditional acceptance to everyone I met, inviting them to share their burdens and listening empathetically
128.
Advice for System Design Interviews(jacobbrazeal.wordpress.com)
2 points by tibbar 11mo ago|0 comments
129.
by tibbar 11mo ago
People sometimes describe me as a "golden retriever." I.E., I present as quite upbeat. 50% of that is just nervous energy, but the rest might just lie in having a poor memory. I can experience something quite negative, even trauma
130.
We are short DoorDash, Inc(img1.wsimg.com)
7 points by tibbar 1y ago|0 comments
131.
by tibbar 1y ago
> Having been tested for many years of my life (with pretty good results), here are some rules of thumb that I feel helped me Oh, come now, I think Karpathy adequately qualifies his advice. :-)
132.
by tibbar 1y ago
Given the physical limits of the universe and our planet in particular, yeah, this is pretty much always true. The interesting question is: what is that limit, and: how many orders of magnitude are we away from leveling off?
133.
by tibbar 1y ago
Man, watching the robot delivering a package was an uncanny valley moment for me. Frickin' Star Wars droids. I think it would be a little too weird for me to be surrounded by lifelike military-looking androidds.
134.
by tibbar 1y ago
I mean, I agree. There is not yet a clear path/story as to how a model can provide a consistently expert-performance on real-world tasks, and the various breakthroughs we hear about don't address that. I think the industry consens
135.
by tibbar 1y ago
Yeah that is fair. I admit to being a bit bummed out as well. One might almost say that if O3 was effectively GPT5 in terms of performance improvement, that we were all really hoping for a GPT6, and that's not here yet. I am pretty opt
136.
by tibbar 1y ago
The intensely negative reaction to GPT-5 is a bit weird to me. It tops the charts in an elaborate new third-party evaluation of model performance at law/medicine, etc [0]. It's true that it was a bit of an incremental improvement
137.
Nine Aaron Judges(jacobbrazeal.wordpress.com)
2 points by tibbar 1y ago|0 comments
138.
by tibbar 1y ago
I would argue that code review of large changes is hard because we don't write good PR descriptions. If you just get 200 files to review in alphabetical order, sure, that's basically an impossible task. Instead, the PR description
139.
by tibbar 1y ago
To slightly amend my original take, much of this data is specifically for evals . That is, the labs need some sort of data set and grading scheme by which they can measure and direct their progress.
140.
by tibbar 1y ago
I think this is, essentially, a wishful take. The biggest barrier to models being able to do more advanced knowledge work is creating appropriately annotated training data, followed by a few specific technical improvements the labs are work
141.
by tibbar 1y ago
> Removed verbose explanations mixed with instructions Is Claude rewriting generic instructions once, or is it rewriting the core task statement each time? If so, I'm not sure how you prevent information leakage: Claude might easily
142.
by tibbar 1y ago
Left unsaid is why this practice can be so meaningful. I think it's just that: these are the questions you wish someone else would ask you. When we're stressed, angry, grieving, lost, I think we all yearn to have someone care ab
143.
by tibbar 1y ago
Yeah, so much of this is not about facts as much as judgment: knowing what are the most themes and factors to making a system work well. What are the parts that, if you get right, everything else will follow. Knowing what areas it's ok
144.
by tibbar 1y ago
I think a big difference between being a professional and a beginner is, a beginner will work very hard on the most obvious part of the problem. And might get something reasonable after a while. But the professional knows that the obvious p
145.
by tibbar 1y ago
Experimentation is a toolkit for optimizing things. That's it. For example, if you are Uber, you might offer a bonus to people to work more. How much money do you offer? Is there a relationship between how much you pay and how well peo
146.
by tibbar 1y ago
It's an amazing rant, but he's still going off on tangents: what is worthwhile in life? And how horrible is Oracle? No really, let's just ROAST oracle. Wait that reminds us of Larry Ellison. He really is the worst, have you h
147.
Why are you telling me this?(jacobbrazeal.wordpress.com)
2 points by tibbar 1y ago|0 comments
148.
by tibbar 1y ago
Statsig is big enough that its acquisition is interesting in its own right. Or maybe I'm biased because I spent quite some time setting up a Statsig integration at $FORMER_EMPLOYER. But if I've done that, odds are that a lot of th
149.
by tibbar 1y ago
Nope, it was a live coding interview, less than an hour. I personally did a double-take at the beginning of the interview. However, the long tail of people who can code up algorithms and apps quickly is very long indeed. There are certainly
150.
by tibbar 1y ago
I mean, you don’t hire the best engineers by just blasting off a LinkedIn ad. Or by cold-emailing them. You hire them by already being their friend and offering a massive chunk of cash and equity to work on an interesting project, plus a va

More ›